2013-05-20T00:07:04 *** monstr_ has joined #rtems 2013-05-20T00:13:25 *** mgahlot|away has joined #rtems 2013-05-20T01:19:45 *** arvind_khadri has joined #rtems 2013-05-20T02:06:28 *** mgahlot|away has quit IRC (Remote host closed the connection) 2013-05-20T02:40:22 *** mgahlot|away has joined #rtems 2013-05-20T03:30:31 *** monstr_ has quit IRC (Ping timeout: 276 seconds) 2013-05-20T04:11:54 *** MegaAlex|away is now known as MegaAlex 2013-05-20T04:36:32 *** S_Somani has joined #rtems 2013-05-20T04:39:18 *** mgahlot|off has joined #rtems 2013-05-20T04:39:52 *** S_Somani has left #rtems 2013-05-20T04:42:25 *** mgahlot|away has quit IRC (Ping timeout: 248 seconds) 2013-05-20T05:10:13 *** MegaAlex is now known as MegaAlex|away 2013-05-20T05:22:00 *** freenix has joined #rtems 2013-05-20T06:22:57 antgreen, are you around ? 2013-05-20T06:23:48 kiwichris: you just caught me. I'm packing for a trip right now, but I'll have time to look into the toolchain issue this week 2013-05-20T06:24:12 great and thanks, the current is I have these patches ... 2013-05-20T06:24:18 like, tomorrow morning 2013-05-20T06:24:38 http://git.rtems.org/rtems-source-builder/tree/rtems/config/4.11/rtems-moxie.bset?id=06c02001ea105fe2e05f1f7c431bcd49fd33fada 2013-05-20T06:25:03 http://git.rtems.org/rtems-tools/tree/tools/4.11/binutils/moxie 2013-05-20T06:25:04 ok, bookmarked 2013-05-20T06:25:17 and I seem to still be a -EB error in libgcc 2013-05-20T06:25:22 thanks 2013-05-20T06:25:35 have a good trip 2013-05-20T06:26:04 hmm.. but I see the -EB patch in gas 2013-05-20T06:26:46 so you are taking binutils 2.23.2 and patching it with those two patches? 2013-05-20T06:27:42 it's a business trip, which is always a great time for me to hack :-) so I'm looking forward to it 2013-05-20T06:29:00 yes that is correct. we assumed this was missing and the patches apply without error 2013-05-20T06:31:24 Hmm I had an old version in the prefix, I wonder if this was being picked up 2013-05-20T06:33:36 probably 2013-05-20T06:36:11 It should cause the prefix is not in the path 2013-05-20T06:36:17 sorry should not 2013-05-20T06:43:27 I believe the compiler will look in the prefix in addition to the path 2013-05-20T06:43:34 during build time 2013-05-20T06:46:51 Before the path ? 2013-05-20T06:48:21 *** mgahlot|off has quit IRC (Remote host closed the connection) 2013-05-20T06:49:35 yes it worked once I had cleaned it up. I will need to consider this a little more 2013-05-20T08:10:45 *** gedare has joined #rtems 2013-05-20T08:19:19 good morning 2013-05-20T08:28:54 *** monstr_ has joined #rtems 2013-05-20T09:00:53 hi, where can i download the moxie tools? It seems the url on the moxie wiki page is wrong. 2013-05-20T09:01:57 freenix: does rtems-source-builder have it? 2013-05-20T09:02:30 I have not tried to build the tools from source. 2013-05-20T09:02:53 But There are no rpm packages 2013-05-20T09:03:00 then i think you have to build from source 2013-05-20T09:03:18 moxie is new in rtems 2013-05-20T09:03:40 ok. 2013-05-20T09:25:13 *** weiY has joined #rtems 2013-05-20T09:33:13 *** DrJoel has joined #rtems 2013-05-20T09:33:14 *** DrJoel has joined #rtems 2013-05-20T09:33:14 *** ChanServ sets mode: +o DrJoel 2013-05-20T09:37:54 *** weiY has quit IRC (Ping timeout: 252 seconds) 2013-05-20T09:40:23 *** embz has joined #rtems 2013-05-20T09:40:57 *** weiY has joined #rtems 2013-05-20T09:49:51 gedare.. are you recovered from all the graduation celebrations? 