2013-09-02T00:00:03 Mingw build ? 2013-09-02T00:00:07 yes, mingw 2013-09-02T00:00:30 I have to head out for a while and I will take a look when I return 2013-09-02T00:00:48 it's probably just me then. I'll figure it out 2013-09-02T00:01:04 It is a change that has broken on Windows. 2013-09-02T00:02:26 We have added support to manage RAP files in archive format files and I suspect the file move call is POSIX and not windows 2013-09-02T00:04:32 that might be it, although it didn't complain about other commands like stat and unlink which is located before that particular line 2013-09-02T00:05:20 http://stackoverflow.com/questions/10846740/does-windows-support-link-system-call ?? 2013-09-02T00:05:43 hmmm... maybe I can put a quick define for this. Thanks 2013-09-02T00:05:44 Maybe we need to look at that call. I will need to check the code and see what is happening 2013-09-02T00:05:48 Thanks 2013-09-02T00:06:01 This means we need a recent Windows which maybe ok. 2013-09-02T00:07:04 Can use on XP and later 2013-09-02T00:08:16 sounds good to me. In the meantime I'll just replace that call with CreateHardLink() or something 2013-09-02T00:09:15 Let me know if it builds 2013-09-02T00:09:22 I am off out 2013-09-02T00:09:33 ok, thanks for the help 2013-09-02T00:32:00 kiwichris, it was a pain, but I managed to build it. I had to include the Windows API and also VS2010 headers though, and also add a define to specify the minimum version of windows to target. Pretty ugly if you ask me because it's another dependency 2013-09-02T00:32:56 I needed the VS2010 headers for sal.h, which the Windows API needs for some reason when compiling via mingw. I haven't looked too much into this though 2013-09-02T01:01:03 ok.. after building the new rtems-ld from git upstread, with the CreateHardLink() wrapper.. the error I was getting is gone and the RAP seems to have been created successfully. I haven't tried loading the RAP file yet though 2013-09-02T01:43:18 *** sebhub has joined #rtems 2013-09-02T01:44:03 good morning 2013-09-02T01:47:43 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-02T01:56:37 *** MegaAlex|away has quit IRC (Ping timeout: 248 seconds) 2013-09-02T01:57:44 *** MegaAlex|away has joined #rtems 2013-09-02T01:58:52 the9a3eedi, thans 2013-09-02T02:09:26 I would send a patch, but the approach I took is terrible. Not sure what the correct approach to fixing this problem is.. 2013-09-02T02:09:43 do you think I should send to the list anyway? 2013-09-02T02:13:07 I will sort it out 2013-09-02T02:13:39 I need to take a look at the change Peng has made and see where we can make changes. 2013-09-02T02:13:53 the9a3eedi, we have support for debugging via gdb as well 2013-09-02T02:36:49 kiwichris, nice. I might need that soon.. though I'm not sure how it's going to work with Eclipse, which is what I'm using 2013-09-02T02:37:48 the9a3eedi, it will be part of the RTEMS gdb. You just set a breakpoint on a function in a RAP file and once it is loaded and the code runs the breakpoint is hit 2013-09-02T02:38:37 great 2013-09-02T03:02:07 *** arvind_khadri has joined #rtems 2013-09-02T04:01:00 *** sebhub has quit IRC (Read error: Operation timed out) 2013-09-02T04:01:50 *** sebhub has joined #rtems 2013-09-02T04:26:15 *** the9a3eedi has quit IRC (Ping timeout: 245 seconds) 2013-09-02T04:47:55 kiwichris: how would this work with JTAG <-> gdb debug solutions? 