2010-08-30T02:12:11 *** projectgus has quit IRC 2010-08-30T02:59:42 *** exception13 has joined #rtems 2010-08-30T03:21:35 *** madrazr has joined #rtems 2010-08-30T04:24:54 *** exception13 has left #rtems 2010-08-30T04:25:19 *** exception13 has joined #rtems 2010-08-30T04:25:43 *** exception13 has left #rtems 2010-08-30T04:28:14 *** exception13 has joined #rtems 2010-08-30T07:10:21 hello 2010-08-30T07:11:07 i need help with debuggin rtems booting with u-boot 2010-08-30T07:13:24 < rokka> i think something goes wrong after line 158 in c/src/lib/libbsp/arm/shared/start/start.S 2010-08-30T07:13:27 < rokka> it jumps to 0x000000d2 and then to 0x0000000c and then to twiddle 2010-08-30T07:13:30 k 2010-08-30T07:13:47 i don't see any boot messages from rtems 2010-08-30T07:14:28 last row is: "Transferring control to RTEMS (at address a0000040) .." 2010-08-30T07:14:37 then maybe one corrupted character .. 2010-08-30T07:46:55 *** madrazr has quit IRC 2010-08-30T09:22:20 *** madrazr has joined #rtems 2010-08-30T10:42:51 *** tuxmaniac has joined #rtems 2010-08-30T10:49:44 *** madrazr has quit IRC 2010-08-30T11:45:12 *** madrazr has joined #rtems 2010-08-30T11:52:14 *** kuzew has quit IRC 2010-08-30T11:53:28 *** kuzew has joined #rtems 2010-08-30T12:34:55 *** tuxmaniac has quit IRC 2010-08-30T12:40:04 *** madrazr has quit IRC 2010-08-30T15:35:56 *** exception13 has left #rtems 2010-08-30T18:30:38 *** projectgus has joined #rtems 2010-08-30T19:13:47 *** DrJoel has joined #rtems 2010-08-30T19:13:47 *** ChanServ sets mode: +o DrJoel 2010-08-30T19:43:31 *** DrJoel has quit IRC 2010-08-30T20:25:29 *** DrJoel has joined #rtems 2010-08-30T20:25:29 *** ChanServ sets mode: +o DrJoel 2010-08-30T20:25:36 *** DrJoel has quit IRC 2010-08-31T01:47:32 *** projectgus has quit IRC 2010-08-31T01:49:27 *** lekernel has quit IRC 2010-08-31T01:51:01 *** lekernel has joined #rtems 2010-08-31T02:08:35 *** _Lucretia_ has quit IRC 2010-08-31T05:21:34 *** madrazr has joined #rtems 2010-08-31T05:59:44 *** exception13 has joined #rtems 2010-08-31T07:01:37 *** DrJoel has joined #rtems 2010-08-31T07:01:37 *** ChanServ sets mode: +o DrJoel 2010-08-31T07:03:13 *** DrJoel has quit IRC 2010-08-31T07:07:50 *** _Lucretia_ has joined #rtems 2010-08-31T07:21:38 *** madrazr has quit IRC 2010-08-31T07:23:28 *** madrazr has joined #rtems 2010-08-31T07:54:21 *** lekernel has quit IRC 2010-08-31T08:10:12 *** madrazr1 has joined #rtems 2010-08-31T08:11:11 *** madrazr has quit IRC 2010-08-31T08:11:27 *** DrJoel has joined #rtems 2010-08-31T08:11:28 *** ChanServ sets mode: +o DrJoel 2010-08-31T08:33:41 *** madrazr1 has quit IRC 2010-08-31T08:45:15 *** lekernel has joined #rtems 2010-08-31T09:36:14 *** gedare has joined #rtems 2010-08-31T09:40:10 good morning 2010-08-31T09:43:04 DrJoel: can you provide me with a trace of the execution for the rtems_semaphore_obtain on psim that shows the +50 overhead 2010-08-31T09:43:17 (I don't have a psim setup :) 2010-08-31T09:43:28 I don't know how to trace on psim. :( Let me look 2010-08-31T09:43:42 *** exception13 has left #rtems 2010-08-31T09:44:02 *** exception13 has joined #rtems 2010-08-31T09:44:19 OK. I'm going to spend a couple minutes on this today, but I have other obligations to attend. 2010-08-31T09:44:34 no one has submitted his code to google yet 2010-08-31T09:44:35 life has caught up :) 2010-08-31T09:44:39 I understand.. yesterday was my wife's birthday so you didn't get much time 2010-08-31T09:44:43 i think i will this afternoon 2010-08-31T09:44:45 Fallenou: when's the deadline? 2010-08-31T09:44:57 Can one of you post a message to the rtems-gsoc list with details 2010-08-31T09:45:07 humm a soft dead line in a dozen of days 2010-08-31T09:45:21 i can fw carol smith email sure 2010-08-31T09:45:36 as soon as im back home from the shopping 2010-08-31T09:46:02 thanks .. are you buying us all something nice? LOL 2010-08-31T09:46:22 they are annoying, they do not want any code written after the pencil down date 2010-08-31T09:46:34 lol :) i bought perfume and a comic book 2010-08-31T09:46:44 want some perfume ?:p 2010-08-31T09:46:47 no perfume for me but the comic book sounds nice 2010-08-31T09:46:54 hehe 2010-08-31T09:49:08 thats a comic book based on a web comic strip 2010-08-31T09:49:21 iirc 2010-08-31T09:49:34 which one? 2010-08-31T09:49:53 * DrJoel used to be a big comic book collector.. have some nice Golden Age and early Silver Age 2010-08-31T09:49:55 I made my code submission right before the pencils down date :p 2010-08-31T09:50:04 Just have t ofind it.... 2010-08-31T09:50:38 http://cache.coverbrowser.com/image/action-comics/47-1.jpg is one of my prizes.. first Lex Luther cover 2010-08-31T09:50:52 ok no i checked this one is not based on a web comic strip 2010-08-31T09:50:57 its "lou' 2010-08-31T09:51:44 its pretty new french comic 2010-08-31T09:51:53 adapted on tv afterward 2010-08-31T09:52:15 http://www.coverbrowser.com/image/brave-and-the-bold/28-1.jpg is another (first Justice League appearance) .. also