2013-05-20T09:53:16 hey DrJoel i am..getting there :) 2013-05-20T09:54:44 :) 2013-05-20T09:55:10 You mentioned the strict order has a priority inversion possibility. But doesn't the current simpler algorithm also suffer from one 2013-05-20T09:55:17 Just better defined :) 2013-05-20T09:55:21 *** freenix has quit IRC (Ping timeout: 240 seconds) 2013-05-20T09:55:27 the current algorithm doesn't have a true priority inversion 2013-05-20T09:55:53 or, maybe it does 2013-05-20T09:56:00 it just seems less likely to me to be a problem 2013-05-20T09:56:12 just holds the elevated priority too long.. I agree.. much less likely to be a problem 2013-05-20T09:56:22 got a minute to chat gsoc students? 2013-05-20T09:56:29 where? 2013-05-20T09:56:42 and, not really.. i need to get back to work 2013-05-20T09:57:05 OK. I am flip flopping on a couple of students and want to get a sanity check 2013-05-20T09:57:12 ok pm me 2013-05-20T10:01:50 gedare, about the strict order the mutex must be acq/release using LIFO order 2013-05-20T10:02:05 and the better way is not use this LIFO order 2013-05-20T10:02:18 about the solution i also post a mail about this issue 2013-05-20T10:02:21 weiY: yes I know. that is not such a problem, although if you can eliminate the requirement while correcting the bug it would be good 2013-05-20T10:02:34 you suggested a heap-based approach to tracking the lock holders? 2013-05-20T10:02:58 yeah, or red-black tree 2013-05-20T10:03:39 and i donot really look deep into the mail, maybe i am misunderstand the problem 2013-05-20T10:04:05 ok. well, now i have provided the test case to show the bug :) 2013-05-20T10:04:36 ok, i am now reading your mail carefully 2013-05-20T10:04:38 it would be good to propose a design for this, it can be considered in the scope of your GSoC if you would like. 2013-05-20T10:04:39 Hi 2013-05-20T10:04:42 hello embz 2013-05-20T10:04:51 I've pulled rtems 4.11 2013-05-20T10:05:01 performed the bootstrap 2013-05-20T10:05:25 in the bootsrap procedure this was common "acinclude.m4 is unchanged" 2013-05-20T10:05:31 several lines showing that 2013-05-20T10:05:33 yeah, i am also study the linux priority-inversion mutex design and other materials. later i will summer it again 2013-05-20T10:05:54 the make is suprizingly fast 2013-05-20T10:06:09 embz: i think that is ok. it just means that file was already in place.. make should not be that fast, a couple minutes i would guess 2013-05-20T10:06:16 after make install i get this in the install dir 2013-05-20T10:06:20 i would suspect it did not make correctly :) 2013-05-20T10:06:26 /opt/rtems-4.11$ ls 2013-05-20T10:06:26 arm-rtems4.11 bin include lib libexec make share 2013-05-20T10:06:35 yep it didn't 2013-05-20T10:06:39 weiY: ok good. you might be able to find some academic article that describes this issue. 2013-05-20T10:07:03 embz: how did you "configure" RTEMS? And, are you following a tutorial / guide? 2013-05-20T10:07:08 yep 2013-05-20T10:07:13 I configured rtems 2013-05-20T10:07:29 ../rtems/configure --target=arm-rtems4.11 --enable-rtemsbsp=arm --enable-tests=samples 2013-05-20T10:08:07 and your mean "that is not such a problem" is not the same as the mail i post in the past? 2013-05-20T10:08:27 that is not a valid --enable-rtemsbsp argument. It must be a valid BSP name. Do you have real hardware? 2013-05-20T10:08:48 i see a mail "Semaphore Issue", is it this issue? 2013-05-20T10:08:57 weiY: i think most users can accept LIFO semantics for releasing semaphores. 