2013-09-02T05:00:34 *** MegaAlex|away has quit IRC (Ping timeout: 246 seconds) 2013-09-02T05:01:46 *** MegaAlex|away has joined #rtems 2013-09-02T05:37:30 *** arvind_khadri has quit IRC (Ping timeout: 245 seconds) 2013-09-02T05:55:23 *** arvind_khadri has joined #rtems 2013-09-02T06:53:14 *** marcinb has joined #rtems 2013-09-02T07:21:41 *** MegaAlex|away has quit IRC (Ping timeout: 245 seconds) 2013-09-02T07:22:48 *** MegaAlex|away has joined #rtems 2013-09-02T07:26:11 *** marcinb has quit IRC (Quit: Leaving) 2013-09-02T07:51:42 *** MegaAlex|away has quit IRC (Ping timeout: 264 seconds) 2013-09-02T07:52:18 *** MegaAlex|away has joined #rtems 2013-09-02T08:05:17 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-02T08:09:10 *** Thorn has quit IRC (Ping timeout: 256 seconds) 2013-09-02T08:37:54 *** MegaAlex|away has quit IRC (Ping timeout: 264 seconds) 2013-09-02T08:38:48 *** MegaAlex|away has joined #rtems 2013-09-02T09:12:05 *** arvind_khadri has joined #rtems 2013-09-02T09:19:39 *** arvind_khadri has quit IRC (Read error: No route to host) 2013-09-02T10:34:11 *** sebhub has quit IRC (Remote host closed the connection) 2013-09-02T10:56:10 *** monstr has quit IRC (Ping timeout: 246 seconds) 2013-09-02T12:00:46 *** arvind_khadri has joined #rtems 2013-09-02T13:29:15 *** arvind_khadri has quit IRC (Ping timeout: 260 seconds) 2013-09-02T19:07:01 *** the9a3eedi has joined #rtems 2013-09-02T20:09:03 *** peerst has quit IRC (Ping timeout: 240 seconds) 2013-09-02T20:39:13 *** freenix has joined #rtems 2013-09-02T21:15:55 *** antgreen has joined #rtems 2013-09-02T21:38:16 *** peerst has joined #rtems 2013-09-02T21:45:57 Hi kiwichris. How about use `#if __USER_LABLE_PREFIX__`? I find that to bfin and h8300, this macro is defined `#define __USER_LABEL_PREFIX__ _` ,with `_` at last. 2013-09-02T22:30:11 freenix, I would just prepend it to "rtems". 2013-09-02T22:30:38 #define RTL_ENTRY_POINT __USER_LABLE_PREFIX__ # "rtems" 2013-09-02T22:30:43 or whatever it is 2013-09-02T22:32:39 Then the `#define SYM(_x) _x` can also use this way? like this `#define SYM(_x) __USER_LABLE_PREFIX__ # _x` 2013-09-02T22:38:57 Yes 2013-09-02T22:39:07 Thanks. Got it. 2013-09-03T00:15:01 *** monstr has joined #rtems 2013-09-03T01:31:39 *** arvind_khadri has joined #rtems 2013-09-03T01:49:22 *** sebhub has joined #rtems 2013-09-03T01:50:09 good morning 2013-09-03T02:49:54 *** the9a3eedi has quit IRC (Ping timeout: 264 seconds) 2013-09-03T02:50:04 *** the9a3eedi has joined #rtems 2013-09-03T04:28:08 *** the9a3eedi-2 has joined #rtems 2013-09-03T04:31:18 *** the9a3eedi has quit IRC (Ping timeout: 264 seconds) 2013-09-03T04:56:40 *** the9a3eedi-2 has quit IRC (Ping timeout: 245 seconds) 2013-09-03T05:11:47 *** monstr has quit IRC (Remote host closed the connection) 2013-09-03T05:20:38 *** sebhub has quit IRC (Ping timeout: 240 seconds) 2013-09-03T05:21:01 *** sebhub has joined #rtems 2013-09-03T07:04:40 *** longqm has joined #rtems 2013-09-03T07:12:24 *** antgreen has quit IRC (Quit: Leaving) 2013-09-03T07:56:26 *** bradfa has joined #rtems 2013-09-03T07:57:56 *** sebhub has quit IRC (Ping timeout: 245 seconds) 2013-09-03T08:43:17 *** freenix has quit IRC (Remote host closed the connection) 2013-09-03T08:56:11 *** bradfa has left #rtems ("Leaving") 2013-09-03T09:03:30 *** sebhub has joined #rtems 2013-09-03T09:11:28 *** antgreen has joined #rtems 2013-09-03T09:14:02 *** longqm has quit IRC (Quit: Leaving) 2013-09-03T09:54:32 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-03T09:57:26 *** MegaAlex|away has quit IRC (Ping timeout: 240 seconds) 2013-09-03T09:58:11 *** MegaAlex|away has joined #rtems 2013-09-03T10:34:26 *** sebhub has quit IRC (Remote host closed the connection) 2013-09-03T12:00:53 *** antgreen has quit IRC (Ping timeout: 248 seconds) 2013-09-03T12:16:16 *** antgreen has joined #rtems 2013-09-03T12:20:44 *** arvind_khadri has joined #rtems 2013-09-03T12:46:35 *** antgreen has quit IRC (Ping timeout: 256 seconds) 2013-09-03T13:00:26 *** antgreen has joined #rtems 2013-09-03T13:13:55 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-03T13:46:29 *** antgreen_ has joined #rtems 2013-09-03T13:47:33 *** antgreen has quit IRC (Ping timeout: 248 seconds) 2013-09-03T15:18:59 *** rtdin has joined #rtems 2013-09-03T15:34:40 *** gedare has joined #rtems 2013-09-03T15:57:30 *** gedare has quit IRC (Ping timeout: 245 seconds) 2013-09-03T16:16:22 *** antgreen_ has quit IRC (Quit: Leaving) 2013-09-03T16:26:33 *** rtdin has quit IRC (Ping timeout: 250 seconds) 2013-09-03T16:57:14 *** dmbaturin has joined #rtems 2013-09-03T17:04:37 Are there prebuilt binary images of RTEMS, or one should always build it from source? 2013-09-03T18:53:03 *** rtdinesh has joined #rtems 2013-09-03T19:16:17 *** the9a3eedi has joined #rtems 2013-09-03T20:28:54 Question about RTL.. if I embedded the symbol table into my image, I'm supposed to initialize RTL (as well as the symbol table) before I load anything, right? Or is it done automatically? 2013-09-03T20:36:56 the9a3eedi, it should be done automatically. It uses a weak symbol trick 2013-09-03T20:40:50 hmm.. I see 2013-09-03T21:00:51 *** rtdinesh has quit IRC (Ping timeout: 250 seconds) 2013-09-03T21:52:12 kiwichris, I've been getting a segfault in rtems_rtl_symbol_hash() when loading a rap file, which is why I was asking.. 2013-09-03T21:52:37 but the rap file was made using the latest git rtl-host, and the target code wasn't updated in a while, so maybe that's why.. 2013-09-03T21:56:45 *** antgreen has joined #rtems 2013-09-03T21:59:05 actually, no it seems to be occurring in rtems_rtl_symbol_global_find() 2013-09-03T22:58:30 *** rtemsLogger` has joined #rtems 2013-09-03T23:01:11 *** rtemsLogger has quit IRC (Ping timeout: 264 seconds) 2013-09-04T00:11:58 ah, it seems that rtl is actually null, which means it's never been initialized... kiwichris I guess this means the symbol trick isn't working for me? 2013-09-04T00:12:46 the9a3eedi, it works by linking an ELF object file with the symbol table to the base image 2013-09-04T00:13:16 the9a3eedi, this creates a new base image 2013-09-04T00:14:22 I see.. the way I was doing it was that the symbol table object file was linked with all the rest of the object files, not with the base image. 2013-09-04T00:14:39 Thanks 2013-09-04T00:14:52 Try just loading that object file. 2013-09-04T00:14:57 the symbol one. 2013-09-04T00:15:01 It should also work 2013-09-04T00:15:10 That is the other method of managing symbols 2013-09-04T00:15:10 I did that before and it works 2013-09-04T00:15:17 I wanted to embed it in the image this time though 2013-09-04T00:15:51 Interesting. This is a good idea. I will talk to Peng about doing this. 2013-09-04T00:16:29 We would have to add special code to manage this in the RAP format 2013-09-04T00:17:28 With the latest code in Peng's github repo you could create an ra archive so one file then load the symbols first then the app 2013-09-04T00:17:35 This issue is the load order 2013-09-04T00:19:53 kiwichris, actually the only reason why I did it the way I did was that I was using eclipse, and adding stuff to the build process isn't very flexible as editing your own makefile lol 2013-09-04T00:20:23 so I just had it generate a symbol c file after every compile. Then in the next build the symbol file gets compiled and linked together with the rest of the files 2013-09-04T00:20:34 If you use rtems-ar you can create an archive of RAP files. 2013-09-04T00:21:03 You only need to regenerate the symbols once per base image build 2013-09-04T00:21:51 yes, I know. It's merely me adapting the process to eclipse. 2013-09-04T00:22:18 well, that helps a lot. Thanks :D 2013-09-04T00:29:58 the9a3eedi, no problem. Nice to have someone using the code. 2013-09-04T00:30:22 the9a3eedi, it is not announced and there is not a lot of doco 2013-09-04T00:38:58 you know what would be a nice idea with RTL? If somehow the shell can be used to call any function simply by typing it in along with args, much like a certain OS does. 2013-09-04T00:39:01 I think RTL can make this possible 2013-09-04T00:42:03 There is a command 2013-09-04T00:43:05 really? hmmm 2013-09-04T00:44:35 ah, rtl-shell... why didn't I notice it before? 2013-09-04T00:45:01 I'll check it out 2013-09-04T01:36:31 *** the9a3eedi has quit IRC (Read error: Connection reset by peer) 2013-09-04T01:43:20 *** the9a3eedi has joined #rtems 2013-09-04T02:04:14 hmm.. I'm having some trouble understanding how to link the symbol table into the base image. It's supposed to be something like sparc-rtems4.10-gcc baseimage.exe symboltable.o -o "newbaseimage.exe" ? 