2010-11-22T00:09:21 *** projectgus has quit IRC 2010-11-22T01:50:12 *** sebhub has joined #rtems 2010-11-22T01:50:22 good morning 2010-11-22T02:54:03 *** bazinski has joined #rtems 2010-11-22T05:22:15 *** kiwichris has joined #rtems 2010-11-22T05:22:36 sebhub, hi 2010-11-22T05:32:43 *** kiwichris has quit IRC 2010-11-22T05:46:30 *** exception13 has joined #rtems 2010-11-22T06:15:41 *** lcpfnyvc has quit IRC 2010-11-22T06:37:53 *** lcpfnyvc has joined #rtems 2010-11-22T06:44:40 *** green` has joined #rtems 2010-11-22T06:45:52 *** antgreen has quit IRC 2010-11-22T06:45:52 *** green` has quit IRC 2010-11-22T06:46:05 *** sebhub has quit IRC 2010-11-22T06:46:41 *** antgreen has joined #rtems 2010-11-22T07:07:15 *** mavu has joined #rtems 2010-11-22T08:07:45 *** DrJoel has joined #rtems 2010-11-22T08:07:45 *** ChanServ sets mode: +o DrJoel 2010-11-22T08:08:13 *** DrJoel changes topic to "Google Code In starts November 22" 2010-11-22T09:10:35 *** exception13 has left #rtems 2010-11-22T09:32:47 *** lekernel__ has joined #rtems 2010-11-22T09:40:44 *** mavu has quit IRC 2010-11-22T09:46:17 hi 2010-11-22T09:46:24 how can I register a new filesystem with rtems? 2010-11-22T09:46:43 I understand I should create a structure "rtems_filesystem_operations_table" with function pointers in it, then how to I make RTEMS care about it? 2010-11-22T09:49:27 lekernel__: hold on a minute.. looking.. Chris changed this recently 2010-11-22T09:50:32 *** bazinski has left #rtems 2010-11-22T09:50:43 (i'm porting yaffs2 btw) 2010-11-22T09:51:59 YEAH! What's the license? 2010-11-22T09:53:16 GPL 2010-11-22T09:53:21 It appears that he added a list of filesystem names to libio.h and some magic to confdefs.h so there is a table rtems_filesystem_table which has all known filesystem types and an initialize routine. This is used by the mount command and mount-mgr.c code 2010-11-22T09:53:33 without linking exception, but this can still be useful as a separate library 2010-11-22T09:53:48 It will be useful to some folks I am sure. :-D 2010-11-22T09:54:28 We can add something to confdefs.h to either directly support it or allow user to add their own filesystem types similar to the driver table "extras" 2010-11-22T09:59:45 i'm for the "extra filesystems" option 2010-11-22T10:00:17 or, isn't there a way to register it by calling some function? the NFS filesystem was supposedly using it (though it didn't work for me) 2010-11-22T10:06:08 I think an extra filesystem option is good but let's ask Chris 2010-11-22T10:07:39 mh there's CONFIGURE_HAS_OWN_FILESYSTEM_TABLE 2010-11-22T10:07:46 I can just use that 2010-11-22T10:08:57 That will work for now .. but is a lot more trouble. 2010-11-22T10:10:32 *** exception13 has joined #rtems 2010-11-22T10:19:27 *** antgreen has quit IRC 2010-11-22T10:47:49 what is mt_entry->mt_fs_root.node_access doing? 2010-11-22T11:07:28 dammit, what is this "node" thing all about ?!??? 2010-11-22T11:11:17 do I need to allocate a "node" for every file and directory? 2010-11-22T11:19:33 lekernel__: ask kiwichris.. he implemented RFS which is a real on-disk filesystem and should be able to help explain things .. you shouldn' 2010-11-22T11:19:54 shouldn't have to do much expect plug your FS hooks into the top level structure 2010-11-22T12:00:09 *** exception13 has left #rtems 2010-11-22T12:43:57 *** antgreen has joined #rtems 2010-11-22T13:53:27 *** Fallenou has joined #rtems 2010-11-22T13:53:29 hi 2010-11-22T13:53:53 is there any google code-in for rtems ? :) 2010-11-22T14:02:48 *** aantn has joined #rtems 2010-11-22T14:03:30 DrJoel: hey, are you Joel Sherrill? 2010-11-22T14:06:18 hi, everyone. I took the Google Code In task on logging information (http://www.google-melange.com/gci/task/show/google/gci2010/rtems/t129018536044) 2010-11-22T14:07:24 my irc connection isn't great at the moment, but I'm looking for feedback. please leave comments on the task's page :) 2010-11-22T14:17:50 oh nice 2010-11-22T14:17:54 you're a student ? or a mentor ? 2010-11-22T14:20:32 aanjhan: yes .. sorry.. my wife got a new netbook and setting it up is distracting me :) 2010-11-22T14:21:00 aantn: I am reading it now.. feedback from anyone in IRC would be appreciated.. 2010-11-22T14:21:13 aanjhan: did you ever talk with Chris about the Capture engine 2010-11-22T14:21:42 aantn: the GUI part is for http://wiki.rtems.org/wiki/index.php/Capture_Engine .. 2010-11-22T14:22:46 aantn: had you been working on this ahead of time? Or are you just this quick? 2010-11-22T14:23:34 DrJoel: ha, thanks. I have some experience with embedded systems (for robotics) so I dug through my notes on the topic 2010-11-22T14:23:44 Which one do you think is likely to be the best? 2010-11-22T14:23:52 DrJoel: I wasn't aware of the capture system - I'm looking at it now 2010-11-22T14:24:02 DrJoel: I don't think you can classify one as the best 2010-11-22T14:24:03 * DrJoel presented on some robotics work last month... 2010-11-22T14:24:07 DrJoel: they all do different things 2010-11-22T14:24:15 DrJoel: cool, what was the topic? 2010-11-22T14:25:45 hmmm.. best has many attributes.. most formats, easiest to use, supports reading dynamic sources.. for dynamic.. say you had an Ethernet link from the board to a PC.. could you visualize live and update as data came in? 2010-11-22T14:26:15 make a template and score one or two 2010-11-22T14:26:35 if you want a follow up task to play with one of the visualization tools, we can create a task for that and get you a log file 2010-11-22T14:28:54 DrJoel: we can score them, but I think that each tool excels at doing something different 2010-11-22T14:29:32 DrJoel: for example, TAU is great profiling, but it's fairly worthless if you want to just graph time-series data 2010-11-22T14:29:41 *great for profiling 2010-11-22T14:29:42 hi 2010-11-22T14:29:46 hey mwalle 2010-11-22T14:30:15 if you were doing a robot and wanted to visualize what was in a log from it, which one would you pick? :D Have an opinion. :D 2010-11-22T14:30:22 aantn: out of curiousity, where are you in general terms geographically? 2010-11-22T14:30:28 DrJoel: israel 2010-11-22T14:31:08 cool! How did you hear about Google Code In? 2010-11-22T14:31:35 OK then for TAU what data would we need to provide. And which would be better for time-series? 2010-11-22T14:31:35 DrJoel: I'm 18 and post-high school (but still eligible), learning in a one-year religious program in israel 2010-11-22T14:32:04 DrJoel: I took part in GHOP, which was a similar contest that google ran a few years ago 2010-11-22T14:32:05 Ahhh.. very cool! I am not the eligibility judge .. just org administrator and mentor for some tasks 2010-11-22T14:32:16 Then you had been waiting for a new program 2010-11-22T14:32:45 DrJoel: I've been involved a lot with GNOME and Zeitgeist (another open source project) since then 2010-11-22T14:33:13 DrJoel: I think I'd use RRDTool for time-series data 2010-11-22T14:33:27 DrJoel: should I reformat the wiki page to list tools by purpose? 2010-11-22T14:34:36 please reformat it 2010-11-22T14:36:39 aantn: if you have two top candidates, we can ask the Capture developer for logs and you can transform/render them 2010-11-22T14:37:01 but we will make that a second/thrid task and he can mentor them. He is in Sydney and not online here yet 2010-11-22T14:37:17 aantn: what's zeitgeist? 2010-11-22T14:37:39 DrJoel: okay, no problem 2010-11-22T14:39:49 it would be really cool to get some visualization for the captured info 2010-11-22T14:40:09 lekernel__: it's a bit outdated, but http://gnomejournal.org/article/70/an-introduction-to-gnome-zeitgeist 2010-11-22T14:40:39 DrJoel: yeah, TAU can actually capture data automatically 2010-11-22T14:41:03 aantn: from socket? what sources? 2010-11-22T14:41:14 brb 2010-11-22T14:49:10 * DrJoel is back 2010-11-22T14:52:19 DrJoel: I'm checking right now. There's a script which automatically instruments your code. I think that the information is logged to file, but I'm checking that now 2010-11-22T14:53:58 remember .. running RTEMS is more like your robots with a single application in Linux terms.. only one process and often no disk. So ideally we would be feeding data by Ethernet back to a "real computer" to do the visualization. Automagic stuff designed for Linux is most likely not going to work the same 2010-11-22T15:10:36 *** errordeveloper has quit IRC 2010-11-22T15:31:50 DrJoel: okay, TAU is interesting, but I'm having trouble getting it to work on my desktop. it _should_ work for anything, but it's primary focus is MPI clusters, so I removed it from the article for now 2010-11-22T15:33:33 aantn: OK. I am sure one is worth using. Can't be worse than GNU plot LOL 2010-11-22T15:34:18 DrJoel: I'm not listing it because it's included in Sage, but pyplot is a nice replacement for GNUplot 2010-11-22T15:34:40 I think I looked at that for some plotting app recently .. 2010-11-22T15:35:26 http://www.rtems.org/ftp/pub/rtems/people/joel/coverage/erc32.html .. all of the time-series plots linked to on the right are generated from the data in the test coverage 2010-11-22T15:38:15 DrJoel: it's pretty cool. last year I wrote a python script that used pyplot to help calibrate sensors and graph the results 2010-11-22T15:38:42 DrJoel: Okay, I think the page is better now. http://www.rtems.com/wiki/index.php/LoggingTools 2010-11-22T15:39:22 DrJoel: I reorganized everything, added RTEMS' own profiling tools, and wrote better some explanations 2010-11-22T15:40:37 DrJoel: I know that you want an out-of-the-box solution for RTEMS, but it's a bit difficult to pinpoint one 2010-11-22T15:42:46 aantn: We hoped for an out of the box solution but are realistic to know that there will be work.. Identify best tools and what needs to be done. :-D 2010-11-22T15:45:03 Simile looks very nice :D 2010-11-22T15:45:29 DrJoel: yeah 2010-11-22T15:45:54 DrJoel: I couldn't help myself. :) I just added on a section with "Ideas for Future RTEMS Improvements" 2010-11-22T15:45:57 This is close to what I wouldn't mind seeing http://www.simile-widgets.org/timeline/ 2010-11-22T15:46:10 DrJoel: that discusses how to better integrate these tools with rtems 2010-11-22T15:47:29 DrJoel: I'm going out to run for a bit. I'll leave IRC open and will be back in about ~45 minutes 2010-11-22T15:47:31 The first idea is very close to what kiwichris has in mind. The data sent from the embedded system to the host needs to be compact but converting it to a standard log format on host for subsequent processing opens a LOT of possibilities 2010-11-22T15:47:35 ok 2010-11-22T15:48:23 you can get back to your wife's netbook :P 2010-11-22T15:50:09 just doing the malware scan on the initial OS image :-D 2010-11-22T15:53:23 *** kiwichris has joined #rtems 2010-11-22T15:59:51 * kiwichris waves to Joel 2010-11-22T16:00:19 kiwichris: hey.. when aantn gets back.. speak with him about logging and the capture engine.. 2010-11-22T16:00:29 Sure. 2010-11-22T16:00:45 Finally got the test script stuff sorted and running now on 4.10 2010-11-22T16:00:55 Has already written this much for capture/logging visualization research task.. http://www.rtems.com/wiki/index.php/LoggingTools .. 2010-11-22T16:01:40 if you have some capture engine logs worth trying to visualize we can add another task for him to try that 2010-11-22T16:01:47 marvelous on test script 2010-11-22T16:01:52 neat 2010-11-22T16:03:07 My test scripts print the results as they are run now. 2010-11-22T16:04:08 http://www.simile-widgets.org/timeline/ looks like a nice web visualization engine... if the capture logs can be converted to input for that, that would be great 2010-11-22T16:04:13 MARVELOUS on test data 2010-11-22T16:05:22 ohhh and goto to each task you might mentor and click the star on the top.. otherwise you won't get updates on it via email 2010-11-22T16:06:07 Ok I will do that today 2010-11-22T16:07:07 we have had some interest today on a few tasks 2010-11-22T16:07:30 That os good. 2010-11-22T16:07:58 I also used a couple of scripts to generate all the moodle and wiki tasks. we have about 120 pending to be published 2010-11-22T16:08:34 That is a lot. 2010-11-22T16:08:55 almost all of them.. the ones left are mostly one of a kind that have to be created by hand 2010-11-22T16:09:18 * DrJoel is about to head home.. file server needs a reboot and I don't feel like hanging around for it 2010-11-22T16:09:22 *** DrJoel has quit IRC 2010-11-22T16:26:49 *** antgreen has left #rtems 2010-11-22T16:28:42 *** projectgus has joined #rtems 2010-11-22T17:07:56 hey, I'm back 2010-11-22T17:08:16 are any of the google code in mentors around? 2010-11-22T17:13:48 Hi 2010-11-22T17:30:48 kiwichris: I need to wait for the mentor to mark a task as completed, right? 2010-11-22T17:31:16 Is this for Google Code In ? 2010-11-22T17:44:55 kiwichris: yes 2010-11-22T17:45:11 kiwichris: I claimed http://www.google-melange.com/gci/task/show/google/gci2010/rtems/t129018536044 2010-11-22T17:53:27 good night everyone :) 2010-11-22T17:56:03 aantn, sorry bout missing this message. I am working in a VM and miss them. 2010-11-22T17:56:15 I will take a look at this today 2010-11-22T18:00:30 *** aantn has quit IRC 2010-11-22T18:31:35 *** aantn has joined #rtems 2010-11-22T18:33:41 kiwichris: do you have Joel's email? I want to contact him so that he'll be able to get back in touch with me after the task is closed 2010-11-22T18:40:23 *** aantn has quit IRC 2010-11-22T20:09:46 *** DrJoel has joined #rtems 2010-11-22T20:10:06 *** DrJoel has joined #rtems 2010-11-22T20:10:06 *** DrJoel has joined #rtems 2010-11-22T20:10:06 *** ChanServ sets mode: +o DrJoel 2010-11-22T20:10:11 DrJoel, hi 2010-11-22T20:10:42 Hi .. trying a new client on my G2 2010-11-22T20:10:48 Must be late 2010-11-22T20:11:13 aantn wants to close a task. He wanted to email you 2010-11-22T20:14:21 Only about 8pm.. I will close the task. He knew the material. Do you have a log from the trace engine? 2010-11-22T20:14:32 No 2010-11-22T20:14:44 He would like to send you am email so he can stay in touch 2010-11-22T20:15:10 If you see him feel free to give it to him 2010-11-22T20:15:27 he will likely be back for gsoc LOL 2010-11-22T20:15:33 :) 2010-11-22T20:18:27 *** DrJoel_ has joined #rtems 2010-11-22T20:20:13 http://www.simile-widgets.org/timeline/ would be a nice representation of the capture or trace kit records .. if you have (or can get) any data sets to work from, we can add a task to transform each into the Timeline format 2010-11-22T20:22:54 http://www.simile-widgets.org/timeplot/ is a nicer way of showing my gnuplots for coverage 2010-11-22T20:23:23 yeah it is nice. Does it take gnu plot format files ? 2010-11-22T20:24:18 I don't think so. So the trick would be to rewrite that part of my shell script to generate this instead 2010-11-22T20:24:30 Ok 2010-11-22T20:24:39 The timeline format would be nice for capture/trace log visualization. 