have B&B 29-30 and JLA 1-200 2010-08-31T09:52:28 thanks for the mail gedare 2010-08-31T09:52:51 i guess i will try to cheat 2010-08-31T09:53:01 and submit an up to date version 2010-08-31T09:53:06 Why 2010-08-31T09:53:12 mh? 2010-08-31T09:53:35 They don't really use the code. they just want samples of work done. 2010-08-31T09:53:41 because i dunno if the last commit of the pencil down date does not break anything 2010-08-31T09:53:46 I'm not even convinced my patch would commit ;) 2010-08-31T09:54:13 i find it ridiculous to put on the web some code that does not work 2010-08-31T09:54:19 Fair enough 2010-08-31T09:54:32 with our name next to it 2010-08-31T09:54:34 :p 2010-08-31T09:55:00 gedare: rtems psim build does not have tracing enabled apparently. :( 2010-08-31T09:55:14 is it hard for you to build a psim target? 2010-08-31T09:55:35 bleh Ok. I will manually track that code to see if I can figure it out. I'm not familiar with the tools, so it would take me awhile to get everything going 2010-08-31T09:55:57 Maybe I can do it later on a Fedora client. I have to compile everything on my Ubuntu desktop :) 2010-08-31T09:56:36 :) psim isn't hard to use with the scripts 2010-08-31T09:56:51 what is psim ? 2010-08-31T09:56:54 By the way, depending on how function calls are handled, I would estimate an extra 5-6 instructions per 'scheduling call' 2010-08-31T09:57:00 on average. 2010-08-31T09:57:21 some targets (sparc) can hide this if the function call is "free" 2010-08-31T09:57:29 the testsuite to calculate function call overhead? 2010-08-31T09:57:33 but the average will still win out because of window spill/fill 2010-08-31T09:57:41 psim is a simulator for powerpc 2010-08-31T09:57:48 oh ok 2010-08-31T09:57:51 I think it runs with gdb or something 2010-08-31T09:58:01 better than qemu ? 2010-08-31T09:58:31 different than qemu. :) I don't know if qemu emulates the ppc very well. 2010-08-31T10:00:09 ok 2010-08-31T10:02:12 *** madrazr has joined #rtems 2010-08-31T10:03:40 We have a qemuppc BSP but I only use it for coverage. You will need the couverture tree. I suppose it would work equally well in this case 2010-08-31T10:11:28 *** madrazr1 has joined #rtems 2010-08-31T10:11:48 *** madrazr has quit IRC 2010-08-31T10:30:10 *** madrazr1 is now known as madrazr 2010-08-31T11:01:47 DrJoel: the improvements in time appear to be from not flashing the ISRs. The time savings is about 4-5 on sis, which might be from not context switching. 2010-08-31T11:02:31 or just not doing that trap. :P 2010-08-31T11:02:41 gedare: that may be OK or it might be increasing the interrupt latency.. the trap is probably a savings though. 2010-08-31T11:03:05 I am concerned that some common operations seem to have grown ... 50 instructions is a lot 2010-08-31T11:03:20 We'll have to go case-by-case on the flashing issue later. 2010-08-31T11:04:07 I agree. FWIW it looks like there are a lot of 25-50 instruction jumps on psim 2010-08-31T11:04:15 by jumps i mean increases between pre and post. 2010-08-31T11:04:30 I was trying to get that in a spreadsheet but haven't yet... sorry 2010-08-31T11:04:46 It's ok. I've spent my RTEMS quota for today :) 2010-08-31T11:04:50 I think a build of psim with --enable-sim-tracXXX should do it. 2010-08-31T11:05:02 me too.. I actually have to eat lunch and get ready for a 1pm meeting 2010-08-31T11:05:40 ok. thanks for the help.:) 2010-08-31T11:05:41 I may be able to get you a trace on qemuppc easier than psim if that would help 2010-08-31T11:05:55 You would need a before trace and num and both after 2010-08-31T11:06:18 oK. it also might be useful to see how good these averages are 2010-08-31T11:06:36 or if there is some weird outliers. 2010-08-31T11:07:26 but a trace would be helpful to see what code path is taken through the new scheduler/ready queue, and how many calls are made. 2010-08-31T11:10:52 I might be able to correlate the number of extra calls to the overhead. On powerpc I think the function call overhead will be more dependent on register pressure, whereas on sparc it is dependent on the call depth. 2010-08-31T11:12:04 Alright, time for me to re-focus. 2010-08-31T11:12:36 *** gedare has quit IRC 2010-08-31T11:59:00 *** DrJoel has quit IRC 2010-08-31T13:05:51 *** madrazr1 has joined #rtems 2010-08-31T13:06:10 *** madrazr has quit IRC 2010-08-31T13:06:17 *** madrazr1 is now known as madrazr 2010-08-31T13:06:23 *** madrazr has joined #rtems 2010-08-31T13:09:18 *** madrazr1 has joined #rtems 2010-08-31T13:09:31 *** madrazr has quit IRC 2010-08-31T13:09:34 *** madrazr2 has joined #rtems 2010-08-31T13:10:54 *** madrazr2 is now known as madrazr 2010-08-31T13:11:00 *** madrazr has quit IRC 2010-08-31T13:11:00 *** madrazr has joined #rtems 2010-08-31T13:13:59 *** madrazr1 has quit IRC 2010-08-31T13:32:32 *** madrazr1 has joined #rtems 2010-08-31T13:32:40 *** madrazr has quit IRC 2010-08-31T13:51:11 *** exception13 has left #rtems 2010-08-31T13:51:16 *** exception13 has joined #rtems 2010-08-31T14:33:39 *** krzabr has joined #rtems 2010-08-31T14:54:47 *** kuzew has quit IRC 2010-08-31T14:55:34 *** kuzew has joined #rtems 2010-08-31T15:52:00 *** krzabr has quit IRC 2010-08-31T16:47:55 * Fallenou submitted his code on the google page 2010-08-31T16:55:58 *** madrazr1 has quit IRC 2010-08-31T16:57:38 Fallenou, mh? where? 