2015-04-20T00:00:05 [trac] GSoC/2015 edited by punitvara (empty comment) http://devel.rtems.org/wiki/GSoC/2015 2015-04-20T00:07:45 *** edsiper_ is now known as edsiper 2015-04-20T00:22:01 *** javamonn has joined #rtems 2015-04-20T00:25:37 *** javamonn has quit IRC (Client Quit) 2015-04-20T00:29:44 *** monstr has joined #rtems 2015-04-20T00:35:35 *** zoso has joined #rtems 2015-04-20T01:08:26 *** ketul has joined #rtems 2015-04-20T01:24:23 [git] 26c142e by Sebastian Huber: [score: Refactor SMP cache manager support] http://devel.rtems.org/changeset/26c142e/rtems 2015-04-20T01:24:23 [git] 1981b46 by Sebastian Huber: [score: Add _ISR_lock_ISR_disable/enable()] http://devel.rtems.org/changeset/1981b46/rtems 2015-04-20T01:47:52 *** gigetoo has quit IRC (Ping timeout: 256 seconds) 2015-04-20T02:04:37 *** kiwichris has quit IRC () 2015-04-20T02:32:22 *** andrei_ has joined #rtems 2015-04-20T03:06:24 *** ketul has quit IRC (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client) 2015-04-20T03:11:44 *** ketul has joined #rtems 2015-04-20T03:40:05 beng-nl: hi 2015-04-20T03:40:48 beng-nl: did you look at work done ? 2015-04-20T03:45:05 *** ketul has quit IRC (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client) 2015-04-20T03:45:30 *** ketul has joined #rtems 2015-04-20T03:52:42 *** Hesham has joined #rtems 2015-04-20T04:21:26 *** zoso has quit IRC (Ping timeout: 256 seconds) 2015-04-20T06:24:25 ketul: no, is it on your github? 2015-04-20T06:24:55 should I mail it to you? 2015-04-20T06:24:59 github is better 2015-04-20T06:25:03 ya also on github 2015-04-20T06:25:30 beng-nl:and also mailed on devel mailing list 2015-04-20T06:26:07 beng-nl:you can find it here https://github.com/ketul93/RTEMS-on-BBB/tree/master/GPIO-Test 2015-04-20T06:27:12 i see it now 2015-04-20T06:27:28 i see the patch, good work! 2015-04-20T06:27:57 ya thanks. any comments from your side would be great 2015-04-20T06:28:51 it's a bit too 'narrow' (specific) to merge mainline 2015-04-20T06:28:53 if that is what you want 2015-04-20T06:49:07 *** andrei_ has quit IRC (Ping timeout: 246 seconds) 2015-04-20T06:51:02 beng-nl: that work I done is specifically for LEDs. But I can do it generic. 2015-04-20T06:51:16 I think now its easy to that. 2015-04-20T06:52:35 * to do that 2015-04-20T06:55:00 beng-nl: and have you seen that videos? 2015-04-20T07:09:49 *** Hesham has quit IRC (Quit: Leaving.) 2015-04-20T07:22:45 *** andrei_ has joined #rtems 2015-04-20T07:49:27 yes i did! 2015-04-20T07:52:50 *** Hesham has joined #rtems 2015-04-20T08:03:03 beng-nl: now what kind of changes you suggest to do.? 2015-04-20T08:04:07 what do you want to achieve? 2015-04-20T08:04:09 mainline it? 2015-04-20T08:05:00 ya as proposed in GSoC 2015-04-20T08:06:06 beng-nl: 2015-04-20T08:11:15 ketul: ok. so for that you'd have to implement it to a more generic rtems gpio api. if it exists. i was just looking.. 2015-04-20T08:11:19 have you? 2015-04-20T08:11:46 functionality with an interface that is specific to one bsp is much less valuable than a general one 2015-04-20T08:12:40 admittedly i haven't found a general gpio api 2015-04-20T08:15:20 beng-nl: general api means you want to have it for all boards that supports rtems? Actually I haven't got what you mean to say. 2015-04-20T08:15:59 i mean yeah, is there a way to control gpio's (an api) that is the same for all bsp's, so you could write the same application for multiple bsp's 2015-04-20T08:16:09 and not change the gpio calls 2015-04-20T08:17:31 Actually I can . But as I proposed for bbb . I am concentrating mainly on bbb. 2015-04-20T08:17:37 i understand 2015-04-20T08:17:53 i'm just suggesting that the interface in bbb should be like everywhere else. but it seems that doesn't exist. 