2010-11-22T20:25:22 *** DrJoel has quit IRC 2010-11-22T20:26:23 *** DrJoel_ is now known as DrJoel 2010-11-22T20:26:25 *** DrJoel has joined #rtems 2010-11-22T20:26:25 *** ChanServ sets mode: +o DrJoel 2010-11-22T20:28:06 night 2010-11-22T20:28:10 *** DrJoel has quit IRC 2010-11-22T21:32:05 *** arvind_khadri has joined #rtems 2010-11-22T21:35:22 anyone in here spend much time with MOTLoad on PowerPC? 2010-11-22T21:35:58 I'm wondering if there's a way to get MOTLoad to boot a .ELF image instead of a simple binary .boot, which in turn might (I hope) give function names in stack traces 2010-11-22T21:36:56 I suspect the list maybe better. 2010-11-22T21:37:02 What stack trace data are you after ? 2010-11-22T21:39:53 just symbol names, so function names, from an exception/fault 2010-11-22T21:40:05 I know I can look them up, but it'd be nice just to have them dumped to the console 2010-11-22T21:40:22 I'm going to have a quick experiment myself before I ask the list, just thought I'd put it out there in case someone knew :) 2010-11-22T21:40:25 You need to add a symbol table loader to be able to do this. 2010-11-22T21:40:56 I have one in the RTL code I am working on but it is not in CVS yet 2010-11-22T21:41:12 oh ok, so possibly not supported atm? 2010-11-22T21:41:29 I saw that the build process produced an ELF binary (with a symbol table) and a .boot raw binary file (no symbols) for my platform 2010-11-22T21:41:51 I've just been booting the .boot file up until now, so I was wondering if there was a way to load/boot the ELF, and if so I was hoping that would mean I had a symbol table 2010-11-22T21:41:53 Not currently. A solution many have used is to dump with addresses and then capture and process on a host to map to functions 2010-11-22T21:42:16 yeah, I was being lazy and hoping I didn't have to do that (have set it up on other platforms, but not before on RTEMS) :) 2010-11-22T21:42:27 Typically boot loaders only load relocated, ie static ELF files. 2010-11-22T21:42:52 ok, that makes sense I guess 2010-11-22T21:43:05 In time when the run time loader is available and you arrange to load the base image symbols this should be possible. 2010-11-22T21:43:30 MOTLoad does have some level of ELF support - there's an ELFLoad command that allows you to load an ELF binary and then call functions in it by name, at least according to the documentation 2010-11-22T21:43:32 Are you at the ANU ? 2010-11-22T21:43:43 yes 2010-11-22T21:43:51 They are then using the symbols. Nice 2010-11-22T21:44:01 I am in Sydney. 2010-11-22T21:44:07 oh, cool... I assumed NZ :) 2010-11-22T21:44:19 A long time ago :) 2010-11-22T21:44:38 Can I ask what you are working on > 2010-11-22T21:44:39 I guess that no matter how much ELF support MOTLoad has, it doesn't help me if RTEMS doesn't know there's a symbol table loaded 2010-11-22T21:44:44 *** DrJoel has joined #rtems 2010-11-22T21:44:44 *** DrJoel has joined #rtems 2010-11-22T21:44:44 *** ChanServ sets mode: +o DrJoel 2010-11-22T21:45:02 I work for these guys: http://www.anu.edu.au/CSEM/machines/Accelerator.htm 2010-11-22T21:45:09 Yeap unless you glue them together 2010-11-22T21:45:09 we're refitting our accelerator control system to use EPICS 2010-11-22T21:45:45 Nice. Joel and I visited the LBL in Berkley 2010-11-22T21:45:54 last month 2010-11-22T21:45:55 ACTION 2010-11-22T21:46:08 nice :) 2010-11-22T21:46:23 DrJoel, huh ? 2010-11-22T21:47:38 don't you think it would nice for me to visit to .au 2010-11-22T21:48:00 It would be lovely. Nice wine here and friendly people :) 2010-11-22T21:48:25 kiwichris: I'm guessing your Chris Johns? Do you work on RTEMS full time? 2010-11-22T21:48:34 I must admit to liking NZ wine more 2010-11-22T21:49:03 Yeah that is me and of the last few years yes I have been working full 2010-11-22T21:49:17 nice :) 2010-11-22T21:49:33 DrJoel, both sides of the "ditch" have nice wine. 2010-11-22T21:49:47 DrJoel, no need to pick from just one when both are nice 2010-11-22T21:50:02 LOL agreed 2010-11-22T21:50:24 Some good beer as well but do not tell Ralf that :) 2010-11-22T21:50:52 projectgus, if you need anything just send me an email 2010-11-22T21:52:25 thanks kiwichris. You're welcome to visit if you ever find yourself in Canberra, as well. 2010-11-22T21:52:49 projectgus, I would love to visit. Are you using RTEMS now ? 2010-11-22T21:52:50 *** arvind_khadri has quit IRC 2010-11-22T21:53:06 looks like a nice place to visit.. cool pic on web site 2010-11-22T21:53:18 * kiwichris needs out for school pick up 2010-11-22T21:53:46 kiwichris: nope, entire system running on rtVAX boards & OpenVMS 2010-11-22T21:53:47 bed time here... Justin is working on a code in project :D 2010-11-22T21:54:12 gus.. how old is all that? 2010-11-22T21:54:15 I'm hopefully a week or so from putting in the first piece of hardware running RTEMS+EPICS, for one small piece of LINAC function 2010-11-22T21:54:45 I know all physics experiment sites keep stuff running forever 2010-11-22T21:54:46 DrJoel: some of the VME I/O boards I've written drivers for have been older than me (I'm 27.) 2010-11-22T21:55:00 the rtVAX computers are 25ish years old, I think... 2010-11-22T21:55:12 apparently they got rid of the PDP-11 about 3 years ago :P 2010-11-22T21:55:22 (rtVAX are the VME board computers) 2010-11-22T21:55:42 SLAC has 25+ year old Multibus I boards still in service 2010-11-22T21:55:55 wow 2010-11-22T21:56:13 they work until they break and then you are stuck 2010-11-22T21:56:14 I think in this case the original automation was done in the late 80s, so the oldest pieces of hardware have probably been in service since then 2010-11-22T21:56:36 we have a few like that, where we have one spare and any more will cost $7000 or so (for a 4-channel counter card) 2010-11-22T21:56:56 that is how SLAC is.. some may be newer boards but basically mid 80s vintage 2010-11-22T21:56:56 that's one of the reasons for doing the RTEMS+EPICS rewrite/update, once it's up and running then we'll be replacing the oldest and hairiest pieces of hardware 2010-11-22T21:57:36 the VAXes stay around because some of the professors have old Fortran programs that they've lost the source for, but need to run every now and again 2010-11-22T21:57:39 EPICS+RTEMS is gaining in popularity. Price is right. LOL 2010-11-22T21:57:48 that's it! 2010-11-22T21:57:54 That's funny 2010-11-22T21:58:48 I've had nothing much to complain about so far, although I haven't done much interacting directly with RTEMS - it's all via EPICS abstraction layers, and I've used the POSIX compatibility features for most of the drivers 2010-11-22T21:59:06 chris and I are talking about quarterly or so controlled tool and source releases for sites like you guys. you need control and media 2010-11-22T21:59:06 but yeah, happy with RTEMS - unobtrusive and functional is pretty much ideal :) 2010-11-22T21:59:27 it is posix :D 2010-11-22T21:59:33 well bed time 2010-11-22T21:59:35 *** DrJoel has quit IRC 2010-11-22T22:05:32 *** arvind_khadri has joined #rtems 2010-11-22T22:05:32 *** arvind_khadri has joined #rtems 2010-11-22T22:09:26 *** Omnipotence has joined #rtems 2010-11-22T22:13:41 is anyone here an RTEMS developer? I have some questions 2010-11-22T22:13:47 I am a code-in participant 2010-11-22T22:27:49 Yeap I am 2010-11-22T22:33:51 kiwichris: just getting back to my prev. question about stack traces, do you know if anyone has built a better way to symbol-ify stack traces apart from pasting addresses into gdb's "info symbol"? 2010-11-22T22:33:56 I'm getting my results out just fine doing that 2010-11-22T22:34:28 but I've used scripts where you just feed it a strack and a symbol table (or binary w/ symbol table) and it dumps a "smart" stack dump 2010-11-22T22:34:32 for instance, Android has one 2010-11-22T22:34:36 *** arvind_khadri has quit IRC 2010-11-22T22:34:43 just curious... I might throw one together if it doesn't yet exist 2010-11-22T22:35:18 well, actually, my question got answered on the code-in site 2010-11-22T22:35:22 thanks though 2010-11-22T22:35:48 I have seen them on specific sites. A little perl, python or awk and *-rtems4.11-addr2lin 2010-11-22T22:36:12 Omnipotence, great 2010-11-22T22:36:29 heh, addr2line is not a tool with which I am familiar :) 2010-11-22T22:36:35 I am the one making you guys a new logo 2010-11-22T22:36:41 I see now that I already have it, thanks :) 2010-11-22T22:36:51 The problem with RTEMS is the BSP's tend to have different stack trace formats 2010-11-22T22:37:04 Omnipotence, cool ! 2010-11-22T22:37:31 (I'm fairly new to the intricacies embedded development... am used to just getting a core file to load into gdb so everything is nice and easy!) 2010-11-22T22:38:06 I would love to be able to have RTEMS generate a core file for GDB. 2010-11-22T22:38:34 hehe. I guess it needs to generate it _to_ somewhere though, right? 2010-11-22T22:38:45 You need a more system level view to allow this, eg a way to save a core in an embedded syctem that is not working any more 2010-11-22T22:38:53 at least it looks like I can throw together a ten-line script on the back of addr2line though - at least for my particular BSP. Thanks. 2010-11-22T22:39:00 kiwichris: yeah, exactly :) 2010-11-22T22:40:09 In a systems (the TAAATS voice switch) we saved this data to static RAM, rebooted then sent it to a logging server and it processed it to make a valid trace. We had a couple of hundred boards on line at once 2010-11-22T22:40:44 The point being we never missed a fault and fixed them all 2010-11-22T22:42:43 so I know you guys are coders not designers, but what do you think of this, before I submit it officially? 2010-11-22T22:42:44 http://img192.imageshack.us/img192/15/rtemslogo1.jpg 2010-11-22T22:43:41 Nice. What happens on a white bg ? 2010-11-22T22:45:31 Also the greens on the web site. Will they match up ? Yours is a little brighter than the one I selected on the web site 2010-11-22T22:46:22 kiwichris: nice! actually, for our install I'm not using the flash on the MVME3100 boards at all so I'd be able to do the same thing. Not that I think I can justify the time to implement something like that, though (hopefully we can keep the fatal faults to a minimum!) 2010-11-22T22:47:06 projectgus, are you network booting ? 2010-11-22T22:47:24 yes 2010-11-22T22:48:00 trying to keep the on-system configuration on our boards to zero (boot script is BootP for everything) so they can be easily interchanged 2010-11-22T22:48:10 ah, ok. didn't know it needed to be the same green, thought it was just a general concept of red and green. will fix! the darker green will make it stay legible on the white bg too. 2010-11-22T22:48:14 The method to use is not that hard. Take a piece of fixed static memory, then define a CRC followed by a count then the length 2010-11-22T22:48:53 Omnipotence, it was just a question and something to think about. The web site colours could change 2010-11-22T22:49:26 In the fault handler copy the stack with the saved registers, inc the count and set the CRC 2010-11-22T22:50:00 Once booted and initialised check the CRC and if the count is not 0 send the block of data to a server. 2010-11-22T22:50:12 so you think you could just set aside space in the standard physical RAM? 2010-11-22T22:50:23 I assume when you said static memory that you had a board with some kind of external SRAM 2010-11-22T22:50:23 We used UDP and a single MTU packet, ie < 1500 bytes 2010-11-22T22:50:53 If it is DRAM I suspect not because a controller reset could effect the memory 2010-11-22T22:51:05 yeah that's what I thought 2010-11-22T22:51:14 Most VME boards have a little SRAM in the clock or something like that 2010-11-22T22:51:19 I could still work out how to talk to the flash controller and get it to do it, more complex though 2010-11-22T22:51:28 Too slow. 2010-11-22T22:51:33 Also wear out 2010-11-22T22:51:48 not sure if the MVME boards have that... there are a couple of serial EEPROMs, though 2010-11-22T22:51:58 This worked so well we ended up feeding warnings etc to the log 2010-11-22T22:52:05 surely not too slow or too high duty cycle for just writing a stack trace every now and again? 2010-11-22T22:52:22 That is confidence. 2010-11-22T22:52:24 :) 2010-11-22T22:52:31 fair enough... :) 2010-11-22T22:53:28 it's a very cool idea, but I honestly hope we don't hit enough faults to make it necessary for me to go through with it :) 2010-11-22T22:53:46 *** antgreen has joined #rtems 2010-11-22T22:53:50 When hardware starts to fail it can help 2010-11-22T22:54:08 we were also discussing buying serial<->ethernet adapters to put on all the serial ports, as they're pretty cheap these days 2010-11-22T22:54:13 Does the RTC have any RAM ? 2010-11-22T22:54:23 then the central server can just log everything that comes off the main console on all the machines 2010-11-22T22:54:28 implement filtering for stack traces 2010-11-22T22:54:32 it's the lazy sysadmin way ;) 2010-11-22T22:55:19 We should talk more. I did a system that buffered all the console data, sort of a tee and then sent it to a central server or let it be viewed by a local web server 2010-11-22T22:55:53 It is not a bad idea and yes they are cheap these days. 2010-11-22T22:56:47 datasheet says the onboard RTC is a DS1375, so 16 bytes of SRAM :) 2010-11-22T22:57:07 Hmm a bit too small. Oh well. 2010-11-22T22:57:14 my previous job was coding for hosted VoIP systems, so we had lots of those kinds of distributed logging/monitoring systems 2010-11-22T22:57:33 mostly implemented by the sysadmins using third-party tools, but sometimes they'd go for something hand-coded 2010-11-22T22:57:46 I am sure. I suppose it depends on the number of ports to monitor and the sapce. 2010-11-22T22:57:55 yeah 2010-11-22T22:58:05 Must head out. Send me an email if you need anything more. 2010-11-22T22:58:09 for us here, the conclusion we came to on the serial<->ethernets is that hopefully the system will be stable enough to not need regular monitoring 2010-11-22T22:58:15 but if that doesn't happen, we'll probably build something with them 2010-11-22T22:58:19 ok, see you :) 2010-11-22T22:58:23 I should do some actual work ;) 2010-11-22T22:58:27 hee 2010-11-22T22:58:55 "need regular monitoring" -> "need regular monitoring at the serial console level" 2010-11-22T22:59:05 it will obviously need regular monitoring at the EPICS/application layer level 2010-11-22T23:15:44 man, you guys inspire me to become a better coder :D 2010-11-22T23:17:04 projectgus, does EPICS define any system monitoring ? I have seen SLAC and LBL working but know little about EIPCS 2010-11-22T23:18:31 there's a bunch of things built in, like EPICS records go to alarm states when they see unusual or out-of-band values (or communication failures) 2010-11-22T23:18:43 then EPICS has its own client applications for alarm logging and notification 2010-11-22T23:19:01 you can also just do general logging/charting of any parameters in your system 2010-11-22T23:19:24 I'm as new to EPICS as I am to RTEMS, and our system isn't up and running yet, so I'm sure there is more that I'll learn about after we roll it out :) 2010-11-22T23:20:08 Met Eric Norum last month after working with on RTEMS for many many years. 