2010-08-31T16:57:57 http://code.google.com/p/google-summer-of-code-2010-rtems/source/list 2010-08-31T16:58:00 i see only r1 2010-08-31T16:58:25 in downloads 2010-08-31T16:58:56 ah ok 2010-08-31T16:59:13 i followed that : http://groups.google.com/group/google-summer-of-code-students-list/web/code-submission-guidelines-2010 2010-08-31T16:59:20 btw cvs head is still no working right? 2010-08-31T16:59:37 yes 2010-08-31T16:59:44 still not working 2010-08-31T17:00:16 ok just drop me a note, when its done, so i can try to fix that isr type issue 2010-08-31T17:00:34 ok ! 2010-08-31T17:00:35 *** madrazr has joined #rtems 2010-08-31T17:02:56 *** madrazr has quit IRC 2010-08-31T17:16:00 *** DrJoel has joined #rtems 2010-08-31T17:16:01 *** ChanServ sets mode: +o DrJoel 2010-08-31T17:16:20 mwalle: have you checked out lm32 lately? It doesn't appear to run on qemu (I think) 2010-08-31T17:18:48 *** exception13 has left #rtems 2010-08-31T17:20:42 the lm32-evr bsp? 2010-08-31T17:20:57 DrJoel, or milkymist? 2010-08-31T17:22:25 $ qemu-lm32_evr -i ticker.ralf 2010-08-31T17:22:25 qemu-system-lm32 is /home2/simulators/bin/qemu-system-lm32 2010-08-31T17:22:25 qemu: fatal: Trying to execute code outside RAM or ROM at 0xbfcf5c68 2010-08-31T17:22:25 IN: PC=bfcf5c68 2010-08-31T17:22:25 ie=00000000 (IE=0 EIE=0 BIE=0) im=00000000 ip=00000000 2010-08-31T17:22:26 eba=04000000 deba=00000000 2010-08-31T17:22:30 r00=00000000 r01=00000000 r03=00000000 2010-08-31T17:22:32 r04=00000000 r05=00000000 r06=00000000 r07=00000000 2010-08-31T17:22:34 r08=00000000 r09=00000000 r10=00000000 r11=00000000 2010-08-31T17:22:36 r12=00000000 r13=00000000 r14=00000000 r15=00000000 2010-08-31T17:22:38 r16=00000000 r17=00000000 r18=00000000 r19=00000000 2010-08-31T17:22:39 ah :) 2010-08-31T17:22:40 r20=00000000 r21=00000000 r22=00000000 r23=00000000 2010-08-31T17:22:42 r24=00000000 r25=00000000 r26=00000000 r27=00000000 2010-08-31T17:22:44 r28=00000000 r29=00000000 r30=00000000 r31=00000000 2010-08-31T17:22:46 Aborted 2010-08-31T17:22:46 try hello.exe 2010-08-31T17:23:11 if its working, its likely that the milkymist port suffers from the same issue 2010-08-31T17:23:30 I haven't changed qemu-lm32 so that means something else broke.. and Fallenou has had some breakage and I wondered if it was the same root cause and you could help pin it down 2010-08-31T17:25:22 oh ! 2010-08-31T17:25:34 that would be interesting 2010-08-31T17:25:52 DrJoel, it has likely something todo with the interrupt 2010-08-31T17:26:02 and todo with recent changes 2010-08-31T17:26:10 since it still work with my git repo 2010-08-31T17:26:18 and does not while merged with cvs 2010-08-31T17:26:34 I have no idea what would have changed to break it.. do you have the per-cpu changes? 2010-08-31T17:26:48 i debugged it with qemu and the program branches to 0 in a ret opcode 2010-08-31T17:27:28 which is a b to $ra and $ra is popped from the stack 2010-08-31T17:27:39 so i assume the stack is broken :) 2010-08-31T17:28:11 we should look at recent commits about lm32 stuff 2010-08-31T17:28:15 i'm no pro with cvs 2010-08-31T17:28:39 Fallenou, use the git repo :) 2010-08-31T17:28:57 it was having trouble and may not be up to date .. :( 2010-08-31T17:29:00 but for now im going to bed, tomorrow is class 2010-08-31T17:29:08 let me run the import by hand 2010-08-31T17:29:35 http://git.serverraum.org/?p=mw/rtems.git;a=summary :) 2010-08-31T17:30:09 thx 2010-08-31T17:30:28 will be a change to cpukit/score/cpu/lm32/cpu_asm.S related to offsets in per-cpu structure (probably) 2010-08-31T17:31:34 Initialized empty Git repository in /home/fallen/rtems-git/rtems/.git/ 2010-08-31T17:31:35 error: The requested URL returned error: 403 while accessing http://git.serverraum.org/?p=mw/rtems.git/info/refs 2010-08-31T17:33:03 try http://git.serverraum.org/git/mw/rtems.git 2010-08-31T17:35:28 gn8 2010-08-31T17:36:42 ok gn8 2010-08-31T17:40:18 night 2010-08-31T17:43:15 *** DrJoel has quit IRC 2010-08-31T17:56:34 there was a change in irq.c 2010-08-31T17:56:43 cpu/lm32/irq.c 2010-08-31T17:56:54 int 2010-07-30 2010-08-31T17:57:00 20h52 2010-08-31T17:57:09 maybe it's the origin of the bug 2010-08-31T17:57:43 *** kiwichris has joined #rtems 2010-08-31T18:01:02 *** projectgus has joined #rtems 2010-08-31T18:04:14 gn8 2010-08-31T22:30:07 *** kristian1aul has joined #rtems 2010-09-01T02:03:27 *** projectgus has quit IRC 2010-09-01T03:13:18 *** madrazr has joined #rtems 2010-09-01T06:15:48 *** exception13 has joined #rtems 2010-09-01T06:50:42 *** DrJoel has joined #rtems 2010-09-01T06:50:42 *** ChanServ sets mode: +o DrJoel 2010-09-01T06:52:22 * DrJoel waves hi 2010-09-01T07:15:07 hi DrJoel 2010-09-01T07:15:14 hello 