2015-04-20T08:19:52 Technically saying for every board register address changes. So for that I need to have all register mapping files available with me. And For testing it I think I would have all the boards. Am I right? 2015-04-20T08:20:06 noo! 2015-04-20T08:20:11 that would be a lot to ask :) 2015-04-20T08:20:26 i just mean, for your bbb implementation, make the function calls the same as a general api, without touching any other bsp 2015-04-20T08:21:17 ketul: example: printf() works everywhere, there is no beagle_printf() 2015-04-20T08:21:20 beng-nl: Dr Joel said Andrei has done some work on GPIO API from last year. 2015-04-20T08:21:39 beng-nl: got 2015-04-20T08:21:39 i found it after googling but i found nothing in mainline 2015-04-20T08:21:40 While he was working on on RPi 2015-04-20T08:21:54 Maybe this can be a start point 2015-04-20T08:22:08 Hesham: ya I took reference from that 2015-04-20T08:22:22 We were discussing about general purpose GPIO API few days ago. 2015-04-20T08:22:45 *** gwty has joined #rtems 2015-04-20T08:23:23 ketul: So can you describe what has been done there? 2015-04-20T08:23:35 beng-nl: but how can I cat general purpose api? like printf() we know. But I want to ask for general purpose api what should be reference point? 2015-04-20T08:24:02 ketul: yeah, good point, i was assuming there was something. but maybe there isn't. 2015-04-20T08:26:21 Hesham: let me fetch the link what he done 2015-04-20T08:27:03 Hesham: https://asuolgsoc2014.wordpress.com/ 2015-04-20T08:28:01 beng-nl: andre's patch https://www.google-melange.com/gsoc/project/details/google/gsoc2014/andremarques/5668600916475904 2015-04-20T08:29:14 ketul: Thanks 2015-04-20T08:31:13 Hesham: np 2015-04-20T08:35:50 beng-nl: lets just consider a case. printf() is same for all the hardware architectures but its compiler specifies the proper address for the stdout. Am I right? 2015-04-20T08:36:13 *** andrei_ has quit IRC (Ping timeout: 246 seconds) 2015-04-20T08:38:33 i wouldn't but it like that, but.. yeah printf() is the same for all architectures, but the i/o (serial write) is bsp-specific 2015-04-20T08:45:33 beng-nl: I have a idea like to do in this fashion but not exactly. http://www.armhf.com/using-beaglebone-black-gpios/ 2015-04-20T08:46:30 beng-nl: i think that is kind of general purpose. Is it? 2015-04-20T08:52:10 i meant the API 2015-04-20T08:52:15 eh, never mind. i think it doesn't exist. 2015-04-20T08:55:34 ok,i got it. I think i should implement a function that can be called by any other program to access the resources 2015-04-20T08:56:23 *** gedare has joined #rtems 2015-04-20T08:56:23 *** ChanServ sets mode: +o gedare 2015-04-20T08:58:47 beng-nl: so what do you suggest to do next ? 2015-04-20T09:01:24 ketul: i'll make a post 2015-04-20T09:04:16 good morning 2015-04-20T09:04:20 hi gedare 2015-04-20T09:04:57 reading back your convo quick in the logger 2015-04-20T09:05:07 Morning gedare 2015-04-20T09:05:52 regarding the GPIO API, we never settled on a satisfactory one last yea 2015-04-20T09:05:52 r 2015-04-20T09:05:57 ok 2015-04-20T09:06:01 So we moved all the code into rpi bsp 2015-04-20T09:06:19 it would be really quite good to continue evolving both rpi and bb in the same direction 2015-04-20T09:06:25 and settle on a common api so we can generalize it 2015-04-20T09:07:21 in terms of the next steps from where you are ketul, you need to review all the style documentations so your code "looks right", and I want you to identify any 'upstream sources' where you copied code from.. whether that is a repo, or "just some code you found online" 2015-04-20T09:07:27 so we'll 'copy' the rpi api first? 