2010-11-22T23:20:23 When do you roll out ? 2010-11-22T23:20:55 the first piece should be going in over the next couple of weeks, but it's only for one small system 2010-11-22T23:21:05 NIce. 2010-11-22T23:21:17 Xmas dead line :) 2010-11-22T23:21:18 we'll be migrating bit by bit to EPICS slowly over the next year or two :) 2010-11-22T23:22:34 *** Omnipotence has quit IRC 2010-11-22T23:47:27 *** arvind_khadri has joined #rtems 2010-11-23T00:57:23 *** arvind_khadri has quit IRC 2010-11-23T01:35:48 *** sebhub has joined #rtems 2010-11-23T01:36:09 good morning 2010-11-23T01:39:54 morning 2010-11-23T02:36:14 *** arvind_khadri has joined #rtems 2010-11-23T04:23:03 *** lekernel__ has quit IRC 2010-11-23T04:23:17 *** lekernel__ has joined #rtems 2010-11-23T05:57:27 *** exception13 has joined #rtems 2010-11-23T06:14:53 *** arvind_khadri has quit IRC 2010-11-23T06:57:23 *** lcpfnyvc has quit IRC 2010-11-23T07:59:33 *** aantn has joined #rtems 2010-11-23T08:08:50 *** aantn has quit IRC 2010-11-23T08:09:51 *** DrJoel has joined #rtems 2010-11-23T08:09:51 *** ChanServ sets mode: +o DrJoel 2010-11-23T08:36:54 joel? 2010-11-23T08:43:41 *** aantn has joined #rtems 2010-11-23T08:52:52 sebhub? 2010-11-23T08:53:18 I just saw your email 2010-11-23T08:54:18 *** aantn has quit IRC 2010-11-23T08:54:34 you are right. Is that on the 4.10 branch also? 2010-11-23T08:54:42 see the email 2010-11-23T08:56:39 patch coming your way now 2010-11-23T08:57:27 this only impacts the head so if that fixes it, I will commit it. 2010-11-23T09:02:47 sorry my fix did introduce dead code if POSIX is enabled 2010-11-23T09:03:03 i change it to your #if !defined(POSIX) 2010-11-23T09:03:29 this is quite brittle if the apis change some time 2010-11-23T09:03:54 yes it is a bit brittle but API additions/deletions are rare 2010-11-23T09:04:55 what is the difference between '#if !defined()' and '#ifndef'? 2010-11-23T09:05:05 nothing 2010-11-23T09:06:02 it occured several times that someone changed #ifndef into #if !defined 2010-11-23T09:08:47 I generally prefer !defined for no particular reason .. so it may have been me... can't say 2010-11-23T09:08:58 did that fix it for you? If so, I will commit it. 2010-11-23T09:09:57 i need to fix my fix now ;-) 2010-11-23T09:10:08 kiwichris: hi 2010-11-23T09:10:43 #if !defined(RTEMS_POSIX_API) || defined(RTEMS_DEBUG) 2010-11-23T09:10:49 i will use this 2010-11-23T09:12:06 That is why I like !defined better .. when you need two conditions it must be used 2010-11-23T09:12:06 ok.. I will commit that 2010-11-23T09:12:33 i am about to commit it now 2010-11-23T09:12:43 apparently you beat me LOL 2010-11-23T09:13:33 * lekernel__ finds the RTEMS filesystem API quite horrible, to be frank 2010-11-23T09:13:56 *** aantn has joined #rtems 2010-11-23T09:14:10 it is not as horrible as it used to be ;-) 2010-11-23T09:15:03 but apart from the API horribility it has several flaws with regard to mount/unmount 2010-11-23T09:15:22 sebhub: do you have to create a "node" (whatever that means) for every file or directory accessed? 2010-11-23T09:15:45 *** exception13 has left #rtems 2010-11-23T09:15:48 no the node producer is evaluate for make or path evaluate 2010-11-23T09:15:54 what should I put in mt_entry->mt_fs_root.node_access in the initialization routine? 2010-11-23T09:15:58 the consumer is node free 2010-11-23T09:16:19 i'm implementing yaffs2, so I guess i'm the "node producer" 2010-11-23T09:17:00 yes, your yaffs2 evalformake_h and evalpath_h produce the node 2010-11-23T09:17:09 do I have to? 2010-11-23T09:17:34 you can use this node to store all your data that belong to it 2010-11-23T09:17:51 is likely some yaffs2 handle 2010-11-23T09:18:01 what should I put in that node? information so that subsequent calls regarding that node will affect the file that was referenced in evalformake/evalpath? 2010-11-23T09:18:10 yes 2010-11-23T09:18:13 ok 2010-11-23T09:18:16 how are nodes freed? 2010-11-23T09:18:21 * DrJoel is going awk 2010-11-23T09:18:34 freenod_h 2010-11-23T09:18:46 aah, ok :) 2010-11-23T09:19:10 ok I think I figured that out 2010-11-23T09:19:11 thanks 2010-11-23T09:19:39 another question, in mt_entry->mt_fs_root.handlers I set a structure containing callbacks that deal with directories 2010-11-23T09:19:43 how do I deal with files? 2010-11-23T09:19:56 http://www.rtems.org/onlinedocs/doxygen/cpukit/html/struct__rtems__filesystem__operations__table.html 2010-11-23T09:20:13 this are file system operations 2010-11-23T09:20:23 the create and manage nodes 2010-11-23T09:20:39 http://www.rtems.org/onlinedocs/doxygen/cpukit/html/struct__rtems__filesystem__file__handlers__r.html 2010-11-23T09:20:45 but the node is an opaque pointer to RTEMS, no? 2010-11-23T09:20:46 these handlers work with nodes 2010-11-23T09:21:28 http://www.rtems.org/onlinedocs/doxygen/cpukit/html/structrtems__filesystem__location__info__tt.html 2010-11-23T09:21:48 node_access and node_access_2 may be used by your yaffs2 layer 2010-11-23T09:22:12 please avoid node_access_2, i think this should go away 2010-11-23T09:22:39 did you consider UFFS? as a flash file system? 2010-11-23T09:23:33 YAFFS has relatively good documentation and portable/clean code 2010-11-23T09:24:02 the problem is that you need a commercial license for commercial products 2010-11-23T09:24:17 no, you need a commercial license for proprietary products 2010-11-23T09:24:18 *** aantn has quit IRC 2010-11-23T09:24:40 ok, you have to release your product under GPL 2010-11-23T09:24:54 yes, that's what I'm doing :) 2010-11-23T09:25:14 ok, in this case yaffs2 should be perfect 2010-11-23T09:25:23 and quite frankly I hate doing this flash filesystem stuff but it needs to be done 2010-11-23T09:25:41 so I want to do it as fast as possible, and YAFFS seems to be the solution that causes the least technical problems 2010-11-23T09:25:47 we need sometime in the future also a flashfile system 2010-11-23T09:30:23 sebhub: in the filesystems I've seen, mt_entry->mt_fs_root.handlers is set to functions that operate on *directories* 2010-11-23T09:30:42 what when I want to access a regular file? 2010-11-23T09:31:27 (I mean: set in the "initialize/fsmount_me_h" function) 2010-11-23T09:32:00 typedef int(* rtems_filesystem_evalpath_t)(const char *pathname,size_t pathnamelen,int flags,rtems_filesystem_location_info_t *pathloc) 2010-11-23T09:32:03 and why have several rtems_filesystem_file_handlers_r tables? one would suffice, no? 2010-11-23T09:33:01 this function will present you a pathloc, which should be initialized with your node_access and handlers 2010-11-23T09:33:03 mh ok 2010-11-23T09:33:15 what an infernal mess :) 2010-11-23T09:33:40 actually it is pretty straight forward, but not documented 2010-11-23T09:34:16 ok, I think I have all the bits now... 2010-11-23T09:34:18 thanks 2010-11-23T09:34:37 the mount_me will set the handlers for the "root directory" this is way it is normally a directory handler table 2010-11-23T09:34:50 yup 2010-11-23T09:35:18 you can send documentation improvements if you have the time 2010-11-23T09:38:37 can I set some pointers in rtems_filesystem_operations_table to NULL? 2010-11-23T09:38:42 like mount_h for example... 2010-11-23T09:39:29 Check this out http://img401.imageshack.us/i/rtemswhitebg.jpg/ 2010-11-23T09:40:01 NO, they must not be NULL 2010-11-23T09:40:08 phew 2010-11-23T09:40:12 there exists default handlers 2010-11-23T09:40:28 all I want is a normal mount 2010-11-23T09:40:40 http://www.rtems.org/onlinedocs/doxygen/cpukit/html/group__LibIO.html 2010-11-23T09:40:47 rtems_filesystem_default_* 2010-11-23T09:42:47 The default implementation returns -1 and sets errno to ENOTSUP. 2010-11-23T09:42:50 oh, yeah... 2010-11-23T09:43:09 no "generic" mount... 2010-11-23T09:43:39 anyway I won't need it, so I won't implement it. people wanting to re-mount filesystem on top of YAFFS will do that themselves. open source rules :) 2010-11-23T09:44:38 a mount is likely not generic, it depends on the file system 2010-11-23T09:45:50 what a mount does is replace the contents of some directory with that from another filesystem 2010-11-23T09:46:03 I've never seen anything behaving in a another way 2010-11-23T09:46:40 why not factorize the path resolution code? 2010-11-23T09:46:48 this will make filesystems a lot easier to write 2010-11-23T09:46:53 yes, it is a mess 2010-11-23T09:47:23 there is a lot of copy and paste in the file system area 2010-11-23T09:47:29 yeah, I see that :( 2010-11-23T09:49:12 the cut and paste has gotten better but if you can see a way to refactor more, it is definitely appreciated, 2010-11-23T09:49:46 the path evaluation is horrible, but it needs some time to clean this up 2010-11-23T09:50:14 for me it has no priority to clean this up 2010-11-23T10:07:35 *** exception13 has joined #rtems 2010-11-23T10:07:51 *** gedare has joined #rtems 2010-11-23T10:16:48 *** arvind_khadri has joined #rtems 2010-11-23T10:19:22 *** madrazr has joined #rtems 2010-11-23T10:19:30 aanjhan: :) 2010-11-23T10:20:11 sebhub: would that make sense for a GSOC project? 2010-11-23T10:21:45 sebhub: why is there a rtems_filesystem_default_evalformake while http://www.rtems.org/onlinedocs/doxygen/cpukit/html/struct__rtems__filesystem__operations__table.html says "There is no default implementation." 2010-11-23T10:23:28 maybe a documentation bug 2010-11-23T10:23:30 it only exists in http://www.rtems.org/onlinedocs/doxygen/cpukit/html/group__LibIO.html, and not defined in my code... 2010-11-23T10:24:46 also, how come there is a default implementation for freenode? 2010-11-23T10:25:04 since the node contents is filesystem dependent? 2010-11-23T10:25:33 i don't know 2010-11-23T10:25:47 i guess someone simply added functions for each handler 2010-11-23T10:26:02 mh 2010-11-23T10:26:13 seems this filesystem stuff could definitely use some clean-up 2010-11-23T10:27:54 one problem is that only few people understand the file system 2010-11-23T10:28:41 it needs also a redesign to support the unmount 2010-11-23T10:28:45 sebhub: agreed 2010-11-23T10:28:55 and factorize the path resolution? 2010-11-23T10:29:20 it would be nice to support forced unmount too (which is easier once the path resolution is factored) 2010-11-23T10:29:20 the default handlers were only added recently to eliminate null tests and it is quite possible it was done mechanically. So some may not needed 2010-11-23T10:29:24 so USB sticks can be supported 2010-11-23T10:29:34 * DrJoel agrees the patch resolution refactorizatino 2010-11-23T10:29:53 from my point of view unmount is not supported at all 2010-11-23T10:30:46 sebhub: hm?! 2010-11-23T10:30:53 the basic problem is that nobody knows the node users 2010-11-23T10:31:00 [/] # ls /flash 2010-11-23T10:31:01 A Million Miles from Earth.fnp Ice Drops.fnp 2010-11-23T10:31:03 (...) 2010-11-23T10:31:10 [/] # unmount /flash 2010-11-23T10:31:11 unmounted /flash 2010-11-23T10:31:12 [/] # ls /flash 2010-11-23T10:31:14 [/] # 2010-11-23T10:31:19 seems to work for me 2010-11-23T10:31:25 (it's using dosfs atm) 2010-11-23T10:31:26 you can unmount file systems that are not in use 2010-11-23T10:31:50 but there is currently no way to determine if a file system is in use 2010-11-23T10:32:14 and unmounting a file system in use leads likely to a corrupt system (heap corruption) 2010-11-23T10:33:17 argh, ok 2010-11-23T10:33:27 so USB sticks won't really work before a while I guess :( 2010-11-23T10:33:54 the work pretty well if you use them right 2010-11-23T10:34:10 I'm doing a product for people who won't use them right 2010-11-23T10:34:14 you can use them dynamically 2010-11-23T10:34:39 it is very easy to do it wrong and this will lead to a useless system state 2010-11-23T10:34:59 or deadlock or starvation 2010-11-23T10:36:08 *** aantn has joined #rtems 2010-11-23T10:43:36 *** aantn has quit IRC 2010-11-23T10:46:24 bye 2010-11-23T10:46:54 *** sebhub has quit IRC 2010-11-23T10:50:41 *** methril_work has joined #rtems 2010-11-23T11:16:08 *** arvind_khadri has quit IRC 2010-11-23T11:16:58 *** arvind_khadri has joined #rtems 2010-11-23T11:30:39 what is evalformake? 2010-11-23T11:31:12 evaluate for create purposes versus an existing file 2010-11-23T11:34:46 'pathloc' is returned with a pointer to the parent of the new node. 2010-11-23T11:34:59 pathloc - IN/OUT (start point for evaluation/parent directory for creation) 2010-11-23T11:35:07 as I understand it, I must not modify pathloc 2010-11-23T11:35:45 *** aantn has joined #rtems 2010-11-23T11:36:32 but the msdos filesystem does modify it, so I haven't understood it correctly 2010-11-23T11:36:55 wow... what an amazing mess :(((( 2010-11-23T11:38:24 *** aantn has quit IRC 2010-11-23T11:38:45 I guess I'll start with a read-only YAFFS filesystem and keep this nicety for later 2010-11-23T11:39:33 *** aantn has joined #rtems 2010-11-23T11:41:53 *** aantn has quit IRC 2010-11-23T11:42:30 *** aantn has joined #rtems 2010-11-23T11:57:38 *** aantn has quit IRC 2010-11-23T12:44:44 *** madrazr has left #rtems 2010-11-23T13:46:05 WTF?!? rtems calls my eval_path with pathloc->mt_entry == NULL 2010-11-23T13:46:45 *** arvind_khadri has quit IRC 2010-11-23T13:46:48 wah... the filesystem infrastructure is definitely not a good point of rtems 2010-11-23T14:33:40 *** gedare has quit IRC 2010-11-23T14:40:24 *** DrJoel has quit IRC 2010-11-23T16:07:06 with my filesystem mounted on /flash, ls /flash calls my "dir_open" callback with "/flash" 2010-11-23T16:07:08 hello 2010-11-23T16:07:14 and with //flash, it calls it with "//flash" 2010-11-23T16:07:20 niiiice :((( 2010-11-23T16:07:51 lekernel__, you may have to step back a little and tell what you are doing so I can help 2010-11-23T16:08:47 i'm just figuring out I have to implement all path decoding and mount point recognition in my filesystem 2010-11-23T16:08:47 lekernel__, I see you are having to explore the eval path handler. 2010-11-23T16:09:16 lekernel__, are you using an existing one as an example to follow ? 2010-11-23T16:10:02 yeah, i'm looking at dosfs/imfs 2010-11-23T16:10:27 can a different implementation of the eval path handler clean up strings that are later passed to the open_dir callback? 2010-11-23T16:10:51 Do you mean malloc in one and clean up in another ? 2010-11-23T16:12:05 my problem is that RTEMS seems to pass whatever path comes from userspace to the dir_open callback 2010-11-23T16:12:13 for example, if your filesystem is mounted on /blah 2010-11-23T16:12:17 Yes that is correct 2010-11-23T16:12:24 and if the user does ls /toto/../blah 2010-11-23T16:12:28 This is done on purpose. 