2010-09-01T07:16:05 i guess what i pointed out in my email was just a define renaming, but its still suspicious 2010-09-01T07:16:31 the name changed.. not suspicious if it was about context switch necessary .. as long as the offset is OK 2010-09-01T07:17:16 I guess we need to trace through _ISR_Handler 2010-09-01T07:18:32 gotta run.. need to take son to school 2010-09-01T07:18:35 *** DrJoel has quit IRC 2010-09-01T08:28:17 *** exception13 has left #rtems 2010-09-01T10:27:25 *** gedare has joined #rtems 2010-09-01T11:30:18 *** gedare has quit IRC 2010-09-01T12:05:29 *** _Lucretia__ has joined #rtems 2010-09-01T12:06:24 *** _Lucretia_ has quit IRC 2010-09-01T12:21:00 *** gedare has joined #rtems 2010-09-01T12:39:27 *** DrJoel has joined #rtems 2010-09-01T12:39:27 *** ChanServ sets mode: +o DrJoel 2010-09-01T12:42:11 *** exception13 has joined #rtems 2010-09-01T13:34:50 *** madrazr has quit IRC 2010-09-01T15:40:02 *** DrJoel has quit IRC 2010-09-01T16:11:37 hi 2010-09-01T16:12:25 Fallenou, you could try git bisect 2010-09-01T16:12:54 (testing with the lm32-evr ticker.exe" 2010-09-01T16:20:55 oh yes, since i have a git repo now i can bisect :) 2010-09-01T16:21:09 i've never used it, will be fun ! 2010-09-01T16:55:35 *** exception13 has left #rtems 2010-09-01T17:30:32 *** gedare has quit IRC 2010-09-01T18:03:34 *** projectgus has joined #rtems 2010-09-01T18:55:31 *** dr__house has joined #rtems 2010-09-01T18:55:31 *** dr__house has joined #rtems 2010-09-01T20:39:46 fallen@tablet:~/rtems-git/rtems$ git bisect bad 2010-09-01T20:39:46 Bisecting: 12 revisions left to test after this (roughly 4 steps) 2010-09-01T20:39:46 [f18fdb2fdc515d3761c7e595af4f215bf7a9783d] 2010-08-16 Ralf Corsépius 2010-09-01T20:39:51 mwalle: almost done with bisecting ! 2010-09-01T20:40:04 i've been through 411 commits ^^ 2010-09-01T20:40:17 will finish the 4 more steps tomorrow 2010-09-01T20:40:40 Fallenou, woo that is a large number. What is the problem you are looking at ? 2010-09-01T20:40:57 kiwichris: lm32 does not work anymore (lately) 2010-09-01T20:41:04 lm32-evr and milkymist, both 2010-09-01T20:41:08 hello sample works 2010-09-01T20:41:11 ticker does not 2010-09-01T20:41:23 How does it die ? 2010-09-01T20:41:23 we are guessing some interrupt problem, maybe interrupt stack got fucked up 2010-09-01T20:41:26 but we dunno why 2010-09-01T20:41:48 On a simulator or real hardware ? 2010-09-01T20:42:29 on qemu 2010-09-01T20:42:36 i guess it's the same on the hardware 2010-09-01T20:42:43 i cannot test 2010-09-01T20:42:58 Any instruction trace log available ? 2010-09-01T20:43:18 i closed the terminal but i pasted to someone the beginning of the message 2010-09-01T20:43:23 let me paste it in pastebin 2010-09-01T20:43:32 Ok 2010-09-01T20:43:48 http://pastebin.com/fdGwKeTz 2010-09-01T20:44:14 sometimes the fatal appears right after the CLOCK TICK TEST message 2010-09-01T20:44:30 Always an execute at 0 ? 2010-09-01T20:44:35 yes 2010-09-01T20:44:50 Have you tried gdb ? 2010-09-01T20:44:53 nop 2010-09-01T20:44:59 mwalle did 2010-09-01T20:45:20 qemu has a gdbserver built in and it is very easy to hook up and use, well it is for the i386 and m68k 2010-09-01T20:45:36 i know for the gdbserver 2010-09-01T20:45:38 On the i386 you can then set a watch point in 0 2010-09-01T20:45:46 but i don't know if i will be able to interprete what i would see in gdb :) 2010-09-01T20:45:49 It may also work on this hardware 2010-09-01T20:45:58 Well lets see. 2010-09-01T20:46:18 This will give you the location of NULL pointer access. 2010-09-01T20:46:26 maybe finding the faulty commit would give an obvious reason 2010-09-01T20:46:50 oh btw 2010-09-01T20:47:22 do you know another process than this : git bisect good/bad , then ./bootstrap , then configure and make all again and repeat all the process 2010-09-01T20:47:28 i mean, the bootstrap part is looooooooooooong 2010-09-01T20:47:36 No. 2010-09-01T20:47:44 but i tried sometimes not to do it, but each time there was new files and stuff 2010-09-01T20:47:51 and makefiles were not compatible 2010-09-01T20:47:54 it's a real pain in the ass 2010-09-01T20:48:12 at least when developpign a bsp i can just run the bootstrap in my directory 2010-09-01T20:48:13 Yeah it will move around. --enable-maintainer-mode (or what ever) can help 2010-09-01T20:48:17 here i gotta run it everywhere :( 2010-09-01T20:48:22 Yeap. 2010-09-01T20:48:29 what does the maintainer mode ? 2010-09-01T20:48:34 +do 2010-09-01T20:48:36 I would find the code that is performing the NULL access. 2010-09-01T20:49:50 i'm going to sleep it's almost 4am here :' 2010-09-01T20:49:53 gn8 ! 2010-09-01T20:49:57 cya 2010-09-01T21:04:06 *** Alison_Chaiken has joined #rtems 2010-09-01T21:04:47 *** Alison_Chaiken has left #rtems 2010-09-01T21:13:18 *** dr__house has quit IRC 2010-09-02T01:59:26 *** projectgus has quit IRC 2010-09-02T04:13:36 *** madrazr has joined #rtems 2010-09-02T04:47:30 *** madrazr has quit IRC 2010-09-02T05:47:47 *** madrazr has joined #rtems 2010-09-02T05:56:09 kiwichris: entire copy paste is here : http://pastebin.com/GU0ZNWk0 2010-09-02T05:56:11 hi ! 