2013-05-20T10:09:19 at least, if they need to avoid this PI/PC issue 2013-05-20T10:09:26 weiY: yes 2013-05-20T10:09:31 we have hardware 2013-05-20T10:09:38 ok, i know it is not the issue of LIFO order 2013-05-20T10:09:40 embz: you should identify the correct BSP for your hardware 2013-05-20T10:09:45 raspberies and calray 2013-05-20T10:09:54 ok 2013-05-20T10:10:34 I thought that stating arm I'd get all bsp's built 2013-05-20T10:10:38 weiY: you may like to see http://www.rtems.org/pr2124 2013-05-20T10:10:49 embz: if you leave the option out it will build all ARM bsp's 2013-05-20T10:10:55 +1\ 2013-05-20T10:11:00 oh nice 2013-05-20T10:11:08 ../rtems/configure --target=arm-rtems4.11 2013-05-20T10:11:11 that will build everything. 2013-05-20T10:11:19 (for the ARM) 2013-05-20T10:11:26 What are you trying to do? raspberrypi is the name of the Pi BSP. 2013-05-20T10:11:44 well, not everything, but it will build the samples, like you wanted. 2013-05-20T10:11:58 DrJoel: check pm 2013-05-20T10:12:05 DrJoel: fix your irc client ;) 2013-05-20T10:12:37 ok, i just want to find whether there is a PR. 2013-05-20T10:12:54 pr2124 2013-05-20T10:12:59 i just filed it yesterday 2013-05-20T10:13:00 well, the end goal is to code a scheduler using the "plugable" interface I think is in place 2013-05-20T10:13:06 just to assess overheads 2013-05-20T10:13:21 (I'm not from industry,) 2013-05-20T10:13:28 where are you from? 2013-05-20T10:13:43 i'm a student 2013-05-20T10:14:37 gedare: what's broken in my irc client today? 2013-05-20T10:15:32 DrJoel: i just mean it does not seem to indicate your PMs to you very well :) 2013-05-20T10:15:46 weiY: did you find the PR2124? 2013-05-20T10:16:22 what is a better choice between fix this bug based on strict order code or to aviod strict order using a new design? 2013-05-20T10:16:27 yeah, i have found it 2013-05-20T10:17:04 the fix to this bug must not undo the "stepping down" of priorities when releasing a lock while holding another lock. 2013-05-20T10:17:46 It just needs to implement this "step down" correctly, so that the thread holds the highest priority among its own priority and all of the threads that are blocked on the semaphores that it holds. 2013-05-20T10:18:25 and, if strict order helps make that algorithm more efficient, you can assume strict order 2013-05-20T10:18:45 but if you can implement the algorithm without assuming strict order, that would be good too... 2013-05-20T10:19:06 i think if using strict order the implement should be more easy, hehe 2013-05-20T10:19:25 but if i can avoid the strict order it should be more better. 2013-05-20T10:20:33 and firstly i can consider fix this bug under the strict order 2013-05-20T10:21:17 and then consider to avoid it. but i should think the design more carefully 2013-05-20T10:23:26 ok 2013-05-20T10:24:16 i'm glad you're looking at it. the bug is an important one to fix, and it is an interesting problem to consider :) 2013-05-20T10:24:50 embz: are you planning to implement a custom scheduler, or to add overhead timing to the existing one? 2013-05-20T10:25:08 custom one 2013-05-20T10:25:38 it's a simple limited preemptive scheduler 2013-05-20T10:26:13 ok.. be certain if you publish any timing information that you specify that you implemented a custom scheduler. 2013-05-20T10:26:29 :) 2013-05-20T10:26:41 where do you go to school? 2013-05-20T10:26:51 people are really particular about this performance figures 2013-05-20T10:27:02 I'm from Porto University 2013-05-20T10:27:05 in Portugal 2013-05-20T10:27:11 yeah, i know. and i am also interested in this problem. in the past few day i am study the linux rt-mutex design. and the best way to fix this issue is to inherent the right priority of hang mutex 2013-05-20T10:27:52 embz: ok cool. it is important to be quite clear about how performance measurements are obtained 2013-05-20T10:28:10 yep 2013-05-20T10:28:24 the strict order mutex just simplify the inherence algorithm 2013-05-20T10:28:24 by the way 2013-05-20T10:28:43 weiY: correct, but how to propagate that priority is tricky, especially when there is a chain of priority inheritance 2013-05-20T10:28:43 I've seen that the framework for pluging schedulers is ready for single core 2013-05-20T10:28:53 but is it also available for muti core 2013-05-20T10:28:59 when you have a global queue? 