2010-11-23T16:12:36 then you get "/toto/../blah" passed to your dir_open callback 2010-11-23T16:12:45 We have users where even the overhead of a malloc is too much 2010-11-23T16:12:48 and you have to duplicate path parsing code everywhere 2010-11-23T16:13:02 and C is very nice for string manipulation btw 2010-11-23T16:13:11 very time efficient 2010-11-23T16:13:27 That is correct and it is only eval handlers that suffer from this. Maybe it will get cleaned up maybe not. 2010-11-23T16:13:48 :( 2010-11-23T16:13:55 I am missing your point about C and string parsing. 2010-11-23T16:14:29 well, string parsing in C wastes a lot of programmer time and creates opportunities for various buffer overflows and wild pointer problems 2010-11-23T16:14:40 There are reasons for the code being like this. Some are historic and some relate to the upper layers 2010-11-23T16:14:50 so better use as less as possible 2010-11-23T16:15:58 No doubt. If you want to abstract the eval handler code in RTEMS and clean up I can give you some pointers then send me the patch and I will either approve it and commit or ask for changes 2010-11-23T16:16:27 May I suggest you take the RFS handlers are the starting point. 2010-11-23T16:16:39 They are in rfs/rtems-rfs-rtems.c 2010-11-23T16:19:47 I see you have a flash path. Are you porting or writing a flash file system ? 2010-11-23T16:34:25 kiwichris: he's porting yaffs2 2010-11-23T16:34:51 Ah ok. Nice. 2010-11-23T16:38:07 lekernel__, if need any help please ask. As for eval path, unfortunately it is what it is and it will stay that way until someone cleans it up. Eval path is painful 2010-11-23T18:42:16 *** exception13 has left #rtems 2010-11-23T19:14:08 lekernel_: I had some similar feelings/thoughts when I was looking at specific cases in tftpfs a while ago, and again last week 2010-11-23T19:14:59 it seemed to me in the peanut gallery that it would be better ifall the path munging was in the generic/vfs layer rather than in each specific filesystem 2010-11-23T19:15:31 It would be nice but this means opening up all file systems and fixing them which is difficult. 2010-11-23T19:15:44 I appreciate that, hence "peanut gallery" :) 2010-11-23T19:15:58 I have looked into this before and there are other complicating factors. 2010-11-23T19:16:10 I don't have the spare time and inclination to go and rewrite every filesystem in RTEMS, and I don't expect anyone else to either :) 2010-11-23T19:16:12 For example the atomic nature of a look up. 2010-11-23T19:16:46 I would be happy to fix the known issues but it requires users to sponsor the work 2010-11-23T19:17:05 Quite fair enough :) 2010-11-23T19:17:13 I tend to work on other things with my spare time. 2010-11-23T19:17:19 in RTEMS that is 2010-11-23T19:17:23 Also quite fair enough 2010-11-23T19:17:46 bbl lunchtime :) 2010-11-23T19:18:05 For example I am excited by this http://www.rtems.com/wiki/index.php/LoggingTools 2010-11-23T19:18:18 Hmm yes. Thanks for the heads up. 2010-11-23T19:19:15 I also think a single function or part of RTEMS's file system, eval path, should colour the view of all of RTEMS's file system 2010-11-23T19:20:10 The main issue to sort out in the file system is locking. This way the upper layer can break up the calls. 2010-11-23T20:35:44 *** arvind_khadri has joined #rtems 2010-11-23T20:35:44 *** arvind_khadri has joined #rtems 2010-11-23T20:40:08 *** Omnipotence has joined #rtems 2010-11-23T20:40:34 Omnipotence, hi 2010-11-23T20:41:59 hey 2010-11-23T20:42:17 me again, keep hoping to catch DrJoel :P 2010-11-23T20:42:39 Omnipotence, his time zone will have him tucked up in bed 2010-11-23T20:43:02 The rules of the code-in won't let me accept another task till he officially approves my logo design 2010-11-23T20:43:06 ah, ok 2010-11-23T20:43:20 Ok. 2010-11-23T20:43:52 *** Omnipotence has quit IRC 2010-11-23T20:43:53 His email is joel.sherrill@gmail.com. Send him an email 2010-11-23T20:50:26 kiwichris: at 9pm? 2010-11-23T20:50:41 he's more frail than i would have expected :) 2010-11-23T20:50:44 <-amar btw from gsoc. 2010-11-23T20:50:52 verm__, I knew I would get pinged :D 2010-11-23T20:51:11 hehe 2010-11-23T20:51:12 I suspect he is helping with home work 2010-11-23T20:51:12 hows things? 2010-11-23T20:51:20 Yeah good. 2010-11-23T20:51:22 yeah.. i've been trying to catch him for almost two weeks now 2010-11-23T20:51:33 i figure eventually we'll synch up as my sleep schedual gradually drifts 2010-11-23T20:51:35 Email him then arrange a time. 2010-11-23T20:52:36 He was on late yesterday playing with his new Droid's IRC client 2010-11-23T20:52:38 yeah suppose i should, i wanted to get some requirements together for what the rtems build requirements are 2010-11-23T20:52:44 heh nice 2010-11-23T20:53:05 What do you mean by build requirements ? 2010-11-23T20:55:04 well i know it's cross-built 2010-11-23T20:55:24 but there's also operational testing and regress/unit.. we talked a lot about how it worked and what it does 2010-11-23T20:55:32 but none about how they're actually executed.. ie where and how 2010-11-23T20:56:07 i've been working on my qa/continous integration system.. i was hoping to get some details so i can ensure what i'm writing isn't too locked into ntps' requirements 2010-11-23T20:56:12 RTEMS build on a range of host operating systems including Linux, Windows, MacOS, Solaris etc 2010-11-23T20:56:26 Ah ok. 2010-11-23T20:56:34 right but do you nessicarily need to test building on all those? 2010-11-23T20:56:44 if so it's fine 2010-11-23T20:56:45 No really 2010-11-23T20:56:51 yeah i figured.. 2010-11-23T20:57:00 so once you have your cross-built blobs then what? 2010-11-23T20:57:01 We support Linux as the main development platform. 2010-11-23T20:57:18 Then we test RTEMS with simulators 2010-11-23T20:57:34 hardware-based i'm assuming? 2010-11-23T20:57:41 For example RTEMS runs on a SPARC simulator that is in GDB. Same with PowerPC 2010-11-23T20:57:43 No software 2010-11-23T20:57:52 We also run on things like qemu 2010-11-23T20:57:53 oh really? i figured you'd have some of those really cool hardware sim devices 2010-11-23T20:58:25 Not really. We find they tend to make the testing exclusive rather than inclusive 2010-11-23T20:58:29 that's interesting 2010-11-23T20:58:54 The idea for all user is that the code you get for a specific release is the same for all users 2010-11-23T20:59:11 that makes sense 2010-11-23T20:59:26 There is not special hidden "really really safe" version that you need to join a special club or pay lots for 2010-11-23T21:00:01 Therefore we want all users to be able to run the same tests we run that we use to state the performance of RTEMS. 2010-11-23T21:00:16 yeah i figured as much.. but i was still thinking you'd be testing using hardware sims for various situations, if anything to catch edge cases 2010-11-23T21:00:33 For example if we state our testing of a specific part of RTEMS is 99% coverage then we would the user to be able to check this 2010-11-23T21:00:42 right 2010-11-23T21:01:04 do you have internal unit tests? 2010-11-23T21:01:12 The simulators all us to capture every executed instruction 2010-11-23T21:01:13 ie, tests that run inside rtems to do environmental testing? 2010-11-23T21:01:19 Yes 2010-11-23T21:01:36 that's good, so this isn't going to be too different than what ntp needs 2010-11-23T21:01:54 We run these and capture the instructions then we look at them and say we have tested every instruction 2010-11-23T21:02:01 Great 2010-11-23T21:02:21 it'll still need some custome setup but from what you've told me it sounds fairly trivial.. i'll confirm with joel as well 2010-11-23T21:02:58 kiwichris: yeah that's great but there's also dangers in doing it that way too which i expressed previously 2010-11-23T21:03:02 If you have a Linux host, eg Fedora it is install the yum RPM then yum down the packages for the tools 2010-11-23T21:03:18 hrm i have two vms one that's debian and one that's ubuntu now 2010-11-23T21:03:29 Ubunta is fine 2010-11-23T21:03:36 yeah? cool i'll use that then 2010-11-23T21:03:43 You get the RPMs and rpm2cpio the contents 2010-11-23T21:03:52 You need the CentOS rpms 2010-11-23T21:04:04 Coverage is http://www.rtems.org/ftp/pub/rtems/people/joel/coverage/ 2010-11-23T21:04:31 heh that's pretty neat 2010-11-23T21:04:58 i like it (a lot) when people take the effort to make it look nice 2010-11-23T21:05:36 This as well http://www.rtems.org/ftp/pub/rtems/people/chrisj/pq/html/ 2010-11-23T21:05:36 this is very cool 2010-11-23T21:05:56 We need this because we tend not to build for all possible combination when committing 2010-11-23T21:06:17 I see things are a bit broken. 2010-11-23T21:06:20 right there's no real need usually 2010-11-23T21:06:42 We have a possible 122 different board support packages 2010-11-23T21:07:03 Across 14 architectures and changes are ok on one but not others 2010-11-23T21:07:37 yeah i know how that goes 2010-11-23T21:08:00 Oh the ARM builds are failing because the memory is too full 2010-11-23T21:08:05 I have been looking at autotest. 2010-11-23T21:08:31 The idea is a simple way to do "make test" 2010-11-23T21:09:35 autotest is ok.. it depends on what you're use-case is 2010-11-23T21:09:42 are you trying to do test-driven development? 2010-11-23T21:09:58 otherwise it's massivly overkill 2010-11-23T21:09:59 Not sure what that means 2010-11-23T21:10:12 test-driven development is when you write your unit test _first_ 2010-11-23T21:10:15 then write your function 2010-11-23T21:10:36 it does two things: forces you to think about what you're really trying to do 2010-11-23T21:10:48 Sometimes. It is more a means to know we are not regression and we also have full coverage 2010-11-23T21:10:53 and ensures you've got a test that actually tests for what it's supposed to (it has to by nature since it was written first) 2010-11-23T21:11:14 if you're just looking for a simple target to run at the end of building i would not use autotest 2010-11-23T21:11:28 I tend to develop tests as I am writing the code to make sure. A sort of unit test 2010-11-23T21:11:39 then you want autotest :) 2010-11-23T21:11:42 autotest is intelligent 2010-11-23T21:11:49 since it'll only re-run tests for the code you've changed 2010-11-23T21:12:00 We need to be able to say the test runs on this BSP or not 2010-11-23T21:12:17 NOw that is a nice feature. 2010-11-23T21:12:37 yeah it is 2010-11-23T21:13:06 kiwichris: you can use buildbot for that 2010-11-23T21:13:06 I also want to know if something is added to the code base that changes the fail count 2010-11-23T21:13:33 i could help you guys set that up easily 2010-11-23T21:13:38 would probably be extremely handy 2010-11-23T21:13:40 Yes I agree. 2010-11-23T21:13:45 I agree 2010-11-23T21:14:00 and we don't have to throw away any of what you've already done just integrate it 2010-11-23T21:14:02 I have been working of late to get the various test scripts into some sort of order 2010-11-23T21:14:10 NIce 2010-11-23T21:14:10 ie, buildbot triggers and handles the builds while your reporting mechs do the reporting 2010-11-23T21:14:26 I am sold. 2010-11-23T21:14:26 kiwichris: we can do that inside buildbot too 2010-11-23T21:14:35 it can handle the order in which to run the tests 2010-11-23T21:14:43 I am rather busy until the new year. 2010-11-23T21:14:52 Can we plan for post new year to do this ? 2010-11-23T21:14:54 same pretty much 2010-11-23T21:15:20 well, hmm where would you set this up? 2010-11-23T21:15:33 In OAR Corp machines. 2010-11-23T21:15:49 We use the GSoC money to buy build machines to support the project 2010-11-23T21:16:04 We have a new 8 core machine coming 2010-11-23T21:16:19 even if we start after the holiday it would be nice to get a slowish or spare machine or jail/vm somewhere so i can setup buildbot and play around 2010-11-23T21:16:28 i don't have much time 2010-11-23T21:16:38 but i can certainly find time in the intervening 6-8 weeks 2010-11-23T21:16:53 We have a couple of those machines available. The main server can host buildbot and we can run the builds on remote machines 2010-11-23T21:17:05 that way when you're free i won't have to worry about annoying bits like setup and familiarising myself with rtems 2010-11-23T21:17:07 I am out until the new year. 2010-11-23T21:17:25 yeah that's fine, i'm good with finding my own way around :) 2010-11-23T21:17:27 Cool. Please send Joel and email and feel free to cc me 2010-11-23T21:17:52 i'd just rather get an early start beacuse i'll get bored on ocassion and feel like working on something else.. you know how that goes 2010-11-23T21:18:31 yep i will 2010-11-23T21:18:39 that same gmail address? 2010-11-23T21:19:03 chrisj@rtems.org 2010-11-23T21:19:11 what about joels? 2010-11-23T21:19:47 joel.sherrill@OARcorp.com 2010-11-23T21:20:07 got it 2010-11-23T21:20:37 setting this up should be really easy 2010-11-23T21:20:45 where you off to for the holidays, booking off time to spend with the family? 2010-11-23T21:21:05 No just at home this year. Family heading over to us. 2010-11-23T21:21:41 ah, nice. 2010-11-23T21:55:50 *** methril_ has quit IRC 2010-11-23T22:09:33 *** methril_ has joined #rtems 2010-11-23T22:15:38 *** arvind_khadri has quit IRC 2010-11-23T22:36:50 *** christopher_ has joined #rtems 2010-11-24T00:11:20 *** projectgus has quit IRC 2010-11-24T00:46:36 *** lcpfnnvcy has joined #rtems 2010-11-24T00:47:58 morning 2010-11-24T03:05:22 *** aanjhan has quit IRC 2010-11-24T03:05:27 *** aanjhan has joined #rtems 2010-11-24T03:11:23 *** lekernel__ has quit IRC 2010-11-24T03:39:09 *** sebhub has joined #rtems 2010-11-24T03:39:15 good morning 2010-11-24T04:33:45 *** lekernel__ has joined #rtems 2010-11-24T05:48:18 *** exception13 has joined #rtems 2010-11-24T06:27:41 *** lekernel has joined #rtems 2010-11-24T06:31:23 *** lekernel__ has quit IRC 2010-11-24T06:34:26 *** christopher_ has quit IRC 2010-11-24T06:38:29 kiwichris: can I set the dir_open and dir_close callbacks to NULL, if opening the directory always succeeds if the node was successfully created before? 2010-11-24T06:39:05 apparently, node_type is called before, so I shouldn't need to check that I'm really dealing with a directory there 2010-11-24T07:29:39 ok, never mind, it turns out i'll need to allocate memory in dir_open 2010-11-24T07:43:10 *** DrJoel has joined #rtems 2010-11-24T07:43:10 *** ChanServ sets mode: +o DrJoel 2010-11-24T08:28:57 *** antgreen has quit IRC 2010-11-24T08:29:58 joel? 2010-11-24T08:30:22 * DrJoel waves to sebhub ... email reply is in the ether 2010-11-24T08:31:14 different topic ;-) 2010-11-24T08:31:25 i try to build the rtems graphics toolkit 2010-11-24T08:31:31 ohhh.. probably harder to answer LOL 2010-11-24T08:31:32 very good wiki site btw 2010-11-24T08:31:43 ok the graphics toolkit? 