2013-09-04T02:04:28 because that's not working for me 2013-09-04T02:04:41 *** sebhub has joined #rtems 2013-09-04T02:05:21 good morning 2013-09-04T02:06:40 morning 2013-09-04T02:37:47 *** monstr has joined #rtems 2013-09-04T03:04:31 *** arvind_khadri has joined #rtems 2013-09-04T03:09:56 *** the9a3eedi has quit IRC (Ping timeout: 260 seconds) 2013-09-04T03:10:39 *** the9a3eedi has joined #rtems 2013-09-04T03:13:46 *** the9a3eedi-2 has joined #rtems 2013-09-04T03:17:52 *** the9a3eedi has quit IRC (Ping timeout: 264 seconds) 2013-09-04T03:44:37 *** freenix has joined #rtems 2013-09-04T04:01:28 *** the9a3eedi-2 has quit IRC (Read error: Connection reset by peer) 2013-09-04T04:01:39 *** the9a3eedi has joined #rtems 2013-09-04T04:34:32 Hi! 2013-09-04T05:18:38 Hi :) 2013-09-04T05:28:24 the9a3eedi: Are there any binary images available, or I should just set up the host and build my own? 2013-09-04T05:32:42 RTEMS? None as far as I know. 2013-09-04T05:34:14 Why? Just curious, I understand I'll need to build my own anyway. 2013-09-04T05:34:30 dmbaturin, it's pretty easy to build RTEMS though. The source has a few samples that can be built automatically. 2013-09-04T05:35:50 Well, I'm not an expert.. and not an official RTEMS dev, and I still consider myself pretty new, so I can't say for sure.. but RTEMS is meant for embedded devices. It's supposed to be customized per application, so having pre-built binaries doesn't make sense 2013-09-04T05:37:27 Indeed, it just might be easier for newcomers to look at a working example on a simulator to know what to expect when it builds properly. :) 2013-09-04T05:37:56 Well, if it builds properly, you get an executable :P 2013-09-04T05:37:59 *** longqm has joined #rtems 2013-09-04T05:39:29 but you can easily build examples from the source. 2013-09-04T05:41:06 Ok, will just setup the host. 2013-09-04T05:43:39 the9a3eedi: What is the "classic API"? Where is it defined and should I ever use it? 2013-09-04T05:45:03 dmbaturin, the "classic" api is the default api that RTEMS provides, thats not POSIX or ITRON. It's what people usually use I think. Details about the C user's guide are in the C user's guide 2013-09-04T05:45:20 http://rtems.org/onlinedocs/doc-current/share/rtems/html/ check here 2013-09-04T05:47:49 (I'm going to use RTEMS to learn about hard RT OSes, so I may have a lot of dumb questions as I never worked with one before) 2013-09-04T05:47:58 RTEMS has a "super core" that actually does all the OS stuff. The APIs (including classic, POSIX, and ITRON) merely interface with this supercore.. that way the different APIs are all interoperable. 2013-09-04T05:48:30 I consider myself pretty new as well, so I'm not an expert 2013-09-04T05:48:45 you should read the Getting Started and C User's Guide, they explain a lot 2013-09-04T05:49:20 Are there Ada bindings for the classic API? 2013-09-04T05:50:16 Don't know much about Ada, so I'm not sure how it works. There's an Ada user's guide 2013-09-04T05:50:52 Also, what is BSP? 2013-09-04T05:54:32 Board Support Package 2013-09-04T05:55:01 It's how RTEMS supports multiple architectures and subarchitectures 2013-09-04T05:55:18 Ah, got it. 2013-09-04T05:58:08 Also, I'm not sure if this is a correct question. How do people prove their code execution takes expected time? 2013-09-04T06:09:35 different methods, like the analysis of code. One way of demonstrating that I heard of is that the code takes expected time you can hook up an oscilloscope to gpio, and the time between gpio on and off manually. 2013-09-04T06:10:05 I haven't actually gone into this much though, but this is what I know 2013-09-04T06:31:41 Well, GPIO and oscilloscope is a test, not proof. :) 2013-09-04T06:56:04 yeah, which is why I said "demonstrate". For proof one of the main ways is through formal verification, but this can be very difficult. 