RTEMS IRC Log, 27 Jul 2009

*** tuxmaniac has quit IRC00:23
*** ChanServ has quit IRC00:23
*** dr__house has quit IRC00:24
*** ChanServ has joined #rtems00:28
*** irc.freenode.net sets mode: +o ChanServ00:28
*** tuxmaniac has joined #rtems00:28
*** chrisjohns has quit IRC00:48
*** kiwichris has joined #rtems00:49
lee_kiwichris, morning?02:30
lee_kiwichris, how was the game on sat?02:30
*** bazinski has joined #rtems03:23
*** cocan has joined #rtems05:02
*** chrisjohns2 has joined #rtems05:20
*** chrisjohns2 has quit IRC05:24
*** chrisjohns has joined #rtems05:24
*** tuxmaniac has quit IRC05:24
*** kiwichris has quit IRC05:24
*** ChanServ has quit IRC05:24
*** kiwichris has joined #rtems05:25
*** tuxmaniac has joined #rtems05:26
*** chrisjohns has quit IRC05:26
*** chrisjohns has joined #rtems05:27
*** cocan_ has joined #rtems05:30
*** cocan has quit IRC05:31
*** chrisjohns has quit IRC05:32
*** chrisjohns has joined #rtems05:33
*** ChanServ has joined #rtems05:38
*** irc.freenode.net sets mode: +o ChanServ05:38
*** cocan_ has quit IRC05:50
*** cocan has joined #rtems05:50
cocanchrisjohns: Hi Chris05:58
*** cocan has quit IRC06:59
*** cocan has joined #rtems06:59
*** th_d has joined #rtems07:05
chrisjohnsHi Thomas07:34
chrisjohnsth_d, hi07:34
*** chrisjohns has quit IRC07:43
*** th_d has quit IRC07:48
*** cocan_ has joined #rtems07:59
*** cocan has quit IRC08:00
*** cocan_ is now known as cocan08:00
*** cocan has quit IRC08:20
*** DrJoel has joined #rtems08:50
*** ChanServ sets mode: +o DrJoel08:50
*** dr__house has joined #rtems10:47
dr__houseDrJoel: good morning :)10:47
*** dr__house has quit IRC10:55
*** dr__house has joined #rtems10:58
dr__housesorry about that, network problems :(10:58
* DrJoel is on his way to lunch .. brb11:41
DrJoeldr__house: before i leave.. how did the numbers in the tarball look to you?11:42
dr__houseDrJoel: they were awesome:D11:42
dr__houseDrJoel: we reached 99.5%!!!!!!!!!!!!!!!!!!11:43
dr__house:D:D11:43
*** cocan has joined #rtems11:43
DrJoeli am picking at the results of running with posix disabled.  Looks like some handlers in the supercore are only used by posix so I am fixing the makefile not to build them11:47
* DrJoel will be back in a couple of hours11:47
dr__houseDrJoel: cool :)11:48
agrierstill no luck with updated newlib.  I have a suspicion I'm missing some patches.11:58
agrierIn file included from /var/obj/gsoc/sourceware/gcc-4_4-branch/newlib/libc/include/dirent.h:6,11:59
agrier                 from /projects/gsoc/sourceware/gcc-4_4-branch/newlib/libc/posix/closedir.c:41:11:59
agrier /var/obj/gsoc/sourceware/gcc-4_4-branch/newlib/libc/include/sys/dirent.h:10:2: error: #error "<dirent.h> not supported"11:59
agrierit's like configure isn't setting the include directories up correctly11:59
*** cocan_ has joined #rtems12:00
*** cocan has quit IRC12:03
*** cocan_ is now known as cocan12:03
agrierI know RTEMS provides a dirent.h12:03
cocanDrJoel: hi12:21
cocanDrJoel: is there a way to get a file from a guest computer (running RTEMS in qemu) on the host computer?12:23
*** cocan has quit IRC13:33
*** cocan has joined #rtems13:37
* dr__house is away: Gone away for now14:01
* dr__house is back.14:04
DrJoelcocan: there are two ways I know of.  (1) to use the virtual fat support in qemu and the RTEMS FAT filesystem and (2) use the RTEMS nfsclient to access the host filesystem.14:34
*** cocan_ has joined #rtems14:46
cocan_DrJoel: great...does anything else needs to be configured in order to make use of FAT instead of IMFS?  http://www.rtems.com/wiki/index.php/Using_the_RTEMS_DOS_File_System14:53
DrJoeli don't think so since FAT is mounted (not the root).  Check out the fileio or network-demo/telnetd for some examples.  Chris should be able to help with any details14:54