2015-04-20T09:07:33 beng-nl: yes that is correct 2015-04-20T09:07:37 ok 2015-04-20T09:07:46 and this should be done consulting with andre marques 2015-04-20T09:08:04 he knows more clearly what is/isn't right about it and what reviewers said it needed 2015-04-20T09:08:27 some aspects that were missing as i recall are interrupt handling and some "set" management for multiple io pins 2015-04-20T09:09:18 ketul: now you should work on improving the code you have written, and write a blog post about your experience, and wait to see if you get accepted to GSoC/RTEMS or not. 2015-04-20T09:09:33 this is the "twiddling thumbs" phase of GSOC. 2015-04-20T09:09:44 :) 2015-04-20T09:11:38 gedare: ya sure. I really learned to much and always happy to do that. 2015-04-20T09:12:18 ok ketul just make sure you pay attention to the small details, they're often the most important. 2015-04-20T09:13:01 gedare: Is it OK to fetch Adapteva/Epiphany GNU tools from within RSB? 2015-04-20T09:13:31 Their upstream code is kinda outdated 2015-04-20T09:14:25 And they use their own Github repos as a base for their SDK tools, not upstream repos. 2015-04-20T09:15:59 gedare: sure. But can you tell me what kind of small aspects.. 2015-04-20T09:16:07 Hesham: regarding limits.h, that sounds suspiciously like a problem in newlib's configure/make files 2015-04-20T09:17:05 gedare: Yeah that's the only thing that stumble me from submitting Epiphany patches now. 2015-04-20T09:17:06 Hesham: it's better to encourage them to submit upstream ;) are their github repos sync'ed to release numbers of the GNU tools? 2015-04-20T09:17:22 ketul: why is the clock driver disabled in your test program? 2015-04-20T09:17:55 beng-nl: it is enabled I removed that comment in the gist. 2015-04-20T09:18:03 ko 2015-04-20T09:18:04 ok 2015-04-20T09:18:04 gedare: I've already pushed RTEMS support to their repos, except newlib 2015-04-20T09:18:06 ketul: I find that newer programmers tend to skim too much and miss important details, so just make sure you pay attention. 2015-04-20T09:18:22 you gotta learn to walk before you run 2015-04-20T09:18:29 and you gotta learn to crawl before you walk 2015-04-20T09:18:33 gedare: GCC 4.9, newlib 1.2.0, binutils-2.23 2015-04-20T09:18:48 Hesham: why are they not submitting up to FSF? 2015-04-20T09:18:53 1.2.0??? 2015-04-20T09:19:16 gedare: 1.20 2015-04-20T09:19:36 Never asked before. 2015-04-20T09:20:27 The RTEMS code is almost ready and it runs fine on their simulator. 2015-04-20T09:20:51 And of course the hardware. 2015-04-20T09:21:47 ketul: i posted some comments to the list 2015-04-20T09:22:05 gedare, Hesham: i hope you agree with the spirit & specifics 2015-04-20T09:22:14 gedare: Ah one issue, I have to force heap allocation to 8 bytes- alignment here -> _Heap_Allocate_aligned_with_boundary( heap, size, 8, 0 ); 2015-04-20T09:22:54 Hesham: please ask them to push upstream it is better for everyone 2015-04-20T09:23:30 gedare: OK, once I get newlib issue sorted out. 2015-04-20T09:23:33 is the minimum heap alignment 8? 2015-04-20T09:24:02 Originally the RTEMS call was _Heap_Allocate_aligned_with_boundary( heap, size, 0, 0 ); 2015-04-20T09:24:10 I changed it to > _Heap_Allocate_aligned_with_boundary( heap, size, 0, 0 ); 2015-04-20T09:24:11 hrm 2015-04-20T09:24:16 where? 2015-04-20T09:24:31 libcsupport/src/cachecoherentalloc.c ? 2015-04-20T09:24:38 heapimpl.h Line 137 2015-04-20T09:25:08 I have to do this due to some double alignment issue that Epiphany restricts. 2015-04-20T09:25:12 oh 2015-04-20T09:25:22 beng-nl: thanks 2015-04-20T09:25:43 gedare: I submitted a patch to their GCC repo earlier to fix this issue, but they said it will break the ABI 2015-04-20T09:25:46 if ( alignment == 0 ) { 2015-04-20T09:25:46 alignment = page_size; 2015-04-20T09:25:46 } 2015-04-20T09:25:51 uintptr_t const page_size = heap->page_size; 2015-04-20T09:26:03 back track why that is not good enough 2015-04-20T09:26:47 if ( page_size == 0 ) { 2015-04-20T09:26:47 page_size = CPU_ALIGNMENT; 2015-04-20T09:26:49 that's the problem 2015-04-20T09:26:56 *** geheimnis` has quit IRC (Ping timeout: 240 seconds) 2015-04-20T09:26:57 is CPU_ALIGNMENT < CPU_HEAP_ALIGNMENT? 