2013-09-04T06:59:26 I'm not 100% sure, but in order to formally prove that an OS can guarantee to do a task in a specific amount of time, you'd need to analyze each algorithm used and show that their run-time is deterministic enough to guarantee that it'll finish the job on time 2013-09-04T07:01:05 from what I can see in google, only a tiny few of RT kernels are formally verified 2013-09-04T07:09:13 kiwichris, turns out that my code _was_ linking properly. The segfault came because I tried to find a symbol before the symbol table was actually loaded.. the rtems_rtl_symbol_global_find() function does not initialize RTL before it actually does the searching, and causes a segfault as a result 2013-09-04T07:09:53 but maybe that function wasn't supposed to be used directly by user code.... 2013-09-04T07:10:42 So RTEMS was not formally verified and there is no reference where I can find what time each API call takes? 2013-09-04T07:18:51 not sure, but the api calls take a deterministic amount of time. Things like task switching are constant and deterministic. Event sending and receiving are also constant. I'm not the right person to ask about this though. 2013-09-04T07:19:33 also, take a look at the Rate monotonic manager in the c user's guide. 2013-09-04T07:19:43 Also, does "single process, multi-threaded" means I only can run one execuctable at the same time? 2013-09-04T07:19:53 * mean 2013-09-04T07:26:58 An RTEMS application only manages threads, not processes. There isn't any isolation between threads, if that makes any sense. 2013-09-04T07:27:23 A thread can directly access the same global variables/functions that another thread does 2013-09-04T07:28:17 So I'm supposed to embed everything into single executable and make sure nothing uses the same resource? 2013-09-04T07:29:50 yes 2013-09-04T07:30:08 RTEMS provides things like semaphores, queues, events, and things like that to help manage resources 2013-09-04T07:31:03 E.g. if I want time sync with NTP/PTP, I can't run a client ported from another OS as a separate process and must extract/rewrite the algorithm and link to my executable? 2013-09-04T07:35:41 not in the same sense as how you "start an exe" in Windows for example, no 2013-09-04T07:36:18 as you said, you need to extract the source and incorporate it into the executable 2013-09-04T07:36:59 however, there's a sideproject that allows you to load functions from a file. It's not exactly the same (it's more like loading a shared library into a running process), but you can use it to do something similar. 2013-09-04T07:37:09 check http://wiki.rtems.org/wiki/index.php/Dynamic_Object_File_Loading 2013-09-04T07:37:29 it's currently WIP though 2013-09-04T07:37:40 Oh, interesting. I think I should start with traditional approach anyway. 2013-09-04T07:39:00 Most of my programming experience is on UNIX, so I've never worked with this type of multitasking. :) 2013-09-04T07:39:37 yeah, it's pretty different. But other RTOS's are similar 2013-09-04T07:40:31 from what I understood, this is done in order to minimize overhead and complexity, which are undesirable in an RTOS 2013-09-04T07:41:02 Yeah, that makes sense. 2013-09-04T07:41:26 Is QNX multiprocess? 2013-09-04T07:42:28 I think so, haven't worked with it 2013-09-04T07:52:52 You might want to send to the mailing list if you want answers from the RTEMS devs (they're on IRC sometimes as well). 2013-09-04T08:01:04 Well, I should read the docs first. :) 2013-09-04T09:33:41 *** freenix has quit IRC (Remote host closed the connection) 2013-09-04T09:47:22 *** MegaAlex|away is now known as MegaAlex 2013-09-04T10:54:01 *** sebhub has quit IRC (Remote host closed the connection) 2013-09-04T11:01:56 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-04T11:46:18 *** monstr has quit IRC (Ping timeout: 264 seconds) 2013-09-04T12:02:48 *** arvind_khadri has joined #rtems 2013-09-04T12:28:36 Does RTEMS have a bugtracker? 