*** ChanServ has quit IRC14:55
cocan_DrJoel: ok...will do so...I am trying to get the file that contains the trace information...in order to decode on the host14:59
* dr__house leaves for the day. Have a good day DrJoel :)15:00
DrJoelgood night.. your new patch is generating a tarball now.15:00
dr__houseDrJoel: cool! :)15:01
*** dr__house has quit IRC15:01
*** cocan has quit IRC15:05
*** ChanServ has joined #rtems15:13
*** irc.freenode.net sets mode: +o ChanServ15:13
*** cocan_ has quit IRC15:18
*** roxanal has quit IRC15:21
tuxmaniacHi DrJoel15:26
tuxmaniacend of day for me. good night. will have to wake p early tomorrow and make more commits :)15:27
*** cocan has joined #rtems15:30
*** roxanal has joined #rtems15:36
cocankiwichris: Hi Chris...are you on?17:43
*** cocan has quit IRC18:19
*** chrisjohns has joined #rtems18:23
*** DrJoel has quit IRC18:35
*** roxanal has quit IRC18:45
*** cocan has joined #rtems19:10
chrisjohnscocan, hi, I am fine.19:14
*** chrisjohns has quit IRC20:43
*** cocan has quit IRC20:45
*** chrisjohns has joined #rtems21:47
*** chrisjohns has quit IRC23:40
*** madrazr has joined #rtems00:04
*** lee21 has joined #rtems00:59
*** madrazr has quit IRC02:07
*** lee21 has quit IRC02:35
*** lee22 has joined #rtems02:36
*** ChanServ has quit IRC03:02
*** tuxmaniac has quit IRC03:02
*** kiwichris has quit IRC03:02
*** lee22 has quit IRC03:02
*** mwalle has quit IRC03:02
*** agrier has quit IRC03:02
*** rokka has quit IRC03:02
*** bazinski has quit IRC03:02
*** lee_ has quit IRC03:02
*** ChanServ has joined #rtems03:03
*** lee22 has joined #rtems03:03
*** tuxmaniac has joined #rtems03:03
*** kiwichris has joined #rtems03:03
*** bazinski has joined #rtems03:03
*** mwalle has joined #rtems03:03
*** rokka has joined #rtems03:03
*** agrier has joined #rtems03:03
*** lee_ has joined #rtems03:03
*** irc.freenode.net sets mode: +o ChanServ03:03
*** lee22 has quit IRC04:08
*** lee21 has joined #rtems04:12
*** dr__house has joined #rtems04:16
*** dr__house has quit IRC04:52
*** madrazr has joined #rtems05:45
*** madrazr has quit IRC06:09
*** madrazr has joined #rtems06:12
*** dr__house has joined #rtems08:39
*** roxanal has joined #rtems08:39
*** switnick has joined #rtems09:26
*** dr__house has quit IRC10:45
*** madrazr has quit IRC11:13
*** madrazr has joined #rtems11:17
*** dr__house has joined #rtems11:31
*** DrJoel has joined #rtems12:12
*** ChanServ sets mode: +o DrJoel12:12
DrJoelroxana and aaron: did you see my email with the framebuffer.h and user side configuration?12:13
DrJoeldr__house: the arm coverage has work left. :( i committed your changes.  Do you have one you want to do next?12:13
dr__houseDrJoel: no, I thought arm would be the next thing. What about other archs?12:15
DrJoelskyeye and tsim are the only simulators with coverage builtin that we have BSPs for.. so back to the test case coal mine. :)12:16
tuxmaniacHi DrJoel12:19
DrJoeli don't have any feel for which would be best for you to tackle.  There are only 29 left.  An easy one is probably the pthread_create with a stack you provide (set stackaddr)12:20
DrJoeltuxmaniac: hello12:20
DrJoelbut there are some posix msgq left, posix key, and posix sporadic server with ~4-5 each so those are good candidates also12:21
tuxmaniacDrJoel: I replioed to your email. I agree a patch over the cvs is easiest. But I have added a few files and folders too and I cant get them in the cvs diff12:21
DrJoeljust tar them up from the top of the tree and I will untar them when I apply th epatch12:24
tuxmaniacok. today I will be commiting more code. Will send across a tarball by end of day today12:25
DrJoelno problem.12:28