2013-05-20T10:29:02 embz: We are sensitive to published benchmarks. We have seen incorrect ones published in the past. We don't mind reviewing anything to help ensure they are correct. 2013-05-20T10:29:03 embz: yes. you can see the smpsimplescheduler 2013-05-20T10:29:18 embz: it is not a "real-time" scheduler for SMP 2013-05-20T10:29:22 embz: It is the same interface for uni and multiprocessor 2013-05-20T10:29:45 for now :) 2013-05-20T10:30:05 and "simple" indicates just that.. it is intended to be simple. It is not designed for efficiency. 2013-05-20T10:30:15 And neither is simplepriority for that matter 2013-05-20T10:30:52 a basic idea is that use two data struct, one stores the priority of hang mutex within the thread. the other one store the highest priority of thread hang on the same mutex 2013-05-20T10:30:55 why is it not a real-time sched for smp? 2013-05-20T10:31:17 It is at least O(n) 2013-05-20T10:31:25 by simple I mean that it is a kind of straightforward extension to regual fixed task priority 2013-05-20T10:31:33 It behaviorally is doing the right thing but not doing it efficiently 2013-05-20T10:31:54 it is. it has the same complexity as fixed priority preemptive 2013-05-20T10:31:59 It is preemptive, priority with time-slicing 2013-05-20T10:32:21 No.. Our fixed priority preemptive is fixed in the number of instructions to block or unblock. It is O(constant) 2013-05-20T10:33:13 I'm assuming that you need to traverse the ready-Q to decide what's the highest priority task ready to execute 2013-05-20T10:33:53 and that n was the task-set size 2013-05-20T10:33:56 that would be an incorrect assumption. We use a bitmap and an array of FIFOs. There is no walking or loop 2013-05-20T10:34:08 ok 2013-05-20T10:34:30 so this sched policy only has to look at a data entry which can be in the task control block 2013-05-20T10:34:35 and then set up a timer 2013-05-20T10:34:39 The Simple Priority Scheduler is the way you describe. A single sorted list .. it is O(tasks) to block but O(constant) to unblock 2013-05-20T10:34:50 No timer required unless you are delaying. Everything is event driven 2013-05-20T10:34:56 yes 2013-05-20T10:35:10 in thsi sched policy you delay tasks as much as possible 2013-05-20T10:35:21 *in this 2013-05-20T10:35:24 please read http://rtems.org/onlinedocs/doc-current/share/rtems/html/c_user/Scheduling-Concepts.html#Scheduling-Concepts 2013-05-20T10:36:01 thanks gedare 2013-05-20T10:36:22 perhaps you are mixing what we have implemented in RTEMS with the "simple limited" that you plan to implement. 2013-05-20T10:37:39 perhaps 2013-05-20T10:38:16 nevetheless only at release of a job do I need to set up a timer 2013-05-20T10:38:39 there is the constant bandwidth server that does something similar you may like to study it. 2013-05-20T10:38:51 it uses timers to avoid letting a task overrun its budget 2013-05-20T10:39:03 so the remainder scheduling code is not to be changed 2013-05-20T10:39:11 yes 2013-05-20T10:39:27 then you can probably just pick the existing functions and provide your own function for the release hook. 2013-05-20T10:39:45 this only works for periodic tasks or tasks scheduled by a server though. 