2010-09-02T05:56:24 hi 2010-09-02T05:56:31 Did you get some sleep ? 2010-09-02T05:56:40 yes sure :) 2010-09-02T05:56:45 i'm up and running ;) 2010-09-02T05:56:52 Great 2010-09-02T05:57:04 5 left commits to bisect 2010-09-02T05:57:38 It looks like an indirect call that is set to 0 2010-09-02T05:58:36 Do you know how to run gdb with qemu ? 2010-09-02T05:58:55 humm i've never done that 2010-09-02T05:59:14 maybe i can give you informations if you tell me how to 2010-09-02T05:59:18 If you want to know just say and I will walk you though it. It should only take a few minutes 2010-09-02T05:59:32 It is easy to use. 2010-09-02T05:59:37 i'm ok to learn : 2010-09-02T05:59:37 :) 2010-09-02T05:59:54 Have a 2nd shell prompt in the same directory as the executable. 2010-09-02T06:00:05 Start qemu with "-s -S" I think. 2010-09-02T06:00:19 The -s is start gdbserver and -S is stop at reset. 2010-09-02T06:00:40 ok 2010-09-02T06:00:44 it's done 2010-09-02T06:01:22 Then start "lm32-rtems4.11-gdb ticker.exe" 2010-09-02T06:01:39 ok 2010-09-02T06:02:03 At the gdb prompt enter "target remote :1234" 2010-09-02T06:02:22 Remote debugging using :1234 2010-09-02T06:02:22 0x04000000 in start () 2010-09-02T06:02:28 Yeap. 2010-09-02T06:02:37 Now "watch 0" 2010-09-02T06:03:17 watchpoint in place general 2010-09-02T06:03:35 Then 'c' 2010-09-02T06:03:41 and qemu should run 2010-09-02T06:03:45 (gdb) c 2010-09-02T06:03:45 Continuing. 2010-09-02T06:03:45 Remote connection closed 2010-09-02T06:03:48 not good ? 2010-09-02T06:03:57 Oh. Did qemu quit ? 2010-09-02T06:04:06 yes 2010-09-02T06:04:25 Ok. Lets try again but this time do "b *0" rather than watch 2010-09-02T06:05:09 same thing 2010-09-02T06:05:17 Hmm. 2010-09-02T06:05:23 (gdb) b *0 2010-09-02T06:05:23 Breakpoint 1 at 0x0 2010-09-02T06:05:23 (gdb) c 2010-09-02T06:05:23 Continuing. 2010-09-02T06:05:23 Remote connection closed 2010-09-02T06:05:42 So it gets to the ticker clock test then qemu exits ? 2010-09-02T06:06:05 i would say there is no change in the qemu output 2010-09-02T06:06:08 same error 2010-09-02T06:06:23 as soon as i type 'c' in gdb 2010-09-02T06:06:45 Does the ticker output appear after the 'c' ? 2010-09-02T06:07:09 yes, beginning of output 2010-09-02T06:07:13 and them the fatal error 2010-09-02T06:07:17 then* 2010-09-02T06:07:18 Ok. 2010-09-02T06:07:21 Hang on 2010-09-02T06:10:23 ok :) 2010-09-02T06:14:42 *** DrJoel has joined #rtems 2010-09-02T06:14:42 *** ChanServ sets mode: +o DrJoel 2010-09-02T06:14:53 * DrJoel waves hi 2010-09-02T06:14:56 Hi Joel 2010-09-02T06:15:31 mwalle: I have a guess on the lm32 failures .. CPU_STACK_ALIGNMENT is 0 and the code to calculate _CPU_Interrupt_stack_high changed. I bet the interrupt stack is misaligned 2010-09-02T06:15:33 hi Joel :) 2010-09-02T06:15:35 * Fallenou going to eat 2010-09-02T06:15:44 hey kiwichris .. how was tennis? 2010-09-02T06:16:12 Not to bad. Had a 3rd tie break win 2010-09-02T06:16:28 :) 2010-09-02T06:16:57 did you get a chance to look at https://spreadsheets.google.com/pub?key=0AngKwYy0qUGZdGVVTE5oLVljYVdUVjlKT19JanQxb2c&hl=en&output=html 2010-09-02T06:17:53 * DrJoel waves to Fallenou 2010-09-02T06:17:55 Yes but a little too complex to follow with a simple look 2010-09-02T06:20:48 that describes posix in a nutshell.. :-D 2010-09-02T06:21:08 one tough nut :) 2010-09-02T06:21:17 The simple thing to look at is the "Missing API" sheet which shows based upon each section (A-C, D-F, etc) what is not in rtems or newlib 2010-09-02T06:21:34 I haven't gone through the networking yet .. so not 100% and haven't doublechecked any of my input 2010-09-02T06:26:40 * DrJoel waves bye.. need to get ready to head to office.. 2010-09-02T06:26:43 be back on later 2010-09-02T06:26:47 *** DrJoel has quit IRC 2010-09-02T06:28:04 *** exception13 has joined #rtems 2010-09-02T06:38:00 Fallenou, I have been looking but I cannot see a way to get qemu to not exit. 2010-09-02T06:38:12 Could you try a breakpoint on Init ? 2010-09-02T06:52:47 how do i do that ? 2010-09-02T06:53:07 * Fallenou is back 2010-09-02T06:56:02 *** madrazr has quit IRC 2010-09-02T07:06:32 (gdb) n 2010-09-02T07:06:32 56 status = rtems_task_create( 2010-09-02T07:06:32 (gdb) n 2010-09-02T07:06:32 Remote connection closed 2010-09-02T07:06:40 it stops there 2010-09-02T07:47:59 *** madrazr has joined #rtems 2010-09-02T07:58:20 *** tuxmaniac has joined #rtems 2010-09-02T08:02:15 *** tuxmaniac has quit IRC 2010-09-02T08:29:08 one more commit to try 2010-09-02T08:29:17 i'm pretty sure this last one will be the bad guy 2010-09-02T08:29:51 _CPU_Interrupt_stack_high = (void *) 2010-09-02T08:29:51 - ((uintptr_t) _CPU_Interrupt_stack_high & ~CPU_STACK_ALIGNMENT); 2010-09-02T08:29:52 + ((uintptr_t) _CPU_Interrupt_stack_high & ~(CPU_STACK_ALIGNMENT - 1)); 2010-09-02T08:30:26 Seems to match what Joel said 2010-09-02T08:42:41 *** DrJoel has joined #rtems 2010-09-02T08:42:41 *** ChanServ sets mode: +o DrJoel 2010-09-02T08:47:36 *** kiwichris has quit IRC 2010-09-02T09:03:55 git bisect confirms DrJoel dd7fe9b8626ea8e485c4e5623258482efb30dbe3 is the first bad commit 2010-09-02T09:05:12 is that the one which touched interrupt stack high? 2010-09-02T09:05:26 I suspect that having CPU_STACK_ALIGN == 0 is a problem 2010-09-02T09:05:35 yes 2010-09-02T09:05:42 15:12 < Fallenou> _CPU_Interrupt_stack_high = (void *) 2010-09-02T09:05:42 15:12 < Fallenou> - ((uintptr_t) _CPU_Interrupt_stack_high & ~CPU_STACK_ALIGNMENT); 2010-09-02T09:05:45 15:12 < Fallenou> + ((uintptr_t) _CPU_Interrupt_stack_high & ~(CPU_STACK_ALIGNMENT - 1)); 2010-09-02T09:05:48 this one 2010-09-02T09:06:19 change lm32.../cpu.h so CPU_STACK_ALIGNMENT is 4 or 8 or whatever it really needs to be. :-D 2010-09-02T09:06:38 that is anding it with ~(0 -1). 2010-09-02T09:07:15 hum seems reasonnable for me to align stack to 4 bytes boundary 2010-09-02T09:07:21 if it can correct this mess 2010-09-02T09:08:29 give that a try.. (if not) see if tracing to that point does cause a fault. 2010-09-02T09:16:16 * Fallenou tries with 4 2010-09-02T09:40:48 DrJoel: that's better :) 2010-09-02T09:40:54 ticker works with this little modification 2010-09-02T09:41:18 for lm32-evr 2010-09-02T09:42:13 *** exception13 has left #rtems 2010-09-02T09:43:41 yeah... submit a pr and patch.. I will merge it :-D 2010-09-02T09:45:02 ok great ! 2010-09-02T09:46:06 looks like me moving the interrupt stack management management into a more portable place highlighted that the lm32 port was lying about its alignment requirements. :) 2010-09-02T09:46:57 what do you mean ? 2010-09-02T10:06:07 It had CPU_STACK_ALIGNMENT = 0 before I made that change. But more than likely the constant wasn't used before I made this change 2010-09-02T10:07:34 hum ok 2010-09-02T10:16:10 DrJoel: can i send you the patch as a git diff ? 2010-09-02T10:16:25 or a git commit patch 2010-09-02T10:16:43 either way.. doesn't matter 2010-09-02T10:17:04 ok good :) 2010-09-02T10:17:13 i'm way more confident with git 2010-09-02T10:27:17 *** exception13 has joined #rtems 2010-09-02T10:55:06 DrJoel: would be cool to add "lm32" in hardware section for a PR :) 2010-09-02T10:56:49 hmmm... hold on 2010-09-02T10:58:46 done.. also added m32r and sparc64 :-D 2010-09-02T10:59:27 ok cool :) 2010-09-02T11:04:01 https://www.rtems.org/bugzilla/show_bug.cgi?id=1697 DrJoel 2010-09-02T11:10:01 ok.. will get it 2010-09-02T11:15:43 *** madrazr has quit IRC 2010-09-02T11:56:54 *** madrazr has joined #rtems 2010-09-02T13:35:11 *** madrazr has quit IRC 2010-09-02T13:39:10 *** madrazr has joined #rtems 2010-09-02T14:43:32 *** n3oo3n_ has joined #rtems 2010-09-02T14:47:34 *** n3oo3n has quit IRC 2010-09-02T15:26:08 *** madrazr has quit IRC 2010-09-02T15:41:45 *** kiwichris has joined #rtems 2010-09-02T15:42:06 hi 2010-09-02T15:43:01 DrJoel, are you about ? 2010-09-02T15:44:03 yes.. what's up 2010-09-02T15:44:41 Are you able to look at pr 1695 ? 2010-09-02T15:45:05 The issue is the function API is defined by the handler API. 2010-09-02T15:45:20 This in turn is defined by the SUS read call. 2010-09-02T15:45:56 pulling it up.. hold on.. 2010-09-02T15:48:38 does bytes ransferred need to be ssize_t ? 2010-09-02T15:50:12 I think that's the solution.. hold on.. reading http://www.opengroup.org/onlinepubs/009695399/basedefs/sys/types.h.html 2010-09-02T15:51:33 See if my comment helps at all 2010-09-02T15:51:49 I wished someone just said change size_t to ssize_t :) 2010-09-02T15:52:43 have tp go cya 2010-09-02T15:52:58 bye 2010-09-02T15:59:52 DrJoel, hi 2010-09-02T16:00:20 when can CPU_STACK_ALIGNMENT be 0 ? 2010-09-02T16:01:51 when there is no masking with (CPU_STACK_ALIGNMENT - 1) ? :p 2010-09-02T16:01:55 when the cpu can do byte aligned stacks 2010-09-02T16:01:56 yep :) 2010-09-02T16:02:29 well, than this http://git.serverraum.org/?p=mw/rtems.git;a=blobdiff;f=cpukit/score/src/isr.c;h=59efc6e6a6a3348221c3cf9e95e437e4b08fc3c3;hp=f71a81aaa782640eba75f91708ac33f7f6f823f0;hb=dd7fe9b8626ea8e485c4e5623258482efb30dbe3;hpb=7db75c5ff52bef147ca6f9fd1c390b01d8f5d6aa breaks things 2010-09-02T16:03:16 unless you should set it to 1 if the cpu can have byte aligned stacks 2010-09-02T16:04:56 ~(0 - 1) == 0 , right? 2010-09-02T16:05:09 right 2010-09-02T16:05:10 yes thats this commit 2010-09-02T16:05:14 that breaks things 2010-09-02T16:06:15 for every arch which specifies CPU_STACK_ALIGNMENT := 0 2010-09-02T16:06:41 is that any? 2010-09-02T16:07:27 so lm32 was the only arch capable of using a non aligned stack ? 