2010-11-24T08:31:48 libpng checks for zlib 2010-11-24T08:32:06 using zlibVersion, but we have z_zlibVersion 2010-11-24T08:32:53 hmmm.. is that how they know the library is present? 2010-11-24T08:33:04 yes 2010-11-24T08:33:18 checking for zlibVersion in -lz... no 2010-11-24T08:33:20 configure: error: zlib not installed 2010-11-24T08:34:28 maybe we should enhance the libpng configure.ac to check also for z_zlibVersion 2010-11-24T08:34:55 that's what I was thinking .. almost have an attempt.. 2010-11-24T08:35:35 could it be done similar to the check for pow on the line above where it checks for z_zlibVersion before making it an error 2010-11-24T08:36:14 AC_CHECK_LIB(z, zlibVersion, , AC_CHECK_LIB(z, zlibVersion, , AC_ERROR([zlib not installed]))) 2010-11-24T08:36:16 is the z_ prefix a standard for zlib or rtems specific? 2010-11-24T08:36:52 It is a standard zlib option which can be selected by the target .. 2010-11-24T08:37:10 at least that's my recollection.. 2010-11-24T08:37:35 did you get the 2nd email with the coverage info? 2010-11-24T08:39:15 yes 2010-11-24T08:39:30 if there is anything that needs updating, let me know 2010-11-24T08:40:55 ok, thanks a lot for this 2010-11-24T08:41:29 no problem. I probably should make them more accessible 2010-11-24T08:43:30 it is really astonishing how well rtems supports posix aware libraries 2010-11-24T08:44:12 sebhub: yep :D 2010-11-24T08:51:22 sebhub: I have had the same problem... it works with zlib 1.2.5 2010-11-24T08:51:56 its not a problem of the zlib or rtems 2010-11-24T08:52:04 its a problem of the libpng configure script 2010-11-24T08:52:56 anyway, fyi it works without configure modifications with the latest versions of libpng and zlib (1.4.4 and 1.2.5 respectively) 2010-11-24T08:53:39 last time I checked the libpng in the RTEMS graphics toolkit was pretty old 2010-11-24T08:53:45 i use libpng 1.4.4 and the zlib from the bsp 2010-11-24T08:54:07 the zlib from the bsp use the z_ prefix 2010-11-24T08:55:13 *** gedare has joined #rtems 2010-11-24T08:58:35 hm, current microwindows has a compile error 2010-11-24T08:59:03 what is it? 2010-11-24T08:59:17 Compiling engine/devblit.c ... 2010-11-24T08:59:19 /home/sh/rtems-graphics-toolkit/microwin/src/engine/devblit.c:439:1: error: conflicting types for 'GdBlit' 2010-11-24T08:59:20 /home/sh/rtems-graphics-toolkit/microwin/src/include/device.h:344:6: note: previous declaration of 'GdBlit' was here 2010-11-24T08:59:21 /home/sh/rtems-graphics-toolkit/microwin/src/engine/devblit.c:729:1: error: conflicting types for 'GdStretchBlitEx' 2010-11-24T08:59:23 /home/sh/rtems-graphics-toolkit/microwin/src/include/device.h:349:6: note: previous declaration of 'GdStretchBlitEx' was here 2010-11-24T08:59:33 why use that z_ prefix btw? 2010-11-24T08:59:59 lekernel: better namespace protection.. I recall conflicts on some items without it 2010-11-24T09:00:00 if it fails with libpng, it probably causes problems with other software packages using autoconf 2010-11-24T09:00:20 mh, works without for me :) 2010-11-24T09:00:36 a crc32 function is not that brilliant in the global namespace 2010-11-24T09:00:48 sebhub: sounds like time to investigate and submit a patch to them. 2010-11-24T09:01:36 personally, I prefer renaming a crc32 function than fixing configure scripts, but that's just me :) 2010-11-24T09:01:36 nope, it makes it hard to run programs that have their own crc32 :P 2010-11-24T09:01:57 easier to maintain external code if it isn't modified 2010-11-24T09:02:16 crc32 not alway use the same generator polynom 2010-11-24T09:03:09 yes, but when porting software, it's often easier to rename its crc32 function (possibly using a different polynom) than fixing its autoconf generated script 2010-11-24T09:04:00 maybe there exists already a zlib autoconf macro 2010-11-24T09:06:45 but when you hack source when you port it, you own that hack. When you fix the configure script to be correct and submit it upstream, it stays fixed. RTEMS doesn't want to own hacked versions of packages, we want upstream to just work 2010-11-24T09:08:11 much like i don't want to own hacked versions of RTEMS :) 2010-11-24T09:09:32 gedare: speaking of which.. is your refactoring patch correct in the PR? I wanted to commit it this morning if I could 2010-11-24T09:09:50 I'm pretty sure that is the up-to-date patch as I have it 2010-11-24T09:10:15 yeah, I replaced the tar-ball for the ppc changes (call depth) 2010-11-24T09:10:21 so that should be good to commit 2010-11-24T09:11:12 DrJoel: Permission to fire. We'll see if anyone complains :) 2010-11-24T09:11:24 i hate to ask .. but which PR 2010-11-24T09:11:29 1647 2010-11-24T09:11:39 * DrJoel warns gedare .. you know we also have to worry about coverage after the commit 2010-11-24T09:11:50 Yeah that is a concern 2010-11-24T09:12:03 I also didn't do a lot of analysis on the code size 2010-11-24T09:12:26 although technically, I think this patch reduces the static size, but uses more of the workspace. 2010-11-24T09:12:36 both are concerns but we needed the refactoring.. now to measure the impact and work on it. 2010-11-24T09:12:45 is the workspace use accounted for in confdefs.h 2010-11-24T09:12:49 Yeah 2010-11-24T09:13:01 also accounted in the spsize test for easy reading 2010-11-24T09:13:06 at least, I gave it the ole college try 2010-11-24T09:13:31 ;) 2010-11-24T09:14:20 that's all any of us can do with so many configurations 2010-11-24T09:14:23 yay, got "ls" (directory listing) to work on yaffs2 :) 2010-11-24T09:14:31 grats 2010-11-24T09:14:38 without locking yet, though 2010-11-24T09:16:19 lekernel: great!! 2010-11-24T09:16:36 gedare: changelog on doc? 2010-11-24T09:17:02 hmmm... don't got 2010-11-24T09:17:46 DrJoel: looks like my ChangeLogs are a little lacking on this PR 2010-11-24T09:17:55 I can generate some but it won't be today 2010-11-24T09:18:05 ok.. just add it to the PR. I have checked out a fresh tree, applied your changes, and will run tests on sparc/sis. If OK. then will commit. 2010-11-24T09:18:24 I will use the one you have for cpukit and then write one for doc. This needs to be closed. :) 2010-11-24T09:18:29 OK thanks 2010-11-24T09:18:35 * gedare busy with dissertation-related work 2010-11-24T09:23:43 was Roxana's work integrated in the Nano-X upstream? 2010-11-24T09:28:03 sebhub: yes. 2010-11-24T09:28:47 never used graphics with rtems, its really interesting 2010-11-24T09:31:42 sebhub: what are you planning to do with microwindows btw? 2010-11-24T09:32:23 we have a request from a medical device company to evaluate gui support on rtems 2010-11-24T09:34:09 i am more interested in FLTK 2010-11-24T09:35:17 sebhub: aaron grier is a good resource for microwindows.. he has an app with it 2010-11-24T09:35:39 sebhub: there is also the infusion pump which has a gui 2010-11-24T09:35:58 yes, its in the wiki 2010-11-24T09:38:32 *** exception13 has left #rtems 2010-11-24T09:48:57 gedare: starting to commit 2010-11-24T09:49:12 great .. fingers crossed 2010-11-24T09:49:49 I don't think coverage should be affected by this patch... 2010-11-24T09:50:24 Will be much worse when we add new schedulers :) 2010-11-24T09:50:44 I started a local set of "scheduler" test suites 2010-11-24T09:51:04 I think that will be the way to go, basically have a set of tests that can be copy-pasted and given scheduler configuration tweaks. 2010-11-24T09:51:21 Unless I actually get scheduler re-configuration to work. 2010-11-24T09:51:31 * gedare back to work 2010-11-24T09:52:46 all committed... 2010-11-24T09:53:00 * DrJoel agrees .. each scheduler plugin will have to have its own test set 2010-11-24T09:55:54 *** antgreen has joined #rtems 2010-11-24T10:01:53 are there scipts available to run the Minesweeper demo mentioned here http://www.rtems.com/wiki/index.php/GSoC_2009_Wrapup? 2010-11-24T10:05:33 just a simple pc386/qemu with the graphics NOT disabled :-D 2010-11-24T10:18:14 *** christopher_ has joined #rtems 2010-11-24T10:37:33 *** christopher_ has left #rtems 2010-11-24T10:44:41 bye 2010-11-24T10:45:40 *** sebhub has quit IRC 2010-11-24T11:01:47 gedare: any patches outstanding? 2010-11-24T11:02:38 DrJoel: red-black tree, PR 1641. There was some discussion about what to do with it 2010-11-24T11:02:39 http://www.rtems.org/ftp/pub/rtems/people/joel/coverage/erc32.html#Ospd is a coverage report with scheduler refactor in 2010-11-24T11:03:00 ping the mailing list so we can start a discussion 2010-11-24T11:04:48 ok. not qute sure how to read the report, although it shows posix enabled, so i think it is in the wrong page :p 2010-11-24T11:05:13 i assume the last two are the pre- and post-patch respectively 2010-11-24T11:05:33 so looks like there are quite a few uncovered areas. ... maybe to do with more inline routines? 2010-11-24T11:05:41 I seem to recall the coverage analysis doesn't do well with inline 2010-11-24T11:06:10 oh i see how to read it now 2010-11-24T11:06:14 not much change 2010-11-24T11:06:23 no.. I didn't see much change 2010-11-24T11:06:28 hard to tell if it would be from the scheduler or other things 2010-11-24T11:06:50 good. 2010-11-24T11:08:11 I don't see anything in http://www.rtems.org/ftp/pub/rtems/people/joel/coverage/erc32-OsPd-20101124-1012/ that would be related to scheduler work. Do you? 2010-11-24T11:08:18 aio on the other hand ... oh vey 2010-11-24T11:13:57 OK I refreshed the rbtree discussion. 2010-11-24T11:14:40 thanks.. enjoy your holiday! 2010-11-24T11:15:20 PR 1674 is also re-opened by ralf, to apply to 4.10 branch 2010-11-24T11:15:27 thanks, you too 2010-11-24T11:22:57 ok.. will look at that.. 2010-11-24T11:32:07 oh joel you're around.. i just sent you an email 2010-11-24T11:32:09 <-amar from ntp 2010-11-24T11:41:45 and I replied :-D 2010-11-24T11:42:22 Chris knows the details of his daily builder but I can do a lot of the sysadmin and give you an account whereever you need it. But with THanksgiving coming up, some may be next week 2010-11-24T11:42:49 yeah no rush there the next several days are busy for myself as well 2010-11-24T12:37:08 * lekernel sees the getcwd implementation and laughs... or not 2010-11-24T12:37:45 3 hours to get it working, nice! 2010-11-24T12:38:10 if that's the one I remember, it was copied from newlib 2010-11-24T12:38:12 seriously the filesystem does need big rewrite 2010-11-24T12:39:25 lekernel: start a place in OpenProjects in the wiki for Filesystem Improvements and put notes in it. It could be a good GSOC project if we can provide guidelines 2010-11-24T12:40:49 yup 2010-11-24T12:46:30 are you talking about the actual filesystem or file access? 2010-11-24T12:46:34 (or both?) 2010-11-24T12:48:32 actually getcwd itself is OK. but the mess it does with the eval_path isn't 2010-11-24T12:49:08 Whatever falls into the category of improvement.. it will have to provide common code and touch implementations and clean up places 2010-11-24T12:49:56 what kind of media is typically used by vendors that use rtems? 2010-11-24T12:51:13 There has been no survey.. I am assuming Flash or NFS 2010-11-24T12:51:23 * DrJoel waves bye.. time to head home for holidays 2010-11-24T12:51:26 *** DrJoel has quit IRC 2010-11-24T12:51:29 ah 2010-11-24T14:07:52 http://www.rtems.com/wiki/index.php/Refactor_the_filesystem_infrastructure 2010-11-24T14:57:07 *** christopher_ has joined #rtems 2010-11-24T15:34:37 *** antgreen has quit IRC 2010-11-24T15:54:02 *** rtemsLogger` has joined #rtems 2010-11-24T15:54:19 *** rtemsLogger has quit IRC 2010-11-24T16:00:00 *** projectgus has joined #rtems 2010-11-24T16:04:03 *** rtemsLogger has joined #rtems 2010-11-24T16:04:16 *** rtemsLogger` has quit IRC 2010-11-24T16:21:15 *** exception13 has joined #rtems 2010-11-24T16:33:06 *** antgreen has joined #rtems 2010-11-24T16:34:51 *** antgreen has quit IRC 2010-11-24T16:36:56 *** antgreen has joined #rtems 2010-11-24T16:55:51 *** gedare has quit IRC 2010-11-24T17:04:08 *** antgreen has quit IRC 2010-11-24T17:05:42 *** antgreen has joined #rtems 2010-11-24T18:11:27 *** aanjhan_ has joined #rtems 2010-11-24T18:12:12 *** aanjhan has quit IRC 2010-11-24T18:21:17 *** christopher_ has quit IRC 2010-11-24T18:33:27 kiwichris: around? 2010-11-24T18:43:52 Yeah 2010-11-24T18:44:07 Just getting the cricket on a screen next to me :) 2010-11-24T18:44:35 *** exception13 has left #rtems 2010-11-24T18:45:25 kiwichris: hehe 2010-11-24T18:45:33 is the testbox you currently use heavily loaded? 2010-11-24T18:46:08 Not really. It is just the daily build so it is fine to keep it doing this tsak 2010-11-24T18:46:34 The rtems.org for BB master should be fine if my understanding is this is just the user interface and job control 2010-11-24T18:47:04 rtbf32a also generates the CVS commit mail messages but this is not much of a load 2010-11-24T18:48:54 http://wiki.rtems.org/wiki/index.php/RTEMS_Build_Farm 2010-11-24T18:49:01 Take munin link 2010-11-24T18:49:25 kiwichris: it can get loaded when handling builds 2010-11-24T18:49:56 Yes builds will load it. It will also pressure the disk space. I see rtbf32a disk needs checking 2010-11-24T18:50:15 HOw often does the build bot run a build ? 2010-11-24T18:50:19 for the master not much disk space is needed, depends on what we want to keep around 2010-11-24T18:50:25 i'll make it launch after every commit 2010-11-24T18:50:27 Great 2010-11-24T18:50:34 we can operate that in two ways 2010-11-24T18:50:48 usually what i'll do is coalease builds every 10 mins 2010-11-24T18:51:05 Ah ok. 2010-11-24T18:51:15 so 1 commit will start a build, but any other commit within 10 mins will reset the timer 2010-11-24T18:51:23 and all those commits go into the next build if one doesn't happen for 10 mins.. 2010-11-24T18:51:35 we can change that of course i need to see how long things take and what's involved based on the hardware you have 2010-11-24T18:51:43 The builds take many hours 2010-11-24T18:52:33 is that for 1 architecture including tests? 2010-11-24T18:52:45 how much to build for 1 arch, and how long do the test take (roughly) 2010-11-24T18:53:02 No idea. Can BB gather this info ? 2010-11-24T18:53:59 yep it's all detailed.. http://termite.services.openoffice.org/buildbot/waterfall 2010-11-24T18:54:05 some OO builds triggering.. has an ETA too 2010-11-24T18:54:12 I will take a look 2010-11-24T18:54:47 http://termite.services.openoffice.org/buildbot/builders/Solaris-Intel/builds/2686 details for one build 2010-11-24T18:55:12 Nice. 2010-11-24T18:55:35 Can we set this up and perform some runs then figure out what to do ? 2010-11-24T18:55:38 we can customise it as much as we want too 2010-11-24T18:55:58 yeah we can do that 2010-11-24T18:56:18 Given the different speeds of machines this will help. 2010-11-24T18:57:10 yeah.. i was thinking i'd put one slave each on the two 4 core machines 2010-11-24T18:57:15 so we can run 8 parallel builds 2010-11-24T18:57:22 Does this mean we teach BB about the arch and the BSPs and then let it sort out the builds ? 