2013-05-20T10:39:52 since normal tasks do not "release" more than one time. 2013-05-20T10:40:07 *** arvind_khadri has quit IRC (Ping timeout: 264 seconds) 2013-05-20T10:40:48 lets see 2013-05-20T10:40:58 I need to first get a hold of the rtems environment 2013-05-20T10:41:05 and test some simple stuff 2013-05-20T10:41:21 before getting into the more complicated details 2013-05-20T10:41:22 :) 2013-05-20T10:41:28 yup. you might want to try a simulator first 2013-05-20T10:41:39 for the ARM you can use the realview BSP on Qemu 2013-05-20T10:41:45 yep 2013-05-20T10:41:58 definitely better to try that first 2013-05-20T10:46:46 gedare: duplicated your core dump on the scheduler simulator :( 2013-05-20T10:47:00 well, i'm glad to hear it is not my problem ;) 2013-05-20T10:47:16 Don't blame it on me yet :) 2013-05-20T10:47:26 pfft, who else wrote code for schedsim?? 2013-05-20T10:47:33 everybody :) 2013-05-20T10:47:36 hah 2013-05-20T10:47:54 i guess in theory it could be unveiling a latent bug in rtems 2013-05-20T10:48:10 I just glued existing code together. The _Scheduler_priority_Ready_queue_first() call is returning NULL. 2013-05-20T10:48:16 Could also be a violation of some assumptino 2013-05-20T10:48:25 if you "watch" the Thread_Heir you should see where it gets assigned NULL 2013-05-20T10:48:34 That's how I got that information 2013-05-20T10:48:34 hmm. 2013-05-20T10:48:49 right, don't know why i'm advising you on debugging. I guess it is just overflow from last week. 2013-05-20T10:49:06 the Idle thread should be in the Ready Queue. 2013-05-20T10:49:56 I am glad we got ddd working :) 2013-05-20T10:50:14 yeah. even though kiwichris says it is only working by coincidence! 2013-05-20T10:50:31 and, it does seem brittle, but at least they have something to work with. 2013-05-20T10:51:18 That was the point of my wanting testing requirements on the rtl project. If you can't load two modules and debug the base and the dynamically loaded, then we have issues :) 2013-05-20T10:52:32 chris explained it pretty well in the last email i saw from him there is a lot of complications with the "proper" way to debug dynamically loaded modules 2013-05-20T10:52:39 it is beyond the scope of gsoc 2013-05-20T10:53:12 but basically we need to tell gdb how to read a struct that contains debug info for the modules as they load. 2013-05-20T10:53:42 That has to be fixed. It is a bad problem for production use. :( 2013-05-20T10:54:43 well, i don't think chris feels comfortable yet with the state for true production anyway 2013-05-20T10:55:00 that's why this gsoc..it is getting the rest of the archs in order 2013-05-20T10:55:21 probably someone will have to pay for debugger support ;) 2013-05-20T10:55:26 I think he has some upcoming work from Alan. Hopefully that will address this issue 2013-05-20T11:06:07 gedare .. I think there is a bug in your script. The Idle task is blocking on a semaphore 2013-05-20T11:06:19 hmmm 2013-05-20T11:06:35 hmm not being able to debug when using rtems linker is bad 2013-05-20T11:06:42 peerst +1 :) 2013-05-20T11:06:54 is there a workaround? 