2015-04-20T09:27:01 in your cpu.h? 2015-04-20T09:27:27 Nope 2015-04-20T09:27:28 Hesham: uintptr_t _Heap_Initialize() @ score/src/heap.c:204 2015-04-20T09:27:35 They are the same, 8 bytes 2015-04-20T09:27:42 then it should be getting 8 2015-04-20T09:27:54 when alignment is passed as 0 2015-04-20T09:28:00 it should be getting page_size out of Heap_Control 2015-04-20T09:28:04 so check that variable 2015-04-20T09:28:06 and see why isn't it 8 2015-04-20T09:28:19 *** geheimnis` has joined #rtems 2015-04-20T09:28:26 OK, let me debug again and see what page_size value is 2015-04-20T09:31:33 beng-nl: thanks perfect review comments summed up everything nicely. 2015-04-20T09:32:46 *** gwty has quit IRC (Ping timeout: 264 seconds) 2015-04-20T09:35:25 Hesham: since the port is somewhat immature and tools are not upstream, i'd hold off on submitting it until after we release. i expect 4.11 to come out this week or next 2015-04-20T09:35:29 or at least to get "cut" 2015-04-20T09:35:45 gedare: great! 2015-04-20T09:36:04 beng-nl: yeah, way past due, and i really want it done before gsoc starts 2015-04-20T09:36:22 yeah makes sense 2015-04-20T09:36:25 there hasn't been a release since just after i started working on rtems 2015-04-20T09:36:46 correlation = causation, right!? 2015-04-20T09:37:02 *** DrJoel has joined #rtems 2015-04-20T09:37:02 *** DrJoel has joined #rtems 2015-04-20T09:37:02 *** ChanServ sets mode: +o DrJoel 2015-04-20T09:37:03 *** Hesham has quit IRC (Ping timeout: 245 seconds) 2015-04-20T09:37:04 https://xkcd.com/552/ ;-) 2015-04-20T09:37:19 speaking of which, hi DrJoel, got your 'git release' gloves ready? 2015-04-20T09:37:28 except seb wants to merge some patch he sent this morning 2015-04-20T09:38:02 I have a small set of patches required to even build with C++ and networking enabled. Apparently no one else even builds their own BSPs. 2015-04-20T09:38:20 But yes. It is time to start the branch and release process. 2015-04-20T09:38:45 Yesterday I started a build/test run of ~20 BSPs on simulators. I want to see those finish first. 2015-04-20T09:40:03 gedare: Does all look OK to you for gsoc? 2015-04-20T09:40:29 DrJoel: yes until dedup runs we're good 2015-04-20T09:40:44 I just looked at the list and it appears to match my expectations as well. 2015-04-20T09:40:53 Right number of slots, etc. 2015-04-20T09:41:58 great. 2015-04-20T09:42:13 yes i'm satisfied too 2015-04-20T09:44:44 -rw-r--r-- 1 joel rtems 2557406136 Apr 20 09:44 arm-xilinx_zynq_a9_qemu.log 2015-04-20T09:44:51 That doesn't look like it is doing well. :) 2015-04-20T09:45:27 uhhh 2015-04-20T09:45:38 :-) 2015-04-20T09:45:51 apparently qemu in that configuration doesn't like being nohup'ed without a terminal. :( 2015-04-20T09:46:09 I guess I will have to run them one at a time in the foreground to get through them. 2015-04-20T09:48:04 might be another option 2015-04-20T09:48:21 but i'm not a qemu guru 2015-04-20T09:48:39 there might be but I want test results quickly. Long term, we have to figure this one out or we can't do automated testing on check in 2015-04-20T09:50:40 *** monstr has quit IRC (Ping timeout: 256 seconds) 2015-04-20T09:54:13 *** ketul has quit IRC (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client) 2015-04-20T09:58:57 DrJoel: ask sebastian 2015-04-20T09:59:34 DrJoel: we also will need a strategy for real board testing, the zynq is a good candidate 2015-04-20T09:59:36 I will run some tests and see what all simulators have issues 2015-04-20T09:59:40 