2013-09-04T12:29:46 I'd like to find some small issues to look into to familiarize myself with the system. :) 2013-09-04T13:48:02 https://www.rtems.org/bugzilla/ 2013-09-04T13:51:56 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-04T14:18:42 verm__: Thanks! 2013-09-04T14:33:40 sure 2013-09-04T14:34:35 Also, are there any books/articles ones should read before starting to use an RTOS? 2013-09-04T15:42:13 *** marcinb has joined #rtems 2013-09-04T15:43:21 Hello everybody, I am the student doing the ESA summer of code 2013-09-04T15:55:19 hello marcin! 2013-09-04T15:57:56 I logged in here before but it seemed rather empty 2013-09-04T15:59:45 I was wondering how many of the current developers are based at OAR and how many are outside contributors 2013-09-04T16:01:18 2 are at OAR 2013-09-04T16:03:37 and how many people is it in total at the moment? 2013-09-04T16:05:28 hmm i don't know actually :) 2013-09-04T16:05:40 it's an open source project, there are many, many contributors 2013-09-04T16:06:02 RTEMS recently switched to git which keeps the name and information in the patch.. previously it was CVS so who knows how many hundreds have contributed 2013-09-04T16:06:16 you'd probably have to go through rtems-devel and check the bug tracker 2013-09-04T16:06:29 joel would probably have an idea 2013-09-04T16:08:07 no worries, just wanted to get a sense of the size of the community 2013-09-04T16:33:56 *** MegaAlex has quit IRC (Read error: Operation timed out) 2013-09-04T16:43:45 *** MegaAlex has joined #rtems 2013-09-04T16:50:16 so Im guessing I should use the unstable build set to use the new build system you mentioned earlier? 2013-09-04T16:51:55 yes that will be the one you want to use 2013-09-04T16:52:00 marcinb: see your private message 2013-09-04T16:53:18 ok thanks 2013-09-04T16:53:20 *** MegaAlex has quit IRC (Read error: Operation timed out) 2013-09-04T16:59:40 *** MegaAlex has joined #rtems 2013-09-04T17:07:20 *** MegaAlex has quit IRC (Ping timeout: 240 seconds) 2013-09-04T17:11:39 *** marcinb has quit IRC (Quit: Leaving) 2013-09-04T17:19:40 *** MegaAlex has joined #rtems 2013-09-04T17:31:58 *** MegaAlex is now known as MegaAlex|away 2013-09-04T20:50:33 kiwichris, sorry to bother you too much about this, but I'm still unable to figure out how to load this module. 2013-09-04T20:51:10 Assuming I got a bunch of code divided into files, and I compile each file into a .o object file. Whats the proper method of linking them together? using RAP? or ELF? 2013-09-04T20:51:30 or perhaps an archive of RAPs? 2013-09-04T21:03:25 the9a3eedi, there is no specific solution that fits every one so we have provided a range of tools you can use 2013-09-04T21:03:51 the9a3eedi, there are a few common use cases and we have attempted to match up against them 2013-09-04T21:04:27 the9a3eedi, for example consider a version of python made to be loadable. 2013-09-04T21:04:53 we would have a python.rap which is like the standard python command. 2013-09-04T21:05:25 Python also have a number of modules that are loadable at runtime via dlopen etc and these would be separate .rap files. 2013-09-04T21:07:06 ELF loading was implemented first and will stay. It also allows you the ability in a development environment to use the standard libraries via say a NFS mount 2013-09-04T21:07:55 RAP is about delivery of applications for production. It is smaller in size and combines a number of objects into a logical set based on your application. 2013-09-04T21:11:52 kiwichris, I see.. 2013-09-04T21:13:55 I'm just having a lot of trouble loading this RAP file for some reason.. so I'm just trying to figure out if I'm doing something wrong 2013-09-04T21:14:41 I'm getting a segfault rtems_rtl_symbol_hash() in rtl-sym.c right now.. 2013-09-04T21:15:04 that should not be happening 2013-09-04T21:15:52 hold on, let me get a backtrace. I got this problem yesterday.. it had something to do with the strtab 2013-09-04T21:16:25 the RAP file was linked with rtems-ld using *.o as the input, so it's a combination of a bunch of object files 2013-09-04T21:17:02 ok 2013-09-04T21:17:14 Is this with Peng's repo ? 