2010-11-24T18:57:25 perhaps more.. depends 2010-11-24T18:57:41 well, you create a "builder" which is a combination of anything 2010-11-24T18:57:48 it's a unique set, for instance that could be architecture 2010-11-24T18:57:49 We also build the tools which is what the big hardware is typically used for 2010-11-24T18:57:50 compiler version 2010-11-24T18:57:55 or different options to configure etc 2010-11-24T18:58:18 we can schedule things to run however we like 2010-11-24T18:58:33 we can have inter and cross dependencies.. successes and failures it's pretty unlimited 2010-11-24T18:58:35 This is a complex process using mock and other bits and pieces. This is not automatically run rather all handled by Ralf 2010-11-24T18:58:48 Nice 2010-11-24T18:58:49 we'll automate it 2010-11-24T18:59:08 It does not need to be automated and I suspect will not want this to happen. 2010-11-24T18:59:17 no? 2010-11-24T18:59:31 no value in ensuring that process runs once a day atleast to make sure nothing is broken? 2010-11-24T18:59:40 The tools do not change that often and is a very complex. 2010-11-24T19:00:08 I suggest we sort the RTEMS code out first then once working I can approach Ralf. 2010-11-24T19:00:15 yep no worries 2010-11-24T19:00:49 once i get my account(s) it'll take a few hours to get things going the way i want then i'll play around with building rtems etc 2010-11-24T19:00:53 The patches and spec files are in CVS under control/crosspms. Maybe a commit can trigger a build but then Ralf will have had to test this before committing :) 2010-11-24T19:01:02 Nice. 2010-11-24T19:01:33 i'll leave integrating the current test framework you've done until after i have building working correctly 2010-11-24T19:01:39 Once going I will turn off my builder. All I need after that is to get at the stdio output as a file to create the web pages I have 2010-11-24T19:01:40 and we see what the resource usage for that is 2010-11-24T19:01:55 i can trivially automate that in buildbot 2010-11-24T19:02:00 Nice. 2010-11-24T19:02:10 buildbot streams stdio realtime.. and saves the logs.. i can run your script over any output at any stage 2010-11-24T19:02:19 I do not mind if we place this data into bb web pages. 2010-11-24T19:02:28 i'll seperate it out 2010-11-24T19:02:29 The key item is number of duplicate warnings 2010-11-24T19:02:33 but have a link from the bb waterfall 2010-11-24T19:02:47 if you look at the OOo waterfall they do stuff like that 2010-11-24T19:03:02 http://termite.services.openoffice.org/buildbot/builders/Mandriva-2010/builds/1714/steps/shell_4/logs/summary%20log 2010-11-24T19:03:06 http://termite.services.openoffice.org/buildbot/builders/Mandriva-2010/builds/1714/steps/shell_4/logs/warnings 2010-11-24T19:03:55 Ok. I have taken a step further by saying a specific warning is only shown once and there are X duplicates 2010-11-24T19:04:16 that's fine i can run your script it can put it's output in a seperate page 2010-11-24T19:04:22 This happens if a single header used everywhere generates a warning 2010-11-24T19:04:27 and i can also place that output in buildbot as well 2010-11-24T19:04:35 right 2010-11-24T19:04:44 Nice. It is Python and so I can changes it to do whatever you like. 2010-11-24T19:04:54 yep 2010-11-24T19:05:21 you can also submit 'try' runs too 2010-11-24T19:05:31 send a patch to buildbot and tell it to apply and use it in X configs 2010-11-24T19:05:31 How do they work ? 2010-11-24T19:05:39 That is really nice. 2010-11-24T19:06:03 then it'll create a custom build run for you etc 2010-11-24T19:06:06 yeah it's quite handy 2010-11-24T19:06:38 doing just this should save you guys a lot of time 2010-11-24T19:06:40 I bet it is. Makes handling user patches much simpler 2010-11-24T19:06:49 yeah exactly 2010-11-24T19:07:01 or even if you have something you're unsure of you can give it a full run before comitting 2010-11-24T19:07:07 before anyone else sees it 2010-11-24T19:07:29 Yes. I have some patches which are like that now 2010-11-24T19:08:10 heh everyone does that's been doing this for a while.. i think i've got a dozen for each project i work on 2010-11-24T19:08:19 things i just poke at randomly out of boredom on ocassion 2010-11-24T19:20:20 *** methril_work has quit IRC 2010-11-25T00:17:50 *** projectgus has quit IRC 2010-11-25T01:23:31 *** antgreen has quit IRC 2010-11-25T01:31:48 *** sebhub has joined #rtems 2010-11-25T01:31:57 good morning 2010-11-25T02:29:18 *** mwalle has quit IRC 2010-11-25T02:29:53 *** mwalle has joined #rtems 2010-11-25T03:04:26 sebhub, hi 2010-11-25T03:07:34 hi 2010-11-25T03:21:26 Back in a moment 2010-11-25T03:24:59 What does SDHC use as an interface ? 2010-11-25T03:25:21 bus space and bus dma (memory mapped register access) 2010-11-25T03:27:04 Ah ok it is a parallel bus type interface 2010-11-25T03:28:19 i commited a chain api violations patch, this simplifies the scheduler changes review for me 2010-11-25T03:28:47 i did not commit the chain implementation changes 2010-11-25T03:29:02 Can you please do that now ? I am snowed here. 2010-11-25T03:29:21 Not actually, just buried in things to do 2010-11-25T03:29:38 snowed? 2010-11-25T03:29:45 I can clean up an area I see. My main concern is getting const sorted cleanly 2010-11-25T03:29:51 Buried in work :) 2010-11-25T03:30:14 ah, i think snow is something unusual in australia 2010-11-25T03:30:42 We have a few ski fields here. Not as high as you have but still fun. 2010-11-25T03:31:40 On to bus and bus dma. 2010-11-25T03:31:51 I think we should also consider interrupts. 2010-11-25T03:32:07 We have 2 APIs and we should really settle on one. 2010-11-25T03:32:38 which apis do you mean? of interrupt apis we have plenty 2010-11-25T03:33:21 The one used by m68k etc and the i386, PowerPC etc 2010-11-25T03:33:26 http://www.rtems.org/onlinedocs/doxygen/cpukit/html/group__rtems__interrupt__extension.html 2010-11-25T03:33:34 http://www.rtems.org/onlinedocs/doxygen/cpukit/html/group__ClassicINTR.html 2010-11-25T03:34:37 http://www.rtems.org/onlinedocs/doxygen/cpukit/html/irq_8h.html 2010-11-25T03:34:56 Yeap. The interrupt handler APIs 2010-11-25T03:34:58 and more due to copy and paste 2010-11-25T03:36:15 please have a look at c/src/lib/libbsp/shared/src/irq-*.c 2010-11-25T03:37:19 This is the one mapped to FreeBSD. 2010-11-25T03:39:28 no, i did not use the FreeBSD interrupt api for the USB port 2010-11-25T03:43:33 the interrupt api of freebsd is bound to the device object model 2010-11-25T03:43:42 http://www.freebsd.org/cgi/man.cgi?query=bus_setup_intr&apropos=0&sektion=0&manpath=FreeBSD+8.1-RELEASE&format=html 2010-11-25T03:44:15 How did you manage the interrupts ? 2010-11-25T03:44:42 for example the ohci controller driver is split up in two files in freebsd 2010-11-25T03:45:00 one generic and one board specific 2010-11-25T03:45:17 the board specific implements all the clock, io, irq, etc setup 2010-11-25T03:45:37 So the board/bus specific file is RTEMS specific ? 2010-11-25T03:45:53 you can use here custom stuff and don't have to care much about the freebsd apis 2010-11-25T03:46:15 yes, there exists a ohci_lpc24xx.c for example 2010-11-25T03:47:08 So this is a libchip or BSP specific file ? 2010-11-25T03:47:24 bsp 2010-11-25T03:48:12 ./rtems/freebsd/dev/usb/controller/ohci_lpc32xx.c 2010-11-25T03:48:14 ./rtems/freebsd/dev/usb/controller/ohci_lpc24xx.c 2010-11-25T03:48:19 So we can leave interrupts out of this. 2010-11-25T03:48:29 you may have a look at these files of the usb port 2010-11-25T03:48:55 I know the actual ones in FreeBSD. 2010-11-25T03:49:34 ok, they use also rman(9) which is complex 2010-11-25T03:49:58 Does this leave bus and busdma plus the kproc bits and pieces ? 2010-11-25T03:50:10 I have a mmap port which I can send you. 2010-11-25T03:50:37 yes, and mutex etc. 2010-11-25T03:50:52 Is that rman ? 2010-11-25T03:51:13 no, rman is resource management 2010-11-25T03:51:28 ok. I have not looked at that API. 2010-11-25T03:51:41 http://www.freebsd.org/cgi/man.cgi?query=rman&apropos=0&sektion=0&manpath=FreeBSD+8.1-RELEASE&format=html 2010-11-25T03:51:58 i didn't use this, it was to complex 2010-11-25T03:52:15 and only required in the board specific files 2010-11-25T03:52:43 Yes I tend to agree. 2010-11-25T03:53:02 Does this mean the driver defines the address space ? 2010-11-25T03:53:33 the usb port bus space is a simple 1-1 memory mapped one 2010-11-25T03:53:44 phone 2010-11-25T03:54:17 Have you look at the Oxford controller. Its memory is on chip 2010-11-25T04:11:29 *** methril_work has joined #rtems 2010-11-25T04:15:28 back 2010-11-25T04:16:33 the address space assignment works like this: 2010-11-25T04:16:36 /* Register space */ 2010-11-25T04:16:37 e->sc_io_tag = 0U; 2010-11-25T04:16:39 e->sc_io_hdl = USBHC_BASE_ADDR; 2010-11-25T04:16:40 e->sc_io_size = 0x5cU; 2010-11-25T04:17:00 Did you port bus.c from sys ? 2010-11-25T04:17:46 ./rtems/freebsd/kern/init_main.c 2010-11-25T04:17:47 ./rtems/freebsd/kern/kern_module.c 2010-11-25T04:17:49 ./rtems/freebsd/kern/subr_kobj.c 2010-11-25T04:17:50 ./rtems/freebsd/kern/kern_linker.c 2010-11-25T04:17:52 ./rtems/freebsd/kern/subr_bus.c 2010-11-25T04:17:53 ./rtems/freebsd/kern/kern_sysctl.c 2010-11-25T04:17:55 ./rtems/freebsd/local/usb_if.c 2010-11-25T04:17:56 ./rtems/freebsd/local/bus_if.c 2010-11-25T04:17:58 ./rtems/freebsd/local/device_if.c 2010-11-25T04:17:59 thats all from the kernel 2010-11-25T04:18:29 Just looking 2010-11-25T04:20:15 Where is bus_if.c ? 2010-11-25T04:21:12 its a generate file 2010-11-25T04:21:41 related to the kobj(9) framework 2010-11-25T04:21:55 whoa do i see freebsd? 2010-11-25T04:22:06 yes 2010-11-25T04:22:22 i really need to look more at rtems 2010-11-25T04:22:25 i'm avoiding it for now :( 2010-11-25T04:22:59 atleast until i finish the testing stuff.. it's better to not know what i'm testing and focus on treating it as an object, sigh. 2010-11-25T04:23:23 verm__, our networking stack is BSD and we have a port for the USB stack which is what we are talking about 2010-11-25T04:23:28 *** lekernel has quit IRC 2010-11-25T04:23:38 kiwichris: i recognised the files 2010-11-25T04:23:49 are you using the new usb stack? i hope so 2010-11-25T04:24:30 yes, freebsd revision 213272 (trunk) 2010-11-25T04:24:39 excellent 2010-11-25T04:24:49 the stack is excellent 2010-11-25T04:25:09 yeah the stack is definatly great.. the new usb code is well.. new :) 2010-11-25T04:25:11 Hans Petter Selasky did a great job 2010-11-25T04:25:17 yeah 2010-11-25T04:25:26 <-freebsd user since 1992 2010-11-25T04:25:46 heh i remember turfing my bsdi box for freebsd like it was yesterday 2010-11-25T04:26:02 i was surprised beacuse i rarely see anything freebsd in the opensource world :) 2010-11-25T04:26:37 hm, i guess a lot of people use bsd stuff due to its quality and license 2010-11-25T04:26:55 i use it beacuse they take real bugs seriously 2010-11-25T04:27:48 but mostly you're correct, the license is what draws me to it 2010-11-25T04:28:27 infact rtems will be the first gpl software i've ever contributed to officially 2010-11-25T04:29:28 also: http://www.rtems.com/license/LICENSE.NET 2010-11-25T04:29:41 clauses 3,4 have been removed.. do you actually have code that's under the 4-clause license? 2010-11-25T04:30:08 the network stack is very old 2010-11-25T04:30:13 so, yes 2010-11-25T04:30:35 that's too bad 2010-11-25T04:32:38 i'm guessing you're still using the stack from freebsd4? 2010-11-25T04:33:20 it don't know the version 2010-11-25T04:33:33 likely 4 2010-11-25T04:33:42 4 was the best so i'd assume it is as well 2010-11-25T04:34:08 what is the problem with the current? 2010-11-25T04:34:20 nothing really, depends on how involved retms is with it 2010-11-25T04:34:28 do you use firewalling or any other features? 2010-11-25T04:34:39 no 2010-11-25T04:34:46 only ipv4 2010-11-25T04:34:48 with the newest stack you could do a lot of very cool things, like having multiple stacks 2010-11-25T04:34:53 and of course full ipv6 2010-11-25T04:35:11 you'd get pf support for free -- which is much faster and better than ipfw 2010-11-25T04:35:13 yes, i have a experimantal port of the current stack for rtems 2010-11-25T04:35:15 and get QoS via ALTQ etc 2010-11-25T04:35:20 nice 2010-11-25T04:35:26 have you done benchmarking of the two? 2010-11-25T04:35:32 the thing about 4 is it's insanely fast 2010-11-25T04:35:42 it is so experimental that it is not compile clean yet ;-) 2010-11-25T04:35:54 but.. most things are fast when you don't support much 2010-11-25T04:36:08 the 8 stack will be slower on a non smp machine i suppose 2010-11-25T04:36:14 there's other kernel bits in freebsd that slow down the stack as well 2010-11-25T04:36:30 so if you're just pulling that out i'm not sure how much slower it will be than 4.. 2010-11-25T04:36:57 sebhub: well that's probably a moot point in 3-5 years when ipv6 will be a nessicary 2010-11-25T04:37:45 ipv6 is attactive in the deeply embedded sector for example for sensor networks 2010-11-25T04:37:53 yes definatly 2010-11-25T04:38:07 but in the 3-5 year range we're going to see a wide-industry push for dropping ipv4 2010-11-25T04:38:45 there's no point in supporting ipv4 unless you have legacy devices you need to interface with.. 2010-11-25T04:39:13 yes, a stack update is on the list 2010-11-25T04:41:18 in about 6 weeks i'll probably look at the source closely i'm more eager now that i know there's freebsd bits in it :) 2010-11-25T04:42:07 what are you doing know with rtems? 2010-11-25T04:42:15 now with rtems 2010-11-25T04:44:02 i'm setting up a test environment, for tinderbox and regression testing and working on the unit tests 2010-11-25T04:44:09 and in the future continous integration 2010-11-25T04:44:21 the first two i'll be doing shortly 2010-11-25T04:44:44 when doing that kind of work it's better to not know the code you're working with to avoid undue influence 2010-11-25T04:45:42 what do you use for unit tests? 2010-11-25T04:46:25 sebhub, Joel, Ralf and I met Amar at the GSoC summit. 2010-11-25T04:48:00 sebhub, nice changes with the chain violations. Will you commit the clean up to chains itself ? 2010-11-25T04:48:43 i am not sure which version, the cast or union one? 2010-11-25T04:48:48 sebhub: whatever is appropriate. it depends on the situation i've written custom systems.. lately i've been using googletest 2010-11-25T04:48:59 I think the union is the way to go. Fixed for-ever 2010-11-25T04:49:04 ah, ok, i use google test too 2010-11-25T04:49:16 rtems is in a fairly unique situation so it'll be interesting to see what the best solution will be 2010-11-25T04:49:22 i suspect it will be a mix of things 2010-11-25T04:49:25 ok, i will commit the union one 2010-11-25T04:49:42 This is the 3 pointers that overlap with the union ? 