2010-09-02T16:07:38 well at least lm32 :) 2010-09-02T16:07:50 hmmm... there are a number of 0's 2010-09-02T16:07:52 ./avr/rtems/score/cpu.h:#define CPU_STACK_ALIGNMENT 0 2010-09-02T16:07:53 ./bfin/rtems/score/cpu.h:#define CPU_STACK_ALIGNMENT 0 2010-09-02T16:07:53 ./lm32/rtems/score/cpu.h:#define CPU_STACK_ALIGNMENT 0 2010-09-02T16:07:53 ./no_cpu/rtems/score/cpu.h:#define CPU_STACK_ALIGNMENT 0 2010-09-02T16:07:53 ./nios2/rtems/score/cpu.h:#define CPU_STACK_ALIGNMENT 0 2010-09-02T16:07:53 ./m32r/rtems/score/cpu.h:#define CPU_STACK_ALIGNMENT 0 2010-09-02T16:07:55 ./m32c/rtems/score/cpu.h:#define CPU_STACK_ALIGNMENT 0 2010-09-02T16:07:57 ./m68k/rtems/score/cpu.h:#define CPU_STACK_ALIGNMENT 0 2010-09-02T16:08:05 avr, m32r and m32c don't have interrupt support yet 2010-09-02T16:08:11 huhu 2010-09-02T16:08:13 so they aren't broken 2010-09-02T16:09:32 Fallenou, im not sure if the lm32 supports unaligned stacks 2010-09-02T16:09:45 I am prone to change the RISC-ish CPUs to 4, avr to 1, and m68k to 4 2010-09-02T16:09:47 humm well that worked well with 0 2010-09-02T16:10:04 Before the change, it might have done its own computation of _CPU_Interrupt_stack_high 2010-09-02T16:10:24 DrJoel, i dont understand the semantics of this define 2010-09-02T16:10:35 @note This must be a power of 2 either 0 or greater than @ref CPU_ALIGNMENT. 2010-09-02T16:10:35 mwalle: Error: "note" is not a valid command. 2010-09-02T16:10:56 why can that be 0? 2010-09-02T16:11:11 it can be if you can have byte aligned stacks 2010-09-02T16:11:31 so shouldn't it be 1 then? 2010-09-02T16:11:33 byte aligned is what i call "non aligned" 2010-09-02T16:11:41 yep 2010-09-02T16:12:13 when it is a fault or slower, you want it to be non-zero 2010-09-02T16:12:33 so CPU_STACK_ALIGNMENT must be 2^n with (n=0,1,..) 2010-09-02T16:12:36 never 0 2010-09-02T16:13:18 in the most cases 1 for byte alignment, 2 for half word, and 4 for word alignment 2010-09-02T16:13:32 right.. 2010-09-02T16:13:49 ok 2010-09-02T16:13:53 and before cpus that cared probably had their own interrupt stack computation.. with the move to SMP .. it is now common 2010-09-02T16:13:58 so crappy comment 2010-09-02T16:14:07 yes ... 2010-09-02T16:14:11 dunno why i didn't think about that before 2010-09-02T16:14:14 you're fucking right 2010-09-02T16:14:32 shouldn't be 0 2010-09-02T16:14:59 Can one of you make a blanket sweep of the comment across the rtems/score/cpu .h files and submit a patch.. along with the porting guide? 2010-09-02T16:17:13 Fallenou, can you? i'm busy with class atm 2010-09-02T16:17:53 ok but what is a blanket sweep ? 2010-09-02T16:17:59 don't understand the words =) 2010-09-02T16:18:34 remove the comment / replace it with a better one in every cpu.h 2010-09-02T16:19:11 i guess i can put 1 instead of 4 for the CPU_STACK_ALIGNMENT for lm32 2010-09-02T16:19:18 ok i will do it, but tomorrow 2010-09-02T16:19:29 had to much wine at diner with guests 2010-09-02T16:19:42 Fallenou, well we are 32bit so imho this should be 4 for lm32 2010-09-02T16:20:00 Fallenou: don't drink and code. :-D 2010-09-02T16:20:03 what type of wine? 2010-09-02T16:20:09 hummm i knew about 4 bytes alignment restrictions for DMA, but i dunno for other things like stacks 2010-09-02T16:20:20 DrJoel: rosé :) 2010-09-02T16:20:28 even if its supported, its a mess to load a 32bit word from stack :) 2010-09-02T16:20:33 but knowing the type of DDR controller used by milkymist 2010-09-02T16:20:40 4 bytes aligned access should be better :) 2010-09-02T16:21:00 ahh... around my house, we are wine and beer fans.. had a nice bottle of cava last night 2010-09-02T16:21:23 cava N 2010-09-02T16:21:24 ? 2010-09-02T16:21:47 oh a spanish wine 2010-09-02T16:25:32 :) 2010-09-02T16:25:35 sounds nice 2010-09-02T16:25:37 never tested ! 2010-09-02T16:27:03 basically spanish champagne but since it isn't french we can't call it that.. 2010-09-02T16:27:31 hehe 2010-09-02T16:27:55 yeay champagne is just for a precise area in france :) 2010-09-02T16:27:58 in turkey they call it 'local champagne' for the tourists.. 2010-09-02T16:28:41 ahah 2010-09-02T16:29:17 Fallenou, but you must not name your sparkling wine champagne if doesnt come from that region, right? 2010-09-02T16:29:23 sparkling wine is the term in the States 2010-09-02T16:30:56 in germany theres "Wein" (wine), "Sekt" (sparkling wine) and "Champagner" (champagne) 2010-09-02T16:32:07 prost, a votre sante, cheers :) 2010-09-02T16:32:20 im going to bed now :) 2010-09-02T16:32:40 *** exception13 has left #rtems 2010-09-02T16:34:51 night 2010-09-02T16:40:05 gn8 :) 2010-09-02T16:40:10 night 2010-09-02T16:46:59 *** DrJoel has quit IRC 2010-09-02T17:42:04 *** dr__house has joined #rtems 2010-09-02T17:46:17 *** projectgus has joined #rtems 2010-09-02T19:55:36 *** kristianpaul has quit IRC 2010-09-02T19:57:41 *** kristianpaul has joined #rtems 2010-09-02T20:14:48 *** kristianpaul has quit IRC 2010-09-02T20:16:44 *** kristianpaul has joined #rtems 2010-09-02T20:31:46 *** kristianpaul has quit IRC 2010-09-02T20:34:18 *** kristianpaul has joined #rtems 2010-09-02T22:25:00 *** dr__house has quit IRC 2010-09-02T22:29:33 *** _Lucretia__ has quit IRC 2010-09-02T22:37:35 *** _Lucretia__ has joined #rtems 2010-09-03T00:25:57 *** exception13 has joined #rtems 2010-09-03T02:24:28 *** projectgus has quit IRC 2010-09-03T04:34:13 *** methril has joined #rtems 2010-09-03T05:27:06 *** madrazr has joined #rtems 2010-09-03T05:53:33 *** madrazr has