2013-09-04T21:17:14 When I first tried out RTL, I haven't had this issue, but the code I was testing was very small... a simple hello world. This one on the other hand is a more complicated task. So maybe that's why the issue cropped up 2013-09-04T21:17:30 rtems-ld was done with peng's repo, yes 2013-09-04T21:17:37 And the runtime ? 2013-09-04T21:17:48 Have you updated to his target side ? 2013-09-04T21:18:00 not yet, no. I haven't found any updates in the git repo 2013-09-04T21:18:16 I suggest you give this a try 2013-09-04T21:18:26 They are on his github repo 2013-09-04T21:18:51 https://github.com/MrVan/gsoc2013-rtl 2013-09-04T21:19:09 I am close to pulling these changes and pushing them. He has some changes to make. 2013-09-04T21:19:44 hmmmm I'll give this a try. Thanks! 2013-09-05T00:24:59 *** freenix has joined #rtems 2013-09-05T00:25:08 hi kiwichirs. 2013-09-05T00:26:08 freenix, hi 2013-09-05T00:26:20 Just added some comments to github 2013-09-05T00:26:50 ok. 2013-09-05T00:27:01 There is also a bug in the host code when on Windows. There is no 'link' function on Windows 2013-09-05T00:27:11 What is the 'link' call doing ? 2013-09-05T00:27:16 My commits? 2013-09-05T00:27:34 Yes 2013-09-05T00:27:40 I think related to the archive 2013-09-05T00:27:46 which one? I check it now 2013-09-05T00:28:20 Not sure. I have no looked. the9a3eed let me know on IRC the other day 2013-09-05T00:28:55 I tested the code on Linux, but not tried no windows 2013-09-05T00:29:14 ok.i'll try to figure it out. 2013-09-05T00:29:50 I wonder if it is in libiberty 2013-09-05T00:30:45 actually I did not use mktemp in libiberty 2013-09-05T00:31:21 rld-outputter.cpp: if (::link (new_name.c_str (), name.c_str ()) < 0) 2013-09-05T00:32:10 I am not sure about this. 2013-09-05T00:32:38 I'll install mingw and compile this on windows. 2013-09-05T00:32:52 Not sure what ? 2013-09-05T00:32:58 Not sure about what ? 2013-09-05T00:33:29 not sure about `link can not be used in mingw` 2013-09-05T00:33:41 It cannot. 2013-09-05T00:34:07 Also looking at this code with 'rld_XXXXXX'. Where is it created ? 2013-09-05T00:35:14 In the directory which rtems-ld runs. 2013-09-05T00:35:32 I think you should use the libiberty call make_temp_file. 2013-09-05T00:35:55 in make-temp-file.c 2013-09-05T00:36:21 The new archive should be created in the directory of the one being replaced. 2013-09-05T00:36:31 You may not have write perms to the current dir 2013-09-05T00:36:48 yeah. in the dir of the one being replaced 2013-09-05T00:38:00 But make_temp_file in libierty create the file in /tmp dir. 2013-09-05T00:38:31 And I think the code in rtld-process.h should be used and the file added to the temporary file container so if a signal happens it will be cleaned up 2013-09-05T00:39:24 Is this just a rename you are doing with link and unlink ? 2013-09-05T00:39:34 yeah. 2013-09-05T00:39:36 Can the rename call be used ? 2013-09-05T00:40:32 ok. I try rename 2013-09-05T00:40:35 It looks to be in stdio.h in mingw 2013-09-05T00:41:07 how to remove a file in mingw? 2013-09-05T00:41:17 just like unlink in linux 2013-09-05T00:41:24 yes 2013-09-05T00:41:41 The link call is missing in Windows before XP 2013-09-05T00:42:13 On Windows XP and later there is a CreateHardLink 2013-09-05T00:42:30 ok. I'll use rename to replace link. 2013-09-05T00:42:46 Thanks 2013-09-05T00:42:52 is there any function that can replace unlink? 2013-09-05T00:43:03 on mingw 2013-09-05T00:44:29 you can use unlink 2013-09-05T00:44:46 About #define RTL_ENTRY_POINT #__USER_LABEL_PREFIX__ "rtems" , compilation failed with `rap.c:338:5: error: stray '#' in program` 2013-09-05T00:45:28 So I used `GLUE` to fix this. Ok, I'll use RTL_GLUE to replace GLUE. GLUE is just to concat two tokens 2013-09-05T00:46:03 Yeah it might need the glue macro 2013-09-05T00:46:17 Please rename GLUE 2013-09-05T00:46:27 It could be used by user and then we break their code 2013-09-05T00:46:27 ok. 2013-09-05T00:46:42 We need to occupy our own namespace 2013-09-05T00:52:08 hi, kiwichris. I have pushed the commits to github about `GLUE` at https://github.com/MrVan/gsoc2013-rtl/commit/07d14d4e3239188e4f0ee407d7ae3e49fc71741b 2013-09-05T00:52:18 Thanks 2013-09-05T00:55:10 What if a target has something different to _ ? 