2010-11-25T04:49:45 google test works well on rtems 2010-11-25T04:49:58 that's good 2010-11-25T04:50:39 typedef union { 2010-11-25T04:50:40 struct { 2010-11-25T04:50:42 Chain_Node Node; 2010-11-25T04:50:43 Chain_Node *fill; 2010-11-25T04:50:45 } Head; 2010-11-25T04:50:46 struct { 2010-11-25T04:50:48 Chain_Node *fill; 2010-11-25T04:50:49 Chain_Node Node; 2010-11-25T04:50:51 } Tail; 2010-11-25T04:50:52 } Chain_Control; 2010-11-25T04:51:00 Yes this is the one. 2010-11-25T04:51:19 The casts can always break. This should not. 2010-11-25T04:51:40 unless some abi makes crazy things 2010-11-25T04:52:21 on the arm eabi i don't see a problem (pavel pisa mentioned this) 2010-11-25T04:52:27 We do not export an ABI so a change should be reflected in all the code. Anything else is a broken compiler 2010-11-25T04:52:42 Did he see a problem ? 2010-11-25T04:53:13 he said that the overlap may not be as expected an proposed the packed attribute 2010-11-25T04:53:32 it don't see this concern and also tested with an arm eabi tool chain 2010-11-25T04:53:41 Nor do I 2010-11-25T04:54:00 Do we have a chain test we can add a sizeof check to ? 2010-11-25T04:54:23 good idea, i will add this to the test suite 2010-11-25T04:54:25 should "sizeof (Chain_Control) == (sizeof(void*) * 3) ? 2010-11-25T04:55:39 Just hang on while I check my union version. 2010-11-25T04:57:02 Mine matches. 2010-11-25T04:57:57 i am sure this change will break user code, but as till straumann said this is good 2010-11-25T04:58:06 I agree. 2010-11-25T04:58:24 It is a simple mechanical change and low risk 2010-11-25T05:02:20 kiwichris: i'm sure this has come up many times, but why are you guys sticking with cvs and not moving to svn atleast? 2010-11-25T05:02:52 moving from cvs to svn makes no sense in the age of git etc. 2010-11-25T05:03:13 The discussion is rather simple. First, "if it is not broken do not fix it", then if we move to where. This is not completely clear 2010-11-25T05:03:40 if you only work with the cvs head, cvs is fine ;-) 2010-11-25T05:03:42 SVN is out and git seems ok but it has issues. The Postgress move makes for interesting reading 2010-11-25T05:03:44 sebhub: well i don't really want to get into that discussion but my point was "in the age of dcvs using svn is crazy when there is atleast svn around" 2010-11-25T05:04:02 sebhub, yes, I am not prompting CVS is way 2010-11-25T05:04:03 kiwichris: professionally i've moved hundreds of repositories between dozens of RCS 2010-11-25T05:04:36 i can make the leap between any rcs, i'm just very surprised that you guys have stuck with cvs 2010-11-25T05:04:44 and haven't atleast moved to subversion 2010-11-25T05:04:53 http://lwn.net/Articles/409635/ 2010-11-25T05:05:01 yes, i read it 2010-11-25T05:05:20 i can do a perfect 1:1 history retention from cvs to svn and to git 2010-11-25T05:05:41 Our CVS may have corruptions. 2010-11-25T05:05:42 it's a lot of work depending though the problem with cvs is it's very easy to do things it was never meant to do 2010-11-25T05:05:56 Please raise the issue. I am sure many will support a change. 2010-11-25T05:05:59 kiwichris: i spent 2 months moving a cvs repo with over 60,000 commits i can handle anything :) 2010-11-25T05:06:11 and you're right 99% of the time there's a lot of corruptionm 2010-11-25T05:06:16 -m 2010-11-25T05:06:40 if it's to subversion for now i can do that trivially once i get access i'll grab the repo 2010-11-25T05:06:42 and do a conversion 2010-11-25T05:06:55 When you get an account on rtems.org please have a look. The repo is accessible from it as files. 2010-11-25T05:07:05 yep i figured i'll do that 2010-11-25T05:07:08 :) 2010-11-25T05:07:18 Chat to Joel. I think he also have something about. 2010-11-25T05:07:21 i just converted 3 repos from cvs to git two weeks ago i'm very used to the ideas heh 2010-11-25T05:07:27 SVN will not be supported. 2010-11-25T05:07:35 so it would be to git? 2010-11-25T05:07:35 git is fine with me. 2010-11-25T05:07:42 works for me, we can install gerrit 2010-11-25T05:07:47 yes, git is my favorite 2010-11-25T05:07:56 I cannot say. It is a big chance. 2010-11-25T05:08:09 can someone tar up the cvs repo for me and send it to me now? I have some time :) 2010-11-25T05:08:11 this needs to be discussed on the mailing list 2010-11-25T05:08:16 Yes 2010-11-25T05:08:31 i can figure out the issues and throw out a git repo for everyone to play with 2010-11-25T05:08:47 this sounds great 2010-11-25T05:09:01 sebhub: definatly needs to be discussed there, a lot of things need to be considered 2010-11-25T05:09:01 Nice 2010-11-25T05:09:07 but there's no harm in having a repo for people to play with :) 2010-11-25T05:09:26 there are workflow changes, code review features (ie gerrit etc) 2010-11-25T05:09:48 yes, its helpful in the discussion if you can prove that something works and if people can experiment with it 2010-11-25T05:09:58 but i'm fine in putting in the effort to do the conversion i have dozens of scripts i've come up with to help me with conversion 2010-11-25T05:10:06 sebhub: exactly. 2010-11-25T05:10:18 the #1 complaint is "well you say it's easy but.." 2010-11-25T05:10:26 if you hand it to them first and discuss after it's a very different conversation. 2010-11-25T05:10:40 yes 2010-11-25T05:11:35 the first conversion most likely won't be 1:1, no point in putting the effort in unless we decide to migrate 2010-11-25T05:12:03 as long as you can get it converted the fixups are more annoying than anything else, it's corruption that's the most work to fix 2010-11-25T05:24:54 kiwichris: how goes that repo tarball? 2010-11-25T05:45:32 verm__, sorry ? 2010-11-25T05:47:15 *** kiwichris has quit IRC 2010-11-25T05:49:07 oh 2010-11-25T05:49:19 nuts i thought he already said he'd make the tarball oh well 2010-11-25T05:52:54 i thought that too 2010-11-25T05:54:11 sebhub: you have time to zip something over to me? 2010-11-25T05:54:36 i don't have access to the repo files i guess 2010-11-25T05:54:50 oh 2010-11-25T05:54:59 they're on the rtems.org server 2010-11-25T05:55:06 i only have normal cvs access 2010-11-25T05:55:34 ah 2010-11-25T06:49:30 *** antgreen has joined #rtems 2010-11-25T07:14:13 *** exception13 has joined #rtems 2010-11-25T07:50:46 *** antgreen has quit IRC 2010-11-25T08:34:11 *** aanjhan_ is now known as aanjhan 2010-11-25T08:57:16 *** antgreen has joined #rtems 2010-11-25T09:06:00 *** exception13 has left #rtems 2010-11-25T09:51:09 *** christopher_ has joined #rtems 2010-11-25T09:56:46 *** exception13 has joined #rtems 2010-11-25T10:31:37 *** christopher_ has quit IRC 2010-11-25T10:43:46 *** sebhub has quit IRC 2010-11-25T13:07:35 *** christopher has joined #rtems 2010-11-25T13:15:54 *** christopher has quit IRC 2010-11-25T13:35:28 *** christopher has joined #rtems 2010-11-25T14:07:53 *** kiwichris has joined #rtems 2010-11-25T16:27:29 *** projectgus has joined #rtems 2010-11-25T16:45:15 *** exception13 has left #rtems 2010-11-25T18:04:59 *** DrJoel has joined #rtems 2010-11-25T18:04:59 *** ChanServ sets mode: +o DrJoel 2010-11-25T18:05:15 * kiwichris waves to Joel 2010-11-25T18:05:37 * DrJoel waves back.. enjoying a quiet Thanksgiving here .. doing OS upgrades around the house LOL 2010-11-25T18:07:35 will the buildbot kick off building all targets when something is committed? 2010-11-25T18:30:38 It should be. It is just a matter of computing resources we have 2010-11-25T18:31:07 can a build cycle finish fast enough? If the commits start coming quickly 2010-11-25T18:31:41 The issue is knowing when a commit is finished because a commit is not a transaction in CVS. 2010-11-25T18:31:59 If you think you have a problem you should submit the patch for testing. The BB will handle this 2010-11-25T18:32:21 ok.. the work flow may change a bit.. would distcc help here? 2010-11-25T18:32:46 it may 2010-11-25T18:33:13 i want to repurpose england and move it to the farm.. plus we will add another machine from the GSOC money. P 2010-11-25T18:34:15 And after the rtems.org upgrade to a 2U machine, there is that. So 2 4 core and 2-3 2core and one new build farm machine 4-8 core.. distcc should really scream for that 2010-11-25T18:36:22 I agree. I have used it before on old broken hardware and it worked well. 2010-11-25T18:36:47 Linking is the bottle neck. If we get rid of post-link that would help 2010-11-25T18:36:50 well the worst of this lot will be england which is ~2.0 Ghz but decent 2010-11-25T18:37:19 It all evens out and with distcc you can have an agent that handles the load distribution. 2010-11-25T18:37:23 We need to tackle that as a script in the flavor of the sim-scripts. 2010-11-25T18:37:39 what is the name of your autoconf'ed version of my test module? 2010-11-25T18:37:40 This way if Ralf does a mega-tools build distcc should see this load and keep away 2010-11-25T18:38:00 that's good. I suspect most of the time, the machines are idle 2010-11-25T18:38:03 It is called my disk :D 2010-11-25T18:38:22 and this ignores 2 smaller machines hooked to the rack for serial port access :) 2010-11-25T18:38:42 I am not sure I follow the script thing you talk about ? 2010-11-25T18:38:56 what's the name of the cvs module with your autoconf'ed rtems-test module 2010-11-25T18:39:17 It is a patch for rtems-testing 2010-11-25T18:39:43 ohhh... perhaps you should just commit it and force me and anyone else to take the plunge LOL 2010-11-25T18:39:44 distcc should be an autoconf change and configure option. 2010-11-25T18:39:55 Agreeded. 2010-11-25T18:39:59 Agreed 2010-11-25T18:40:29 then not terribly hard and an important step in the buildbot install.. since we have horsepower to take advantage of it 2010-11-25T18:40:44 http://distcc.org/ 2010-11-25T18:40:55 Yes I agree. 2010-11-25T18:41:12 The BB should allow us to select BSPs and targets 2010-11-25T18:41:36 I worried about that.. some rules on what to build would be possible I think... 2010-11-25T18:41:54 Why is this a concern ? 2010-11-25T18:42:15 just thinking it would be nice not to worry about it 2010-11-25T18:42:45 Oh sorry, this is optional for any user to test a patch 2010-11-25T18:43:10 ahhh... OK.. I was thinking it was a required part of the flow that auto-tripped 2010-11-25T18:43:35 My understanding is it is rather flexible. 2010-11-25T18:44:01 well then we will just have to maximize its use.. but your nightly builder could benefit from distcc'ing 2010-11-25T18:44:11 and some email followup for breakages :-D 2010-11-25T18:44:30 Sure, your email address only :D 2010-11-25T18:46:02 NO not that.. I get enough mail LOL 2010-11-25T18:46:20 filter to trash ? ;) 2010-11-25T18:46:43 Hmm... and give up all the rtems email? 2010-11-25T18:47:29 I think a daily build log to a mailling list 2010-11-25T18:47:47 Then anyone interested can join. 2010-11-25T18:47:47 that would be good. We all deal with email lists well 2010-11-25T18:47:56 Yeah. 2010-11-25T18:48:06 well time to run on.. night 2010-11-25T18:48:09 *** DrJoel has quit IRC 2010-11-25T18:48:12 Much more public and if you do not opt in that is your choice. 2010-11-25T18:48:57 *** christopher has joined #rtems 2010-11-25T20:21:29 *** backhappy has joined #rtems 2010-11-25T20:22:45 I have some misunderstanding in u-boot 's root Makefile, why OBJS is a PHONY, but LIBS 2010-11-25T20:22:50 isn't 2010-11-25T20:29:20 I said wrong, OBJS isn't a PHONY, but LIBS is a PHONY, I don't know why 2010-11-25T21:03:50 *** backhappy has quit IRC 2010-11-25T21:54:00 *** methril__ has joined #rtems 2010-11-25T21:55:33 *** methril_ has quit IRC 2010-11-25T22:39:46 kiwichris: can you tar up the repo and send it to me? 2010-11-25T22:40:09 verm__, I will have a look. 2010-11-25T22:40:28 thanks 2010-11-25T22:43:45 *** antgreen has quit IRC 2010-11-25T22:44:01 *** antgreen has joined #rtems 2010-11-25T23:52:57 doing the cvs->svn conversion now 2010-11-25T23:56:18 hmm first error 2010-11-25T23:56:19 * verm__ restarts 2010-11-26T00:24:00 not bad, 24,448 revisions 2010-11-26T00:27:54 alright, started the svn->git conversion this usually takes a while 2010-11-26T00:59:02 *** projectgus has left #rtems 2010-11-26T01:51:34 *** sebhub has joined #rtems 2010-11-26T01:52:02 good morning 2010-11-26T02:33:58 *** DanielCODE has joined #rtems 2010-11-26T02:59:03 hi, is anyone here a Google Code-In mentor? 2010-11-26T03:01:01 i am not a mentor, but maybe i can help you 2010-11-26T03:03:54 I just applied for a task for RTEMS and I need my "Request to Claim" to be accepted: http://www.google-melange.com/gci/task/show/google/gci2010/rtems/t128933972590 2010-11-26T03:04:50 ok, you should wait a bit, its night in the usa 2010-11-26T03:04:58 and also holiday time 2010-11-26T03:05:04 thanksgiving 2010-11-26T03:05:54 is there a problem if I try to do something on the task in the meantime? 2010-11-26T03:06:10 sorry, i don't know 2010-11-26T04:15:10 *** DanielCODE has quit IRC 2010-11-26T04:27:16 *** exception13 has joined #rtems 2010-11-26T04:49:31 *** DanielCODE has joined #rtems 2010-11-26T05:38:55 *** jefecomp has joined #rtems 2010-11-26T05:48:44 Hi everyone! How could I see the list of network drivers available and the list of network cards supported by RTEMS? 2010-11-26T05:54:47 i don't know of such such a list 2010-11-26T05:55:34 which target do you need? 2010-11-26T05:55:56 pc386 2010-11-26T05:56:25 pci network card? 2010-11-26T06:14:24 well the cvs->git conversion went well 2010-11-26T06:17:45 oh, that was fast 2010-11-26T06:18:12 well i already have things pre-prepared from dozens of other projects/companies 2010-11-26T06:18:35 the cvs->svn conversion took 5 mins.. svn->git took about 7 hours 2010-11-26T06:19:48 how big is the git repository? 2010-11-26T06:20:01 100mb, 24,300 commits or so 2010-11-26T06:20:06 goes back to 1995 2010-11-26T06:20:28 that is nice 2010-11-26T06:20:38 really compact 2010-11-26T06:20:43 yeah 2010-11-26T06:21:03 the tree is flattened though unfortunatly so correct cvs mergeinfo wasn't recorded 2010-11-26T06:21:12 i'll see how much of that i can fix 2010-11-26T06:21:51 what about the branches and tags? 2010-11-26T06:22:01 oh those are all there of course 2010-11-26T06:22:07 ok 2010-11-26T06:22:08 but there's no relation that's been recorded 2010-11-26T06:22:19 you can do that with cvs but few do 2010-11-26T06:22:42 it will make for a nicer history to have this information 2010-11-26T06:25:23 sebhub: hmm are there any tags in the cvs repo? 2010-11-26T06:26:07 i don't know, never used them 2010-11-26T06:26:40 yeah looks like there are 2010-11-26T06:27:04 they imported fine into git, 725 of them 2010-11-26T06:27:54 ok 2010-11-26T06:28:39 that's not too bad, i'll see what i can do about detecting the merge history then we can do fixups and apply it to the repo.. if anyone even cares 2010-11-26T06:29:23 i personally do only care about the cvs head 2010-11-26T06:29:37 and its history 2010-11-26T06:29:41 yeah but knowing the merge history can be fairly handy 2010-11-26T06:30:02 that all appears to be there but not knowing the repo we'll have to see what the others say.. i'll wait for joel or steve before giving this out 2010-11-26T06:30:34 yes, they are the important guys 2010-11-26T06:31:21 *** DanielCODE has quit IRC 2010-11-26T06:31:35 building the gcc under mingw is so damn slow 2010-11-26T06:33:13 *** jefecomp has quit IRC 2010-11-26T06:37:19 *** jefecomp has joined #rtems 2010-11-26T06:52:38 sebhub: it's too bad mingw wasn't managed with a nice installer like cygwin 2010-11-26T06:53:19 normally i build the mingw gcc with linux, but sometimes this is broken in gcc 2010-11-26T06:53:35 yeah i wouldn't trust that anyway 2010-11-26T06:53:52 it works normally quite well and fast 2010-11-26T07:18:29 sorry, I was in a meeting. 