2013-05-20T11:07:09 peerst: you can debug RTEMS, but the application modules are not getting the source-level debug to work. the workaround is not to load the RTEMS image in the debugger, but instead to load the symbols of the module 2013-05-20T11:07:17 so, you can debug either RTEMS or application modules, but not both 2013-05-20T11:07:30 oh ok 2013-05-20T11:07:36 thats better than nothing 2013-05-20T11:07:36 this is not so bad, just cumbersome, and frail 2013-05-20T11:07:46 We need to discuss this with Chris to see if we can have a better solution 2013-05-20T11:08:00 but can't I just load symbold from multiple modules in gdb? 2013-05-20T11:08:03 chris plans to eventually store all the debug information in a struct that can be parsed by gdb 2013-05-20T11:08:11 yes that is what you have to do 2013-05-20T11:08:19 but it does not work if the RTEMS image's symbols are loaded 2013-05-20T11:08:21 I mean at the same time 2013-05-20T11:08:32 because it defines a symbol that overlaps the module's address range 2013-05-20T11:08:41 so gdb only uses the symbol given by RTEMS 2013-05-20T11:08:49 hmm 2013-05-20T11:08:55 gedare.. use -v when running the simulator and you get the command lines with numbers in the output: 2013-05-20T11:08:56 Thread Heir: 0x09010001 priority=255 2013-05-20T11:08:56 Thread Executing: 0x09010001 priority=255 2013-05-20T11:08:56 9: semaphore_obtain SEM2 0 2013-05-20T11:08:56 Obtain semaphore (0x1a010002) with timeout 0 2013-05-20T11:08:56 Breakpoint 1, rtems_semaphore_obtain (id=436273154, option_set=0, timeout=0) 2013-05-20T11:08:57 at /home/joel/rtems-4.11-work/rtems//cpukit/rtems/src/semobtain.c:51 2013-05-20T11:08:57 51 the_semaphore = _Semaphore_Get_interrupt_disable( id, &location, &level ); 2013-05-20T11:08:58 (gdb) c 2013-05-20T11:08:58 Continuing. 2013-05-20T11:08:59 Program received signal SIGSEGV, Segmentation fault. 2013-05-20T11:09:10 you can add-symbol-file on each module individually though to get all of your modules 2013-05-20T11:09:27 i see 2013-05-20T11:09:29 Notice that right before line 9 in your script, idle was running. 2013-05-20T11:09:30 oh all modules together OR rtems 2013-05-20T11:09:37 peerst: correct 2013-05-20T11:09:42 brb 2013-05-20T11:10:07 hmm maybe a fake rtems module that has only the symbols (sans the overlapping) 2013-05-20T11:10:41 or remove the overlapping symbols from rtems? 2013-05-20T11:10:43 maybe there is an easy way to prune the symbols out of the RTEMS image 2013-05-20T11:11:06 byb 2013-05-20T11:11:15 i think chris has a better way planned eventually. 2013-05-20T11:11:17 later weiY 2013-05-20T11:12:04 objcopy --strip-symbol 2013-05-20T11:13:11 gedare: or maybe weaken all symbols in rtems 2013-05-20T11:13:25 can be done by objcopy too 2013-05-20T11:13:40 maybe that can work, that would be better if done in post-processing by objcopy 2013-05-20T11:13:47 i dunno for sure 2013-05-20T11:13:58 DrJoel: i don't know why init is grabbing the mutex 2013-05-20T11:14:03 gedare: what symbols overlap the modules? 2013-05-20T11:14:03 or, rather, idle 2013-05-20T11:14:10 are they identifyable? 2013-05-20T11:14:32 they looked liek it in the version i was seeing... some kind of generated symbol that reserves space for the module to be loaded into i think 2013-05-20T11:14:41 i don't really know for sure 2013-05-20T11:14:54 so they can be matched by a pattern? 2013-05-20T11:14:56 it's just a single symbol that covers the entire range the module ends up getting loaded in to 2013-05-20T11:14:58 probably 2013-05-20T11:15:05 oh 2013-05-20T11:15:12 this would be easy to remove 2013-05-20T11:15:20 yeah 2013-05-20T11:15:27 *** weiY has quit IRC (Ping timeout: 260 seconds) 2013-05-20T11:16:02 DrJoel: is there something special about "echo Create 0x0a010001" in the script? 