2013-09-05T00:56:49 this patch can not handle it. 2013-09-05T00:57:20 because I found that __USER_LABEL_PREFIX only is defined empty or _ 2013-09-05T00:58:29 Is it the empty one that fails ? 2013-09-05T00:59:20 sorry 2013-09-05T01:00:36 on sparc, it is empty. #if (RTL_GLUE(__USER_LABEL_PREFIX__, 1) == RTL_GLUE(_, 1)) is used to judge whether __USER_LABEL_PREFIX is _ or empty. 2013-09-05T01:00:46 *** monstr has joined #rtems 2013-09-05T01:02:47 Ah ok 2013-09-05T01:03:21 must go back soon 2013-09-05T01:03:50 *** kiwichris has quit IRC (Quit: This computer has gone to sleep) 2013-09-05T01:26:09 *** kiwichris has joined #rtems 2013-09-05T01:44:20 *** sebhub has joined #rtems 2013-09-05T01:45:42 freenix, hi 2013-09-05T01:47:23 hi, kiwichris 2013-09-05T01:47:42 I have attempted to pull your rtl-host repo and I get conflicts 2013-09-05T01:48:11 good morning 2013-09-05T01:48:20 sebhub, hi 2013-09-05T01:48:20 sorry. I may override the committed patches. 2013-09-05T01:49:35 Another question about RTL: Can I load global variables as part of a RAP module? Doesn't seem to work for me 2013-09-05T01:49:43 back soon 2013-09-05T01:50:17 the9a3eedi. do you mean a rap module in a ra file? 2013-09-05T01:50:38 *** kiwichris has quit IRC (Read error: Connection reset by peer) 2013-09-05T01:50:49 freenix, no, just a regular rap. I created the RAP by using something like "rtems-ld *.o" 2013-09-05T01:50:55 or you just `rap ld ./x.rap` ? 2013-09-05T01:51:46 I'm using dlopen to load it 2013-09-05T01:52:26 which reposity do you use? 2013-09-05T01:52:48 some patches have not been merged to chris's upstream. 2013-09-05T01:53:04 I'm using MrVan's repository 2013-09-05T01:53:24 ok. it is mine 2013-09-05T01:54:00 which platform do you use? 2013-09-05T01:54:07 sparc, leon3 2013-09-05T01:55:05 I use `sparc-rtems4.11-gdb build/sparc-rtems4.11-sis/rtld`. all looks fine . would you please paste the error msg when loading rap file? 2013-09-05T01:56:33 actually, the way I'm using the RTL stuff is that I have copied the core RTL source into my own project in Eclipse 2013-09-05T01:58:00 the error I'm getting is "global symbol not found: " 2013-09-05T01:58:19 the particular symbol it complains about is supposed to be located in the source that was used for the RAP file 2013-09-05T01:59:07 so I think I might have missed something.. I'll take a closer look 2013-09-05T01:59:15 can it work if without my patches? 2013-09-05T01:59:58 actually, I used to get a segfault before your patches, which is why I updated.. so I don't know :P 2013-09-05T02:01:02 you can use rtems-rap -a rap file to see whether the symbol is in the rap file 2013-09-05T02:01:12 rtems-rap -a x.rap 2013-09-05T02:02:43 The rtl-host has been updated, but rtl is under reviewing and checking, and patches have not been merged upstream. This may cause segfault 2013-09-05T02:03:03 yeah 2013-09-05T02:03:24 is rtems-rap -S supposed to show whether the symbol is defined in the rap file as well? 2013-09-05T02:03:53 yeah. 2013-09-05T02:04:03 -a shows the symbol. But I can't tell if it's defined or simply referenced 2013-09-05T02:04:08 -S does not show the symbol 2013-09-05T02:04:25 -a shows all the things in the rap file. 2013-09-05T02:05:15 the global variable in question is a struct pointer, if that makes any difference 2013-09-05T02:05:25 *** kiwichris has joined #rtems 2013-09-05T02:06:48 sebhub, will you commit the mrc/mcr patch ? 2013-09-05T02:07:09 the9a3eedi: if `sparc-rtems4.11-readelf -s *.o` can list the symbol and it's STB_BIND is GLOBAL, then it must be in rap file. 2013-09-05T02:09:31 definitely shows up in readelf -s 2013-09-05T02:09:51