2010-11-26T07:18:44 So, is for pci network cards 2010-11-26T07:19:55 i would ask the mailing list, i work usually with arm and powerpc 2010-11-26T08:43:09 *** exception13 has left #rtems 2010-11-26T09:17:21 *** jefecomp has left #rtems 2010-11-26T09:42:19 *** exception13 has joined #rtems 2010-11-26T10:22:33 *** sebhub has quit IRC 2010-11-26T10:26:31 *** dreamcode has joined #rtems 2010-11-26T10:47:26 *** dreamcode has quit IRC 2010-11-26T11:44:01 *** gedare has joined #rtems 2010-11-26T11:49:26 *** DanielCODE has joined #rtems 2010-11-26T11:50:32 *** dreamcode has joined #rtems 2010-11-26T11:53:52 any Google Code-in mentor here? 2010-11-26T12:08:33 it's 7am where chris is and a holiday in the US where joel is 2010-11-26T12:08:45 assuming that's who you're looking for 2010-11-26T12:17:53 dreamcode: probably not, due to the US holiday. 2010-11-26T12:20:05 ok, thanks. 2010-11-26T12:42:05 *** dreamcode has quit IRC 2010-11-26T12:42:32 *** DanielCODE has quit IRC 2010-11-26T12:55:55 *** DanielCODE has joined #rtems 2010-11-26T14:07:13 *** dreamcode has joined #rtems 2010-11-26T14:33:18 *** antgreen` has joined #rtems 2010-11-26T14:34:43 *** antgreen has quit IRC 2010-11-26T14:41:28 *** methril_work has quit IRC 2010-11-26T15:43:49 *** gedare has quit IRC 2010-11-26T16:41:12 *** dr__house has joined #rtems 2010-11-26T17:31:56 *** DanielCODE has quit IRC 2010-11-26T18:14:12 *** exception13 has left #rtems 2010-11-26T18:23:02 *** dreamcode has quit IRC 2010-11-26T18:34:35 *** dreamcode has joined #rtems 2010-11-26T18:43:07 *** dreamcode has quit IRC 2010-11-26T18:43:37 *** DanielCODE has joined #rtems 2010-11-26T18:54:34 *** dreamcode has joined #rtems 2010-11-26T18:54:34 *** dreamcode has joined #rtems 2010-11-26T18:56:56 *** DanielCODE has quit IRC 2010-11-26T19:00:10 *** dreamcode has quit IRC 2010-11-26T19:10:13 *** DanielCODE has joined #rtems 2010-11-26T19:14:40 *** dreamcode has joined #rtems 2010-11-26T19:14:40 *** dreamcode has joined #rtems 2010-11-26T19:17:41 *** kaspar has joined #rtems 2010-11-26T19:17:51 hello 2010-11-26T19:18:52 has anyone ever heard of XMOS processors? http://www.xmos.com/products/silicon+devices 2010-11-26T19:19:24 I was just wondering about the feasability of porting rtems to them 2010-11-26T19:20:04 i heard a bit 2010-11-26T19:20:13 multi core gigh speed blah blah.. 2010-11-26T19:20:18 no MMU btw 2010-11-26T19:20:31 and not so HUGE datasheets as you may expect 2010-11-26T19:20:33 ah ! 2010-11-26T19:20:47 i forgot to say its compiler is propietary ;-) 2010-11-26T19:20:55 last time i checked 2010-11-26T19:21:43 at least you are happy to play with a bins 2010-11-26T19:24:10 well it is LLVM but as license allows this 2010-11-26T19:24:44 after after realize that i just give away the xmos kit :/ 2010-11-26T19:27:46 we are currently starting a foundation to make xmos more open 2010-11-26T19:28:15 http://groups.google.com/group/xcommons-foundation 2010-11-26T19:30:32 they don't have the best approach to open source at the moment 2010-11-26T19:31:05 but are generally good natured towards the ideas at least 2010-11-26T19:31:22 you have a free/open compiler? 2010-11-26T19:31:35 http://www.xmos.com/products/design-tools-source 2010-11-26T19:31:38 i thought we did 2010-11-26T19:31:47 but what you said is making me double check 2010-11-26T19:32:21 anyway 2010-11-26T19:32:24 that link is there since xmos project started btw 2010-11-26T19:32:41 as they use some open source tools mixed with their thing 2010-11-26T19:33:13 yeah i realize that 2010-11-26T19:34:23 so they have llvm sources and llvm-gcc sources on there 2010-11-26T19:35:23 and it works? 2010-11-26T19:36:19 so i think that is for c only 2010-11-26T19:36:27 c and c++ 2010-11-26T19:37:06 while you need xcc for xc language that allows for threads and channel communication and the like 2010-11-26T19:37:38 he 2010-11-26T19:44:16 *** kristianpaul has quit IRC 2010-11-26T19:45:58 *** kaspar has quit IRC 2010-11-26T19:49:21 *** DanielCODE has quit IRC 2010-11-26T20:17:21 *** kristianpaul has joined #rtems 2010-11-26T20:17:21 *** kristianpaul has joined #rtems 2010-11-26T20:18:17 ok now i can run qemu demo 2010-11-26T20:18:22 but is too slow.. 2010-11-26T21:15:38 *** antgreen` has quit IRC 2010-11-26T21:15:55 *** antgreen` has joined #rtems 2010-11-26T23:07:52 *** antgreen` has quit IRC 2010-11-27T02:00:16 *** dr__house has quit IRC 2010-11-27T03:06:02 *** DanielCODE has joined #rtems 2010-11-27T03:39:43 *** DanielCODE has quit IRC 2010-11-27T03:51:54 *** DanielCODE has joined #rtems 2010-11-27T05:33:17 *** DanielCODE has quit IRC 2010-11-27T07:11:36 *** exception13 has joined #rtems 2010-11-27T09:41:51 *** dreamcode has joined #rtems 2010-11-27T09:50:29 *** DrJoel has joined #rtems 2010-11-27T09:50:29 *** DrJoel has joined #rtems 2010-11-27T09:50:29 *** ChanServ sets mode: +o DrJoel 2010-11-27T09:50:37 * DrJoel waves hi 2010-11-27T09:51:08 Has anyone noticed that a few Code In students are flying through the Wiki tasks? :-D 2010-11-27T10:53:25 DrJoel: so, a couple of us were having a conversation along with chris the other day 2010-11-27T10:54:02 i had brought up why you guys were still using cvs.. i've convered dozens of rcs repos between each other (professionally) i took the time and converted the rtems repo to git 2010-11-27T10:56:15 anyway, the idea was to open a discussion about it.. i converted it just so people could play around 2010-11-27T10:56:39 verm__: there is a git mirror (hidden) at http://www.rtems.org/ftp/private/git/ 2010-11-27T10:56:57 I can email you the script I use.. it updates once an hour 2010-11-27T10:56:59 really? heh noone told me 2010-11-27T10:57:42 I am not personally comfortable with git but some gsoc students wanted it. So I did it for them. I haven't had the spine to bring it up publicly. LOL 2010-11-27T10:57:50 haha 2010-11-27T10:58:19 I suppose it would be a good thing to move it to a more public location on the server and get some feedback. 2010-11-27T10:58:22 using a dcvs does take some getting used to but once you get used to it you'll never think about cvs/svn ever again 2010-11-27T10:58:42 I am sure it is just a matter of breaking nearly 20 year old habbits LOL 2010-11-27T10:59:06 how do you know the git conversion goes well? 2010-11-27T11:00:01 takes a good day or two of pouring over error logs 2010-11-27T11:00:12 first between the cvs->svn conversion then svn->git 2010-11-27T11:00:38 this conversion doesn't have any branch origin info either 2010-11-27T11:00:47 thought noone may even care about that info 2010-11-27T11:01:09 wow, weird 2010-11-27T11:01:31 yours has 93 branches.. mine has 725.. 2010-11-27T11:01:35 send verm__ /tmp/git_it 2010-11-27T11:01:43 send verm__ /tmp/git_it.log 2010-11-27T11:02:11 hmmm... doesn't look like send file is working :( 2010-11-27T11:02:18 we have the same amount of commits though 2010-11-27T11:02:59 DrJoel: in any case.. usually what i do is fixup proper names 2010-11-27T11:03:04 do branch/tag name cleanup 2010-11-27T11:03:26 and if anyone cares fixes the origin of those so you can get a nice history view (this is usually a major pain) 2010-11-27T11:03:45 with git we could setup gerrit 2010-11-27T11:03:51 which would be very nice 2010-11-27T11:04:24 have you ever seen it? 2010-11-27T11:04:55 no I haven't seen it. 2010-11-27T11:05:08 I just sent you the git script that runs once an hour and the last log. 2010-11-27T11:05:08 https://review.source.android.com//#q,status:open,n,z 2010-11-27T11:05:18 it's a code review system 2010-11-27T11:05:27 so those are commits that have been made to anroid 2010-11-27T11:05:36 aahhh.. cool 2010-11-27T11:05:39 https://review.source.android.com/#change,18956 2010-11-27T11:05:45 here's a commit with several dozen deltas 2010-11-27T11:06:02 you can see the review history there near the top 2010-11-27T11:06:08 if you look at the list of deltas you can see one has two comments 2010-11-27T11:06:13 https://review.source.android.com/#patch,sidebyside,18956,1,res/values-mcc214-mnc076/strings.xml 2010-11-27T11:06:14 that one 2010-11-27T11:06:22 if you look to thr ight you can actually make inline ocmments on the diffs 2010-11-27T11:06:25 I suppose I need to learn enough git to start considering the conversion seriously (after we cut 4.10.0) 2010-11-27T11:06:32 and have discussions that way, the comitter can keep submitting new patches 2010-11-27T11:06:37 until it gets reviewed etc 2010-11-27T11:06:44 s/reviewed/approved/ 2010-11-27T11:06:50 I hate to run but I have to help one of my sons with an essay 2010-11-27T11:06:54 *** DrJoel has quit IRC 2010-11-27T11:07:21 wow crazy essays are for sundays :( 2010-11-27T11:52:04 anyone know when aanjhan ranganathan will be avaiable for Google Code-In ? 2010-11-27T11:52:07 *knows 2010-11-27T12:12:36 *** dreamcode has quit IRC 2010-11-27T12:37:50 *** Fallenou_ has joined #rtems 2010-11-27T12:39:50 *** kiwichris has quit IRC 2010-11-27T12:39:56 *** Fallenou has quit IRC 2010-11-27T12:39:56 *** kiwichris_ has joined #rtems 2010-11-27T12:40:07 *** aanjhan has quit IRC 2010-11-27T12:44:36 *** aanjhan has joined #rtems 2010-11-27T13:02:42 *** rokka_ has joined #rtems 2010-11-27T13:05:08 *** rokka has quit IRC 2010-11-27T13:20:34 *** DanielCODE has joined #rtems 2010-11-27T13:27:26 *** dreamcode has joined #rtems 2010-11-27T13:28:21 What happend to the Wiki Infobox, on all the BSP pages is the same Infobox? 2010-11-27T13:28:29 information 2010-11-27T13:29:01 *** dreamcode has quit IRC 2010-11-27T13:32:55 *** verm___ has joined #rtems 2010-11-27T13:34:54 *** dreamcode has joined #rtems 2010-11-27T13:36:04 *** verm__ has quit IRC 2010-11-27T13:36:59 *** kiwichris__ has joined #rtems 2010-11-27T13:37:23 *** kiwichris_ has quit IRC 2010-11-27T13:47:21 *** verm__ has joined #rtems 2010-11-27T14:31:09 *** DanielCODE has quit IRC 2010-11-27T15:55:50 *** dreamcode has quit IRC 2010-11-27T16:21:47 *** DanielCODE has joined #rtems 2010-11-27T16:32:16 *** DanielCODE has quit IRC 2010-11-27T16:45:07 *** DanielCODE has joined #rtems 2010-11-27T16:53:25 *** DanielCODE has quit IRC 2010-11-27T16:57:19 *** methril__ has quit IRC 2010-11-27T17:32:57 *** methril has joined #rtems 2010-11-27T17:49:43 *** methril has quit IRC 2010-11-27T17:59:21 *** methril has joined #rtems 2010-11-27T18:12:25 *** aantn has joined #rtems 2010-11-27T18:15:38 *** aantn has quit IRC 2010-11-27T18:26:45 *** exception13 has left #rtems 2010-11-27T19:07:04 *** aantn has joined #rtems 2010-11-27T19:10:02 *** aantn has quit IRC 2010-11-27T21:18:28 *** methril has quit IRC 2010-11-27T21:19:28 *** methril has joined #rtems 2010-11-28T03:57:36 *** kuzew_ has joined #rtems 2010-11-28T04:01:44 *** kuzew has quit IRC 2010-11-28T04:01:56 *** rokka has joined #rtems 2010-11-28T04:07:35 *** rokka_ has quit IRC 2010-11-28T04:58:26 *** exception13 has joined #rtems 2010-11-28T05:28:34 *** kapeels has joined #rtems 2010-11-28T06:18:24 *** exception13 has left #rtems 2010-11-28T06:23:21 *** kapeels has quit IRC 2010-11-28T06:23:50 *** kapeels has joined #rtems 2010-11-28T06:49:37 *** kapeels has quit IRC 2010-11-28T07:05:55 *** kapeels has joined #rtems 2010-11-28T07:25:50 *** kapeels has quit IRC 2010-11-28T08:21:36 *** Fallenou_ has quit IRC 2010-11-28T08:21:38 *** Fallenou has joined #rtems 2010-11-28T09:22:10 *** methril has quit IRC 2010-11-28T09:42:02 *** lekernel_ has joined #rtems 2010-11-28T09:42:06 hi 2010-11-28T09:42:13 *** kapeels has joined #rtems 2010-11-28T10:18:36 *** kapeels has quit IRC 2010-11-28T10:32:11 *** methril has joined #rtems 2010-11-28T11:55:39 shouldn't the functions rtems_shell_init_env and rtems_shell_env_free be put in include files accessible by the application? 2010-11-28T12:40:20 *** lekernel_ is now known as lekernel 2010-11-28T13:29:59 *** kuzew_ is now known as kuzew 2010-11-28T13:58:52 *** lekernel has quit IRC 2010-11-28T14:00:46 *** lekernel has joined #rtems 2010-11-28T14:22:15 *** methril has quit IRC 2010-11-28T14:22:59 *** methril has joined #rtems 2010-11-28T16:29:22 *** projectgus has joined #rtems 2010-11-28T16:54:05 *** kiwichris__ has quit IRC 2010-11-28T16:55:04 *** aantn has joined #rtems 2010-11-28T17:01:39 *** lekernel has quit IRC 2010-11-28T17:27:21 Hello 2010-11-28T17:28:07 any special option to configure gdb beyond the target and and prefix? 2010-11-28T17:28:19 i'm wroking with rtems4.11 2010-11-28T17:28:35 milkymist port :) 2010-11-28T17:30:56 * projectgus googles milkymist 2010-11-28T17:31:01 wow, interesting 2010-11-28T17:31:50 kristian, FWIW I only used those flags to compile my powerpc gdb but I don't do any remote debugging to my board yet 2010-11-28T17:32:04 I've only used it to load the binary and do some static lookups 2010-11-28T17:32:08 which is pretty undemanding :) 2010-11-28T17:34:50 well i finished compiling and installing but there is no lm32-linux-gdb :/ 2010-11-28T17:35:46 yeah i want do remote debugging :/ 2010-11-28T17:36:08 hmm then I think I am no help 2010-11-28T17:36:19 what did get built, if $(PREFIX)-gdb didn't build? 2010-11-28T17:36:37 sorry, I mean $(target)-gdb 2010-11-28T17:37:30 i did ../gdb-7.1/configure --target=lm32-rtems4.11 --prefix=/opt/rtems-4.11 2010-11-28T17:40:39 *** kiwichris has joined #rtems 2010-11-28T17:40:52 looks right to me kristianpaul but I'm on different arch & 4.10 2010-11-28T17:40:55 kiwichris probably knows 2010-11-28T17:40:58 :) 2010-11-28T17:42:45 Knows what ? 2010-11-28T18:07:37 *** titurel has joined #rtems 2010-11-28T18:24:53 ah my fault i have wrong PATH variable 2010-11-28T18:24:56 let me start over again 2010-11-28T18:25:07 *** kiwichris has quit IRC 2010-11-28T18:29:22 got it :) 2010-11-28T18:53:31 *** aantn has quit IRC 2010-11-28T19:12:12 *** aantn has joined #rtems 2010-11-28T19:42:16 *** titurel has quit IRC 2010-11-28T19:49:20 *** aantn has quit IRC 2010-11-28T23:13:58 *** kapeels has joined #rtems 2010-11-28T23:54:15 *** rfw has joined #rtems 2010-11-28T23:55:04 hi, i don't suppose anybody here can approve my google code-in task?