dr__houseDrJoel: cool, will check them out then12:29
DrJoelpick what looks interesting and doable and I will review the explanations for them to make sure about them12:31
dr__houseDrJoel: sure12:33
switnickDrJoel: Hello12:48
DrJoelswitnick: hi.. have you made any progress?12:49
switnickI think so12:49
switnickI am trying an asm implementation12:49
switnickis there any reason not to do this?12:49
DrJoelof ?12:49
switnickcontext init12:49
DrJoeloffsets into the structure get fixed automatically in C but other that.. no12:50
DrJoeldid you swap the order of loading the 2 halves of the sp in restore?12:50
switnicki did12:50
DrJoeldid that help?12:50
switnicknot really12:50
DrJoelcan you see the value of the sp in gdb as it is set/12:51
switnickyes12:51
switnickit sets fine12:51
DrJoeldoes it return to _Thread_Handler with it ret's?12:51
switnickno12:51
DrJoelwhere does it go/12:51
switnickthe value that pops of the stack is a mystery12:52
DrJoelthat's what confuses me..12:52
switnickwhen I look at the stack is looks ok12:52
switnickbut after the ret an unknown value shows up12:52
switnicknot sure where from12:52
DrJoelis it in memory near the sp? above or below?12:53
switnickyes12:53
switnickat and below i think12:54
switnickobove12:54
switnickabove**12:54
switnickI will try it on the hardware that Eric sent12:55
DrJoelhmm.. maybe we have miscalculated where to set the sp and return address in relation to one another12:55
switnickI have been moving them around and redoing the calcs12:55
switnickseems right12:55
switnickI will send the diff to you12:56
switnickif you would like12:56
switnickor I can just do some more testing on hardware12:56
DrJoelbefore you send me a diff can you look at a trace and see what address the ret instruction reads versus where we put the entry point.  Maybe we can just fix that.12:58
DrJoeltry that .. if no progress in an hour or so, email me a diff and i will take a swing at it12:59
switnickok12:59
switnickjust a second12:59
switnickI believe that pop is post increment13:04
switnickif that is true then the address is corect13:04
DrJoeli don't trust my interpretation when the code doesn't work. :D  If we can put dummy contents above and below where we think the SP should be, then what stack location is read should be obvious13:09
switnickok I will do that13:10
switnickis there a posibility that the problem is with the simulator?13:11
DrJoelpossible .. not likely.. if the code runs on hw and not the simulator yes but otherwise no.13:12
DrJoela lot of C compiler generated code runs on the simulator13:12
DrJoelkiwichris: awake?13:13
*** cocan has joined #rtems13:17
*** cocan has quit IRC13:35
dr__houseDrJoel: leaving early tonight. My back's killing me, catch you tomorrow. Have a good day :)13:42
*** dr__house has quit IRC13:43
*** cocan has joined #rtems14:00
cocanDrJoel: hi....any idea why I am getting this? Program heap: free of bad pointer 1C026C -- range 1BFB78 - 7FFFFF814:04
cocanDrJoel: and also Read IDE Disk Partition Table: error: ide partition table not found: RTEMS inconsistency detected14:05
lee21cocan: think he didn't feel to well and left early, something about his back14:06
DrJoeli'm here.. :)14:06
lee21oops...14:06
cocanlee21: :D14:06
lee21;)14:07
DrJoelcocan: usually that means you freed a pointer that wasn't malloc'ed14:07
DrJoeland in that case .. look at the message.  you definitely freed a pointer not in the heap.14:07
DrJoelCheck your .num file.  It looks like a real variable address14:07
DrJoelwhoops.. misread..14:07
DrJoelprobably a double free or an offset inside a malloc'ed area14:08
DrJoel1C026C is in the heap (range 1BFB78 - 7FFFFF8)14:08
cocanDrJoel: taking a look right now14:09