2013-05-20T11:16:15 i just took it out because I thought it is just echoing to the console 2013-05-20T11:17:48 hmmm, i think i see the problem. the semaphore is created with count=0 2013-05-20T11:18:01 There shouldn't be anything special about that echo. Just help in debugging :) 2013-05-20T11:18:34 It helped me to put comments above commands. The arguments try to match the directives 2013-05-20T11:18:49 need to create -V 1 2013-05-20T11:19:16 there we go. 2013-05-20T11:20:07 Is the simulator sufficient to trip the priority inversion case? 2013-05-20T11:20:08 oh well. i already implemented the sptest that has the behavior I wanted to study ;) 2013-05-20T11:20:15 checking 2013-05-20T11:20:54 I agree it could be used to exericse the core 2013-05-20T11:22:37 um, i'm not sure how to do ENABLE_STRICT_MUTEX=1 with schedsim :) 2013-05-20T11:22:58 Should pass through the same way 2013-05-20T11:24:22 ok trying it out 2013-05-20T11:28:11 it would be helpful if the "task name" were printed 2013-05-20T11:28:26 when? 2013-05-20T11:28:34 all the time :) 2013-05-20T11:28:43 TA01: Obtain semaphore... 2013-05-20T11:28:50 TA01: Releasing semaphore... 2013-05-20T11:28:57 TA02: ClockTick (1) ... 2013-05-20T11:29:08 the user-friendly task name of the executing task 2013-05-20T11:29:17 would help with understanding the flow of execution 2013-05-20T11:32:37 Semaphore create(-p) returned item has not been initialized 2013-05-20T11:38:14 ok i had to make it a binary semaphore too 2013-05-20T11:38:16 not sure why that is 2013-05-20T11:38:23 seems to work eventually 2013-05-20T11:42:06 The primary place to print the name would be in the dispatch wrapper 2013-05-20T11:42:45 maybe i'll get back to it later. moving on to other things now 2013-05-20T11:43:12 We just need requirements and suggestions. So far, it has had really light use 2013-05-20T11:43:17 i know 2013-05-20T11:43:34 you might want to create its own bugzilla area 2013-05-20T11:43:43 You are right though that it could be really handy for debugging/exercising the score 2013-05-20T11:43:46 PRs for it do not belong in RTEMS 2013-05-20T11:43:58 and i think it could be renamed accordingly 2013-05-20T11:47:43 I created a "product" called RTEMS Scheduler Simulator" with a single component. That's enough for now 2013-05-20T11:47:52 *** MegaAlex|away is now known as MegaAlex 2013-05-20T11:48:08 later can we rename it easily and preserve the PRs? 2013-05-20T11:54:50 I think so. Should be a way. It looked possible in Bugzilla 2013-05-20T11:54:54 ok 2013-05-20T11:55:18 check pm ;) 2013-05-20T11:55:43 fix your notifications! :) 2013-05-20T12:02:17 *** antgreen has quit IRC (Quit: Leaving) 2013-05-20T12:05:34 ok.. I turned on some more noises. Hopefully I hit all the important notifications 2013-05-20T12:37:53 * gedare distracts DrJoel with noises 2013-05-20T12:38:00 hi 2013-05-20T12:53:24 http://pastebin.com/xiEWTVRT 2013-05-20T12:53:45 undefined reference to `__libc_init_array' while building nds bsp 2013-05-20T12:53:52 is this a known error? 2013-05-20T13:06:40 i haven't seen it, but i haven't built that bsp either. 2013-05-20T13:07:28 That BSP has built in the not so distant past. Chris would have complained with RSB failing :( 2013-05-20T13:07:38 Where is the symbol referecned? 2013-05-20T13:07:57 sounds like a tools issue with newlib. 2013-05-20T13:10:14 do other arm bsps build? 2013-05-20T13:10:18 arm-rtems4.11-gcc --pipe -o coproc.elf -mcpu=arm7tdmi -mfpu=vfp -mfloat-abi=soft -O2 -DARM7 -I../../../../../../../../rtems/c/src/lib/libbsp/arm/nds/libnds/include -I../../../../../../../../rtems/c/src/lib/libbsp/arm/nds/dswifi/include -DENABLE_WIFI -T ../../../../../../../../rtems/c/src/lib/libbsp/arm/nds/coproc/coproc.ld ../../../../../../../../rtems/c/src/lib/libbsp/arm/nds/coproc/coproc.S ../../../../../../../../rtems/c/src/lib/libbsp/arm/nds/copr 2013-05-20T13:10:18 oc/coproc.c libnds7.rel dswifi7.rel -lc 2013-05-20T13:10:21 yes they do 2013-05-20T13:10:38 well all of them have been building until nds did not 2013-05-20T13:12:19 invoking the make at ~/tools/b-system/arm-rtems4.11/c/nds will lead directly to this 2013-05-20T13:12:55 I can remove this bsp, but at this point it might be messy or wouldn't it? 2013-05-20T13:15:15 I removed the nds from the Makefile at ~/tools/b-system/arm-rtems4.11/c lets see if it builds now :) 2013-05-20T13:17:35 better to only build the one or two bsps you care about .. saves times and disk space. 