quit IRC 2010-09-03T06:38:11 *** methril_work has joined #rtems 2010-09-03T06:38:58 *** madrazr has joined #rtems 2010-09-03T06:39:08 *** madrazr has joined #rtems 2010-09-03T07:12:00 *** kiwichris has quit IRC 2010-09-03T07:35:05 *** exception13 has left #rtems 2010-09-03T08:43:02 *** madrazr has quit IRC 2010-09-03T08:51:03 *** _Lucretia__ is now known as _Lucretia_ 2010-09-03T08:51:11 *** _Lucretia_ has joined #rtems 2010-09-03T09:15:18 *** DrJoel has joined #rtems 2010-09-03T09:15:18 *** ChanServ sets mode: +o DrJoel 2010-09-03T10:13:53 *** methril_work has quit IRC 2010-09-03T10:18:32 *** methril_work has joined #rtems 2010-09-03T10:49:39 *** tuxmaniac has joined #rtems 2010-09-03T10:50:38 *** tuxmaniac has quit IRC 2010-09-03T10:51:00 *** Alison_Chaiken has joined #rtems 2010-09-03T11:20:54 *** madrazr has joined #rtems 2010-09-03T13:29:48 *** exception13 has joined #rtems 2010-09-03T13:55:25 *** Alison_Chaiken has quit IRC 2010-09-03T14:09:11 *** methril has quit IRC 2010-09-03T14:10:49 *** DrJoel has quit IRC 2010-09-03T14:21:27 *** methril has joined #rtems 2010-09-03T14:55:41 *** krzabr has joined #rtems 2010-09-03T15:25:43 *** madrazr has left #rtems 2010-09-03T15:26:19 *** exception13 has left #rtems 2010-09-03T15:55:19 *** DrJoel has joined #rtems 2010-09-03T15:55:19 *** ChanServ sets mode: +o DrJoel 2010-09-03T15:55:25 *** DrJoel has quit IRC 2010-09-03T16:13:31 *** dr__house has joined #rtems 2010-09-03T16:37:06 *** rtemsLogger has joined #rtems 2010-09-03T17:16:23 *** DrJoel has joined #rtems 2010-09-03T17:16:23 *** ChanServ sets mode: +o DrJoel 2010-09-03T17:16:51 *** krzabr has quit IRC 2010-09-03T17:19:46 *** methril_work has quit IRC 2010-09-03T17:28:10 *** DrJoel has quit IRC 2010-09-03T19:26:15 *** kiwichris has joined #rtems 2010-09-03T21:54:01 *** kristianpaul has quit IRC 2010-09-03T21:54:10 *** kristianpaul has joined #rtems 2010-09-03T22:25:04 *** dr__house has quit IRC 2010-09-04T04:22:13 *** exception13 has joined #rtems 2010-09-04T05:23:31 *** tuxmaniac has joined #rtems 2010-09-04T06:03:56 *** tuxmaniac has quit IRC 2010-09-04T08:11:44 *** dr__house has joined #rtems 2010-09-04T08:11:44 *** dr__house has joined #rtems 2010-09-04T09:43:12 *** tuxmaniac has joined #rtems 2010-09-04T09:43:33 *** tuxmaniac has quit IRC 2010-09-04T09:44:13 *** aanjhan has joined #rtems 2010-09-04T10:01:18 *** exception13 has left #rtems 2010-09-04T10:55:52 *** lekernel has quit IRC 2010-09-04T11:22:25 *** madrazr has joined #rtems 2010-09-04T11:33:34 *** madrazr has quit IRC 2010-09-04T11:52:33 *** madrazr has joined #rtems 2010-09-04T14:35:56 *** mw|mobile has joined #rtems 2010-09-04T14:41:59 *** mw|mobile has quit IRC 2010-09-04T14:47:59 *** mw|mobile has joined #rtems 2010-09-04T15:06:29 *** mw|mobile has quit IRC 2010-09-04T16:04:25 *** dr__house has quit IRC 2010-09-04T16:56:34 *** madrazr has quit IRC 2010-09-04T17:11:02 *** madrazr has joined #rtems 2010-09-04T17:24:30 *** madrazr has quit IRC 2010-09-04T17:59:50 *** lekernel has joined #rtems 2010-09-04T18:18:09 *** exception13 has joined #rtems 2010-09-04T18:36:15 *** exception13 has left #rtems 2010-09-04T19:47:47 *** dr__house has joined #rtems 2010-09-04T23:22:35 *** dr__house has quit IRC 2010-09-04T23:24:44 *** dr__house has joined #rtems 2010-09-05T00:12:12 *** dr__house has quit IRC 2010-09-05T02:27:52 *** rtemsLogger` has joined #rtems 2010-09-05T02:27:53 -niven.freenode.net- [freenode-info] if you're at a conference and other people are having trouble connecting, please mention it to staff: http://freenode.net/faq.shtml#gettinghelp 2010-09-05T02:28:29 *** rtemsLogger has quit IRC 2010-09-05T03:54:31 *** methril has quit IRC 2010-09-05T03:55:21 *** methril has joined #rtems 2010-09-05T05:03:53 *** madrazr has joined #rtems 2010-09-05T07:19:08 *** dr__house has joined #rtems 2010-09-05T07:19:08 *** dr__house has joined #rtems 2010-09-05T08:17:57 *** exception13 has joined #rtems 2010-09-05T09:53:59 *** exception13 has left #rtems 2010-09-05T10:58:14 *** exception13 has joined #rtems 2010-09-05T13:12:19 *** dr__house has quit IRC 2010-09-05T13:33:35 *** madrazr has quit IRC 2010-09-05T13:37:22 *** madrazr has joined #rtems 2010-09-05T14:08:28 *** methril has quit IRC 2010-09-05T14:21:08 *** methril has joined #rtems 2010-09-05T15:16:26 *** methril has quit IRC 2010-09-05T15:16:26 *** exception13 has quit IRC 2010-09-05T15:23:10 *** methril has joined #rtems 2010-09-05T16:57:13 *** madrazr has quit IRC 2010-09-05T17:00:41 *** madrazr has joined #rtems 2010-09-05T17:04:59 *** madrazr has quit IRC 2010-09-05T17:33:40 cd .. 2010-09-05T18:06:07 *** Alison_Chaiken has joined #rtems 2010-09-05T18:20:48 *** Alison_Chaiken has quit IRC 2010-09-05T18:21:20 *** Alison_Chaiken has joined #rtems 2010-09-05T18:56:18 *** dr__house has joined #rtems 2010-09-05T20:42:45 *** kiwichris has quit IRC 2010-09-05T20:57:26 *** dr__house has quit IRC 2010-09-05T22:05:30 *** jacob123 has joined #rtems 2010-09-05T22:09:24 *** jacob123 has left #rtems