DrJoelyou can break on that print in gdb and backtrace to see where it came from14:25
cocanwill do that if I will not be able to fix it right now14:29
DrJoel:)14:29
cocanI've just tried fileio example and it works ok...I can mount partitions, create directories and see the changes in the image disk14:40
*** madrazr has quit IRC14:40
DrJoelcocan: wonderful! is this with the virtual fat under qemu?14:43
cocanDrJoel: using a disk image...not the fat: option14:45
cocanDrJoel: using the VFAT in Qemu will map the free space of my HDD as a slave...and it doesn't work..not sure why14:46
cocanDrJoel: but I can always use a disk image and copy the file using mtools14:47
switnickDrJoel: I have a stack question.15:18
DrJoelswitnick: shoot15:20
switnickI was trying to simulate the ret instruction by popping twice from the stack15:21
switnickto see what is there in the trace15:21
DrJoeland ...15:21
switnickthen push them back on15:21
switnickso I get the right address here15:21
switnickbut then when the ret occurs it goes somewhere else15:22
switnicknot to the value I just pushed to the stack15:22
DrJoelvery weird...15:23
DrJoelcan you send me your patch that restores the sp in the right order .. I will play some15:23
switnickok15:24
switnickjust sent the diff15:29
DrJoelok.. i will start building15:29
*** lee21 has quit IRC15:45
* agrier is back in cross-compiler business15:49
agriersourceware newlib HEAD: fail15:49
agriernewlib 1.17.0 with RTEMS patches: success!15:49
DrJoelagrier: what failed on head? which gcc?15:50
agriergcc 4.4.215:50
agrierI can't get newlib head to compile15:51
agrierwent back to 1.17.0 with recent patches and it worked...15:52
DrJoelwhat error?15:55
agrierdirent.h not supported15:55
agrierit looked like newlib was missing an include path, or some autoconf magic wasn't being properly triggered15:56
DrJoelhmm.. ok.. maybe the gcc needed a patch15:56
DrJoeli don't know.. :(15:56
agriersame gcc -- all I changed was newlib from -HEAD to 1.17.0 with RTEMS patches15:56
agrieryou're tracking -HEAD, right Joel?15:58
DrJoelyep.. and gcc -HEAD.  built about a week ago.  need to try it again apparently15:59
agriercould you run a diff against your newlib branch and email it to me, just for kicks?16:01
DrJoelagrier: so small I just pasted it :-D16:04
agrieroh really... wow16:04
agrierthis may be NetBSD-specific.  I am not enough of an autotools avatar to precisely determine what is going on16:04
DrJoelswitnick: why does the Context_Control only have two fields now?  The context save and restore assume a lot more is in there16:05
switnickto make the switch  and init simpler16:06
switnickthis is also the way alot of other rtos' i looked implemented their context switch16:06
switnickseemed to be simpler to me16:06
switnickis there something wrong with it?16:06
DrJoelbut there are registers which have to be preserved across a subroutine call and they must be saved/restored.16:07
DrJoeldid you account for all those pops during the init?16:07
switnickyes16:07
DrJoelok..  just surprised when looking at the code.. no big deal16:07
switnickok16:07
switnickits good to have someone asking these questions16:08
switnickDrJoel: any idea whats going on?16:45
DrJoelswitnick: I think I am onto something16:45
DrJoelthe_context = 0x2BB516:45
DrJoelentry = 0x74A516:45
DrJoelsp = 0x3D5816:45
DrJoel0x3D46 0x2B 0x00 0x00 0x00   0x00 0x00 0x00 0x0016:45
DrJoel0x3D4E 0x00 0x00 0x00 0x00   0xA5 0x74 0xA5 0x7416:45
DrJoel0x3D56 0xA5 0x74 0xA5 0x74   0xA5 0x74 0xA5 0x7416:45
DrJoelI went back to the C init context.. and (think) I accounted for what you push/pop on the stack.  I can not get to the Init task16:46
DrJoelI am writing the entry point into 3 locations in the starting stack which isn't right but I need to track down which one is giving the desired results16:46