2013-05-20T13:17:49 Please report that build error on rtems-devel. It should not be happening 2013-05-20T13:17:55 leaving for appointment 2013-05-20T13:17:56 cya 2013-05-20T13:17:59 *** DrJoel has quit IRC (Quit: A day without sunshine is like .... night) 2013-05-20T13:24:46 the next bsp was built 2013-05-20T13:24:52 which was raspberry 2013-05-20T13:24:54 but now 2013-05-20T13:24:56 make[2]: Entering directory `/home/rgaribay/tools/b-system/arm-rtems4.11/c/realview_pbx_a9_qemu' 2013-05-20T13:24:56 make[2]: *** No rule to make target `all'. Stop. 2013-05-20T13:25:31 and ~/tools/b-system/arm-rtems4.11/c/realview_pbx_a9_qemu$ ls 2013-05-20T13:25:31 config.log make 2013-05-20T13:29:57 *** monstr_ has quit IRC (Ping timeout: 248 seconds) 2013-05-20T13:31:33 *** MegaAlex has quit IRC (Ping timeout: 248 seconds) 2013-05-20T13:32:12 hmm. 2013-05-20T13:32:39 you might just have better luck if you start from configure and pass only the BSPs you want built to --enable-rtemsbsp, like... 2013-05-20T13:35:03 ../rtems/configure --target=arm-rtems4.11 --enable-rtemsbsp="raspberrypi realview-pbx-a9" 2013-05-20T13:35:42 I'll do that 2013-05-20T13:36:03 *** MegaAlex has joined #rtems 2013-05-20T13:36:18 should I start from scratch (i.e. rm -rf build dir?) 2013-05-20T13:38:16 that might be the best 2013-05-20T13:38:23 cleaner that way 2013-05-20T13:39:31 the outcome of ../rtems/configure --target=arm-rtems4.11 --enable-rtemsbsp="raspberrypi realview-pbx-a9" 2013-05-20T13:39:36 states the following 2013-05-20T13:39:46 target architecture: arm. 2013-05-20T13:39:46 available BSPs: raspberrypi. 2013-05-20T13:39:46 'make all' will build the following BSPs: raspberrypi. 2013-05-20T13:39:46 other BSPs can be built with 'make RTEMS_BSP="bsp1 bsp2 ..."' 2013-05-20T13:40:00 hmm 2013-05-20T13:40:11 guess the realview bsp has some other name 2013-05-20T13:40:13 (already performed rm of build dir) 2013-05-20T13:40:22 realview_pbx_a9_qemu 2013-05-20T13:40:35 make RTEMS_BSP="raspberrypi realview_pbx_a9_qemu" 2013-05-20T13:42:25 where are these things documented? 2013-05-20T13:42:41 in ~/tools/rtems/c/src/lib/libbsp/arm the directory is namer realview-pbx-a9 2013-05-20T13:42:49 *named 2013-05-20T13:47:26 make[2]: Entering directory `/home/rgaribay/tools/b-system/arm-rtems4.11/c/realview_pbx_a9_qemu' 2013-05-20T13:47:26 make[2]: *** No rule to make target `all'. Stop. 2013-05-20T13:47:32 still get the same issue 2013-05-20T13:53:31 *** S_Somani has joined #rtems 2013-05-20T14:00:24 hmm. 2013-05-20T14:00:56 the documentation should be in the online manuals somewhere: http://rtems.org/onlinedocs/doc-current/share/rtems/html/ 2013-05-20T14:01:13 i wonder if there was an error in the configure step? 2013-05-20T14:01:24 try adding "--enable-maintainer-mode" to your configure command line 2013-05-20T14:01:38 that option can help to cover up some problems. 2013-05-20T14:11:55 *** S_Somani has left #rtems 2013-05-20T14:23:40 still the same 2013-05-20T14:23:59 a problem like this could only be from bootstrap or configure 2013-05-20T14:24:01 right? 2013-05-20T14:24:25 no problem at all with automake and autoconf? 2013-05-20T14:24:59