sure. 2015-04-20T09:59:54 Yeah.. kiwichris needs to answer that 2015-04-20T10:06:08 *** Hesham has joined #rtems 2015-04-20T10:06:46 gedare: I don't know if this is a bug or not 2015-04-20T10:06:48 if ( alignment == 0 ) { 2015-04-20T10:06:48 alignment = page_size; 2015-04-20T10:06:48 } 2015-04-20T10:06:48 } 2015-04-20T10:07:14 This is not invoked unless boundary != 0 2015-04-20T10:07:37 So it's not aligned at all. 2015-04-20T10:09:36 where is this? 2015-04-20T10:09:51 _Heap_Allocate_aligned_with_boundary 2015-04-20T10:10:23 score/src/heapallocate.c:184 2015-04-20T10:10:38 i see Hesham so when boundary is 0... 2015-04-20T10:10:50 alloc_begin = _Heap_Alloc_area_of_block( block ); 2015-04-20T10:10:52 The code looks like it should be outside that if 2015-04-20T10:10:53 you get that result 2015-04-20T10:11:22 Hesham: you should post to devel to solicit sebh's input on this, it's his code 2015-04-20T10:12:47 but i agree, the minimum alignment on this should be page_size 2015-04-20T10:12:56 really, it should check if alignment < page_size 2015-04-20T10:13:05 i think? 2015-04-20T10:14:15 anyway the code definitely smells funny. 2015-04-20T10:16:11 gedare: Weird, when I moved it outside boundary block, the program doesn't work at all, seems like it get stuck at early init stages. 2015-04-20T10:16:47 Could be honoring the alignment request now and using a lot more memory. 2015-04-20T10:19:22 * Hesham Connecting to the Parallella board, coming back later. 2015-04-20T10:20:05 Hesham: moving it out makes dead code later, where alloc_begin is made 2015-04-20T10:20:16 my guess is there is some implicit assumption about this alignment=0, boundary=0 case 2015-04-20T10:21:38 Hesham: instead, you may want to add code in _Heap_Check_block to set alignment to page_size in case it is 0 (or less) 2015-04-20T10:22:18 well, actually 2015-04-20T10:23:07 Hesham: at score/src/heapallocate.c:233 try setting alignment = min(page_size, alignment) 2015-04-20T10:23:26 no, that won't work either... 2015-04-20T10:24:45 *** Hesham has quit IRC (Ping timeout: 276 seconds) 2015-04-20T10:27:58 gedare is there a wiki page about our licensing requirements/philosophy? 2015-04-20T10:28:03 hesham: it seems that boundary = 0, alignment = 0, is a special case for getting the first address in a block. the problem seems to be that (block + HEAP_BLOCK_HEADER_SIZE) is not aligned right. 2015-04-20T10:28:08 except hesham isn't here. 2015-04-20T10:28:37 DrJoel: i think there is a link to my blog post only, somewhere in coding conventions maybe 2015-04-20T10:28:58 I was wondering if it should be a topic in the developer part of the wiki 2015-04-20T10:29:01 it could be added here: https://devel.rtems.org/wiki/TBR/Website/License 2015-04-20T10:29:12 except for the TBR part 2015-04-20T10:29:17 DrJoel: for that, it is best to be at https://devel.rtems.org/wiki/Developer/Coding/Conventions#Licenses 2015-04-20T10:29:36 that's where it links to my blog post #rant 2015-04-20T10:30:21 Those bullets pretty much sum it up. 2015-04-20T10:30:39 Is the focus on "these are acceptable" ignoring that proprietary licenses are not acceptable? 2015-04-20T10:30:59 hm 2015-04-20T10:31:06 it's a lot easier to be inclusionarfy 2015-04-20T10:31:30 it might be good to put in a list of known-good licenses too i suppose 2015-04-20T10:31:36 and known-bad 2015-04-20T10:31:43 on a wiki page 2015-04-20T10:31:57 oh hey, is the doc/started good to go? 2015-04-20T10:32:01 This covers it pretty well. Just thinking ahead. 2015-04-20T10:32:15 I need to double check it. It is now on the list. 2015-04-20T10:38:00 *** Hesham has joined #rtems 2015-04-20T10:39:42 Did you see this gedare? http://pdos.csail.mit.edu/scigen/scipher.html 2015-04-20T10:41:21 *** ketul has joined #rtems 2015-04-20T10:45:51 are> hesham: it seems that boundary = 0, alignment = 0, is a special case for getting the first address in a block. the problem seems to be that (block + HEAP_BLOCK_HEADER_SIZE) is not aligned right. 