switnickok16:47
switnickdid you look at the trace of what I sent?16:47
DrJoeli got 2 diff files from you in 2 emails?16:52
switnickwasnt sure that the attachment was sent in the first16:52
switnickthey are the same16:52
switnickignore one of them16:52
switnicksorry for the spam16:52
DrJoelyou are the least of my spam.. I feel like I am close...16:52
switnickDrJoel: I sent the result of the trace to you.17:13
switnickI'm going home now17:13
DrJoelok .. thanks.. I think i am onto something.  I think the context init matches the restore but the save doesn't match restore.. having trouble getting into gdb now :(17:13
switnickwill be at my computer later tonight17:14
DrJoelok.. i will work on this a while.. it is close I think17:14
switnickok17:15
switnickdo you have avr-gdb?17:15
switnickthanks for the help17:17
switnicktalk to you tmr17:17
switnickgood night17:17
DrJoelyes.. i can step17:18
DrJoelgood night17:18
*** switnick has quit IRC17:18
agrieris rtems/stringto.h supposed to be checked in to CVS?  ;)18:25
DrJoelyes .. and it is.. (according to my tree)18:28
DrJoel   Repository revision:1.2/usr1/CVS/rtems/cpukit/libmisc/stringto/stringto.h,v18:29
DrJoelaccording to cvs stat18:29
DrJoeldid you bootstrap cpukit?18:29
DrJoelfwiw I switched pc386 fbvga_X to use the frame_buffer_ naming so we can test out the new confdefs.h work18:29
* DrJoel is heading home.. i had enough of a breakthrough for josh that I hope he is moving again18:30
* DrJoel waves night18:30
*** DrJoel has quit IRC18:31
agrierahh bootstrap18:44
*** chrisjohns has joined #rtems18:51
agrierwoohoo it's all working again19:41
agrierI was afraid I was going to have to dig out the goat book19:41
*** switnick has joined #rtems19:43
*** chrisjohns has quit IRC20:00
*** chrisjohns has joined #rtems20:02
*** cocan has quit IRC20:02
agriertime to head home20:12
*** switnick has quit IRC20:20
roxanalcan anyone help me how is preinstall.am generated? it says it is an automatic file21:47
roxanal*help me understand21:47
roxanalI found out ./bootstrap -p :)21:55
*** switnick has joined #rtems22:09
*** switnick has quit IRC22:46
*** lee21 has joined #rtems23:11
*** chrisjohns has quit IRC23:39
tuxmaniacroxanal: yes. and also ensure you do a ./bootstrap again. I sometimes found just ./bootstrap -p isnt enough00:53
*** dr__house has joined #rtems00:57
*** lee21 has quit IRC02:05
*** sebhub has joined #rtems02:16
*** lee21 has joined #rtems02:16
*** dr__house has quit IRC02:59
*** dr__house has joined #rtems04:00
*** madrazr has joined #rtems04:01
*** madrazr has quit IRC04:39
*** madrazr has joined #rtems05:17
*** gnumaniac has joined #rtems05:32
*** dr__house has quit IRC05:33
*** madrazr1 has joined #rtems05:42
*** madrazr has quit IRC05:42
*** bazinski has quit IRC05:42
*** bazinski has joined #rtems05:50
*** gnumaniac has quit IRC05:54
*** dr__house has joined #rtems06:17
*** dr__house has quit IRC07:30
*** dr__house has joined #rtems07:31
*** dr__house has quit IRC07:32
*** madrazr1 has quit IRC08:24
*** DrJoel has joined #rtems08:31
*** ChanServ sets mode: +o DrJoel08:31
*** switnick has joined #rtems08:45
*** cocan has joined #rtems09:11
DrJoelswitnick: did you see the email? does it run on hardware?09:13
switnickDidn't have a power supply last night09:13
switnickjust about to set it up09:13
switnickand yes I did see the email09:14
DrJoelok.. it was running ticker to the end correctly.  But only when I did a printk.09:14
switnickI had a similar problem09:14
DrJoelI suspect we missed a register in the context setup.. like frame pointer or something09:14
switnickok09:14
DrJoeldoes the start.S set anymore registers?  I was going to look at the assembly for printk and see what it did that might help.09:15
DrJoelbut a check on hardware at this point would be good to verify our status09:15
switnickok will do09:15
switnicknot so sure about the setup09:15
switnickmight take some time09:15
DrJoeli wonder if the reg that is supposed to be zero actually is when we initialize the contexts..09:21
tuxmaniacDrJoel: _read_SDR1 is not there in libcpu/spr.h I saw it is used by gen82xx bsp already. But for psim this does not get into the header file. Don't know if I am missing something09:30
tuxmaniachence wrote my own09:30
*** th_d has joined #rtems09:33
roxanalDrJoel: good morning09:35
*** cocan has quit IRC09:36
*** dr__house has joined #rtems09:43
dr__houseDrJoel: good morning09:43
*** madrazr has joined #rtems09:44
*** cocan has joined #rtems09:52
DrJoelhi all10:03
DrJoeltuxmaniac: no problem on writing your own.. just hard code.. easier to reuse :-D10:03
DrJoeldr__house: any thoughts on which case to tackle next?10:03
dr__houseDrJoel: Not yet, looking at them still10:04
roxanalDrJoel: I don't know exactly what happened with the patch I sent you10:04
DrJoeldr__house: anything looking like a likely candidate?10:05
DrJoelroxanal: what do you mean?   It merged cleanly.  All i did was kill tabs and fix formatting that has been there since the file was first submitted 9 years ago10:06
roxanalDrJoel: I can't build rtems anymore there are some weird <<<<<<< fb_vga.c coments that were not put by me10:06
DrJoelthose sound like CVS conflicts.. remove the file and update again10:07
roxanalaaa ok10:07
DrJoeldid you see a "C" in the 1st column on that file?  Instead of a "U"?10:07
roxanalbo10:08
roxanalno10:08
DrJoelhmm.. ok..10:08
roxanalLet me try to delete and update again10:08
roxanalDrJoel: I updated it and looks ok now10:12
*** th_d has quit IRC10:12
DrJoelroxanal: great10:12
DrJoeldr__house: how do the posix message queue or the priority ceiling mutex cases look?10:12
DrJoelfwiw i killed the region resize case last night10:13
tuxmaniacroxanal: it was cvs merge conflicts. you might have changed some things in the same place as that in the cvs. hence the <<< comments10:13
DrJoeland am working on the last killinfo cases.. posix signals are weird10:13
dr__houseDrJoel: oh ok. I thought that might have been interesting :P10:13
roxanaltuxmaniac: tks10:14
dr__houseDrJoel: I meant the region resize10:14
tuxmaniacand to your question about mmap status ages back, it is slightly behind schedule. catching up10:14
tuxmaniacroxanal: ^10:14
DrJoeldr__house: which has more code to kill?10:15
dr__houseDrJoel: I think I shall start off with priority ceiling mutex10:16
roxanaltuxmaniac: don't worry about mmap I can work without it for a while10:17
DrJoeldr__house: ok.. i hope my explanation is right.. verify that it makes sense when you look at the assembly  before doing more10:17
dr__houseDrJoel: sure10:17
dr__houseDrJoel: and then maybe tackle all the mutex cases at once10:18
dr__houseDrJoel: there seems to be one other mutex initialization case10:18
DrJoeli worry that some of these weird small cases might be cases where it isn't a missing test, it is a code rework .. so if you have doubts let me double check10:19
dr__houseDrJoel: in the report for coremutexsurrender.c the explanation says ""10:20
dr__houseThere must not be a test of a priority ceiling mutex getting released10:20
dr__housewhich results in a thread being unblocked and the UNBLOCKED thread's10:20
dr__housepriority being elevated.10:20
dr__houseDrJoel: I didn't get "There must not be .."10:21
DrJoelfirst .. new dedicated test :-D10:23
DrJoeltwo threads... Init has any priority less important than ceiling.. acquires priority ceiling mutex10:23