2015-04-20T10:46:41 DrJoel: yes 2015-04-20T10:48:13 gedare: I found out that the problem happens only when starting allocating from RTEMS_Malloc_Area 2015-04-20T10:48:42 yeah it seems like it is a bit of code for a special case 2015-04-20T10:48:47 There's no problem with when allocating from _Workspace_Area 2015-04-20T10:48:57 are they unified? 2015-04-20T10:49:05 So #define CONFIGURE_UNIFIED_WORK_AREAS 2015-04-20T10:49:17 Will make it work 2015-04-20T10:49:30 I can't find where RTEMS_Malloc_Area should be defined. 2015-04-20T10:49:33 probably has something do with the alignment of the initial heap pointer then 2015-04-20T10:49:34 Or initialized 2015-04-20T10:49:54 *** gigetoo has joined #rtems 2015-04-20T10:50:25 I suspect so, but where it initial heap pointer get initialized? 2015-04-20T10:51:32 For RTEMS_Malloc_Area if that's different than _Work_Area 2015-04-20T10:52:30 bsp_work_area_initialize? 2015-04-20T10:52:59 That's for workspace I think 2015-04-20T10:53:42 bsp_work_area_initialize_default @ libbsp/shared/include/bootcard.h 2015-04-20T10:53:56 RTEMS_Malloc_Initialize(&area, 1, NULL); 2015-04-20T10:54:16 hmm 2015-04-20T10:55:04 ok i'm not sure then 2015-04-20T10:55:08 [trac] #2331/defect (v:4.10) created by dcb (c/src/lib/libbsp/sparc/shared/1553/gr1553b.c: 2 * memory leak ?) http://devel.rtems.org/ticket/2331 2015-04-20T11:04:11 [trac] #2331/defect (assigned) updated by gedare (empty comment) http://devel.rtems.org/ticket/2331 2015-04-20T11:04:57 [trac] #2331/defect (assigned) updated by joel.sherrill (I believe case 1 is a leak on the error case. Case 2 looks OK. The two malloc'ed ...) http://devel.rtems.org/ticket/2331 2015-04-20T11:05:10 Don't know, the previous problem was that when creating a new task from ticker/init.c it was using _Heap_Allocate_aligned_with_boundary( heap, size, 0, 0 ); 2015-04-20T11:05:19 And the allocated block was already used 2015-04-20T11:05:37 This was solved when adding _Heap_Allocate_aligned_with_boundary( heap, size, 8, 0 ); alignment 2015-04-20T11:06:33 So the problem occurs now when only using RTEMS_Malloc_Area 2015-04-20T11:06:55 And that's an alignment problem also... 2015-04-20T11:24:42 *** Hesham has quit IRC (Ping timeout: 265 seconds) 2015-04-20T11:54:23 DrJoel: great and thanks. I am looking to your ppt. 2015-04-20T11:54:26 *** Hesham has joined #rtems 2015-04-20T12:06:25 *** Hesham has quit IRC (Ping timeout: 264 seconds) 2015-04-20T12:10:05 *** dr__house|away is now known as dr__house 2015-04-20T12:16:34 ketul: glad to pass it along. The commands are pretty handy. The API to the multi IO was a prototype. Figuring out how to tie those commands generically to a board will give some neat commands and probably help flesh out the APIs 2015-04-20T12:17:22 *** Hesham has joined #rtems 2015-04-20T12:26:13 DrJoel: Thanks for that. I will surely look into it and will have nice output. Actually gedare pointed put about multiio. But I didn't found any documents related to that. But this ppt will help me a lot. 2015-04-20T12:26:25 *out 2015-04-20T12:32:08 I thought I wrote real docs on it for the shell manual. Maybe I just did them in Doxygen. 2015-04-20T12:33:27 ok 2015-04-20T12:54:25 *** andrei_ has joined #rtems 2015-04-20T12:55:27 hello! 2015-04-20T12:57:59 *** gedare has quit IRC (Ping timeout: 264 seconds) 2015-04-20T12:59:17 hey andrei_ 2015-04-20T13:02:59 sorry if it's not appropriate, but are the slots(students) already decided? 2015-04-20T13:03:02 except deduplication 2015-04-20T13:07:55 I think we can say mostly yes to that. There is still time to impress us or commit a crime and disqualify yourself. :) 2015-04-20T13:12:39 *** gwty has joined #rtems 2015-04-20T13:13:25 out of curiosity, is the number of students higher than before 2015-04-20T13:13:34 as the number of orgs seems to be much smaller? 2015-04-20T13:13:45 Sorry, the number of given slots 2015-04-20T13:15:25 We are on the high side of what we have had in the past. I haven't heard much about other organizations. 2015-04-20T13:17:03 awesome 2015-04-20T13:18:17 if you got time maybe you can have a look at the condvar_wait() pseudocode I wrote here https://docs.google.com/document/d/16EBmWcQw9YccUycofJIlwF98p21PQPq-EJXupes-K80/edit 2015-04-20T13:18:31 Our main concern has been figuring out who to accept and how to best mentor them. :) 2015-04-20T13:19:04 I will look at it. Give me a few minutes. Generating PDFs for an RTEMS class next week 2015-04-20T13:19:12 sure thing 2015-04-20T13:19:43 Did you ever look at the multio git repo? That was my prototype a few years ago for a discrete and analog IO API plus some shell commands. 2015-04-20T13:19:59 I would love to see the shell commands updated and merged. 2015-04-20T13:20:30 hello DrJoel 2015-04-20T13:20:44 DrJoel, do you happen to have the link? 2015-04-20T13:20:51 I didn't look over it, but I can do so now 2015-04-20T13:21:07 https://git.rtems.org/multiio/tree/ 2015-04-20T13:21:45 thanks, let me have a look 2015-04-20T13:22:23 The important parts are the commands and include directory. pcmmio and rtd6425 are specific IO boards for embedded PCs. I still have the hardware but who cares at this point. 2015-04-20T13:32:31 did you get my reply in the proposal drjoel? 2015-04-20T13:35:18 DrJoel: you were talking about slides 54 ,..,.. where given PCMMIO Shell Commands .? 2015-04-20T13:36:46 yes.. the commands start at 54. 2015-04-20T13:36:54 gwty// fuzzy headed.. probably 2015-04-20T13:39:19 DrJoel: pcmmio stands for? 2015-04-20T13:40:17 ketul: pcm-mio is a board, but I m not sure it's coming from that 2015-04-20T13:40:31 http://www.winsystems.com/datasheets/PCM-MIO-G-DS.pdf 2015-04-20T13:40:43 It is coming from that. Poor name. PC multi IO. 2015-04-20T13:43:16 ok got. DrJoel so your suggestion is to use same function names while designing gpio driver for beagle. Am I right? Is it like beagle_din or something? 2015-04-20T13:43:51 *** gedare has joined #rtems 2015-04-20T13:43:52 *** ChanServ sets mode: +o gedare 2015-04-20T13:52:16 *** linuxkernelhacke has joined #rtems 2015-04-20T13:56:17 *** ragunath has joined #rtems 2015-04-20T14:08:45 Finally I figured out what's wrong with limits.h. 2015-04-20T14:08:59 t-rtems wasn't included in gcc tmake_files 2015-04-20T14:09:18 Good catch. 2015-04-20T14:09:39 DrJoel: I got some custom RSB for building Epiphany (from Adapteva repos) 2015-04-20T14:09:52 Should I submit it? 2015-04-20T14:10:12 Please post it. Are all the patches upstream? 2015-04-20T14:10:13 custom RSB patch* 2015-04-20T14:10:41 I will have to submit the issue of t-rtems 2015-04-20T14:10:53 to Adapteva gcc repo 2015-04-20T14:11:23 Currently I am fetching my GCC github fork with this fix 2015-04-20T14:11:32 - newlib from upstream 2015-04-20T14:11:50 - binutils works fine from their Adapteva repo 2015-04-20T14:16:59 By upstream do you mean Adapteva/upstream or the GNU tools upstream repos? 2015-04-20T14:17:33 I guess at this point either. Do they have plans to submit to FSF? 2015-04-20T14:18:37 ketul, andrei_: The key to a multi-IO (e.g. discrete IO, ADC, and DAC) API set from the user application perspective is to not change code when you change hardware platforms. 2015-04-20T14:18:48 So having the name of a board in the API is bad. 2015-04-20T14:19:01 DrJoel: Not sure, gedare asked that I push them to push the latest work upstream 2015-04-20T14:19:30 Even if they are lazy to push upstream, can't we work with their repos? 2015-04-20T14:19:33 https://git.rtems.org/multiio/tree/include/multiio.h shows how I was trying to make the user side of the API target independent 2015-04-20T14:19:46 We can work with their repos. Long term, we push. :) 2015-04-20T14:20:29 DrJoel: yep 2015-04-20T14:21:43 DrJoel: Support for building RTEMS from their repos has been already merged, so it's now their job to push it upstream 2015-04-20T14:21:57 along with their latest changes. 2015-04-20T14:23:33 Yep 2015-04-20T14:24:42 DrJoel: so instead of using gpio_set() I should use rtems_set_dout(). So it can be generic for the user (thats what api means). So like its all about function name. so we can have generic api. Am I right? 2015-04-20T14:26:25 im having a bit of trouble installing waf 2015-04-20T14:26:29 there's no upstream package on ubuntu 2015-04-20T14:26:42 and I haven't yet figured how to proprely get the command to work 2015-04-20T14:26:46 There is a user api (names to be community decided) which is used by any application. How that attaches under the hood to the BSP is something to discuss with andrei_ 2015-04-20T14:27:14 Install it from their upstream code. waf shouldn't really be packaged. It is tied to a program being built. 2015-04-20T14:30:09 *** hermann_19829 has joined #rtems 2015-04-20T14:33:28 DrJoel : what you mentioned above is condvar classic api? 2015-04-20T14:34:51 *** ankur has joined #rtems 2015-04-20T14:38:08 *** ketul has quit IRC (Excess Flood) 2015-04-20T14:38:58 *** ketul has joined #rtems 2015-04-20T14:39:58 andrei_: "install' waf by copying the waf file into your $PATH 2015-04-20T14:40:16 andrei_: no, but it does apply there. DrJoel is confusing you with Andre ;) 2015-04-20T14:40:41 ah 2015-04-20T14:48:21 DrJoel: I have seen andre's patch & work and also done communication for getting my prob solved. 2015-04-20T14:57:33 *** ragunath has left #rtems 2015-04-20T15:08:57 DrJoel: Thanks for the feedback on proposal, got my hands full now 2015-04-20T15:08:58 also 2015-04-20T15:09:11 there is the waf install line 2015-04-20T15:09:12 --rtems-tools=$HOME/Development/rtems/4.11 \ 2015-04-20T15:09:24 I cloned rtems-tools 2015-04-20T15:09:54 andrei_ Posted some comments. 2015-04-20T15:10:02 DrJoel: yeah, I noted, thanks a lot 2015-04-20T15:10:05 + install /home/andrei/gsoc/rtems-source-builder/rtems/development/rtems/4.11/share 2015-04-20T15:10:28 beng-nl.. Am I wrong or are you and andrei_ at the same university? I am not up to speed today so may have screwed that up. 2015-04-20T15:10:28 im at a bit of a loss of what do I have to give 2015-04-20T15:10:30 to waf configure 2015-04-20T15:11:27 https://git.rtems.org/multiio/tree/README.waf has some instructions,. Look at 5 and 6. 2015-04-20T15:12:07 Which git repo are you looking at waf on? 2015-04-20T15:12:10 yeah, I m following that 2015-04-20T15:12:16 the waf worked ok on rtems-tools 2015-04-20T15:12:24 so im looking at multiio 2015-04-20T15:13:28 RTEMS tools path does not contain a 'bin' directory: /home/andrei/gsoc/rtems-source-builder/rtems/build 2015-04-20T15:13:31 andrei_: rtems-tools= means where did you install the rtems gcc/binutils? 2015-04-20T15:13:48 it is the prefix you used for installing tools from rsb 2015-04-20T15:14:35 *** ankur has left #rtems 2015-04-20T15:14:48 *** ankur has joined #rtems 2015-04-20T15:18:46 DrJoel: reportedly yes, but reportedly not yet, i thought? 2015-04-20T15:18:50 andrei_: are you @ VU yet? 2015-04-20T15:19:19 beng-nl : nope, starting this autumn 2015-04-20T15:19:32 okay 2015-04-20T15:19:36 andrei_: look me up when you get here :) 2015-04-20T15:19:50 haha, so you re at vu amsterdam too? awesome 2015-04-20T15:19:54