DrJoelcreates 2nd thread with priority == ceiling priority of mutex10:23
DrJoelInit sleeps .. 2nd thread runs.. blocks on priority ceiling mutex10:24
dr__houseok10:24
DrJoelInit wakes up.. releases mutex.. it is transferred to 2nd thread.. which is already at the priority ceiling so it does NOT need its priority changed.. thus you take the branch and do not have to do the _Thread_Change_priority10:24
DrJoelso this is definitely a case where the annulled instruction shows us we missed a test case :-D10:25
dr__houseok10:26
dr__houseDrJoel: Dinner time, be back in a little while10:26
DrJoeldefinitely easier to create a new test :-D10:26
* dr__house is away: afk10:26
DrJoelno problem..10:26
dr__houseDrJoel: heh, yeah :)10:27
*** madrazr has quit IRC10:28
*** sebhub has quit IRC10:28
*** madrazr has joined #rtems10:48
*** cocan has quit IRC10:54
*** cocan has joined #rtems11:00
* dr__house is back.11:00
DrJoelhow was dinner?11:04
* DrJoel was asked to lunch by his daughter.. she is picking me up in about 25 minutes11:04
dr__houseDrJoel: dinner was usual :)11:07
dr__houseDrJoel: thats nice :)11:07
*** lee21 has quit IRC11:22
dr__houseDrJoel: where should the new test case go in the testsuite?11:26
*** switnick has quit IRC11:30
DrJoelI think you can hit this with the classic api so it should go in the sptests11:33
DrJoelsp63?11:33
dr__houseDrJoel: cool11:38
*** switnick has joined #rtems11:42
*** lee21 has joined #rtems12:09
dr__houseDrJoel: where can I access the logs from rtemsLogger? Are they hosted anywhere?12:23
tuxmaniacdr__house: rtems.com/irclogs12:23
dr__housetuxmaniac: cool, thanks :)12:23
switnickDrJoel: have everything running now12:24
switnickseems to be running very slowly12:25
switnickis this normal?12:25
*** lee21 has quit IRC12:43
*** lee21 has joined #rtems12:47
DrJoelswitnick: on real hardware?12:49
switnickyes12:50
switnickhaving a few interface problems i think12:50
switnickjust wondering how long the hello sample should take to run12:50
DrJoelit is faking out the passage of time by having a special idle task call clock tick until it has gone long enough so definitely no correspondence to real time12:50
switnick??12:50
DrJoelhello is pretty quick normally12:51
DrJoeldoes the simulator tell you how many instructions it executes?12:51
DrJoelor can we approximate it by the number of lines in the trace?12:51
switnickyes12:51
switnickdid some calc12:51
DrJoeland ..?12:51
switnickshould be less than a second12:51
switnickso I must have a problem12:52
switnickI'm going to try and setup ddd12:52
switnickhave a few ideas12:52
switnickwill get back to you with progress12:53
DrJoelok..12:53
roxanalDrJoel: did you got/read my email about the Nano-X license?12:56
DrJoelyes.. need to read that license and think12:56
roxanalok let me know... that is the next step I have to think about12:57
roxanalif we can12:58
roxanalnot have them in I have to find a solution12:58
DrJoelnd if You include a notice stating that the Source Code version of the Covered Code is available under the terms of this License, including a description of how and where You have fulfilled the obligations of Section 3.2.13:00
DrJoelagrier: needs to confirm my reading but I think that the MPL is compatible with the RTEMS GPL+exception intention.  We do not want you to have to distribute any source for anything RTEMS related with your RTEMS application.  So it would be better to avoid MPL I think13:01
DrJoelhttp://www.svgalib.org/ should have similar code which is appropriately licensed for inclusion in RTEMS13:03
roxanalOK ... I was looking at svgalib already ;)13:04
agrierwait wait wait... can't we use the existing routines from vgainit.c ?13:04
agrieror perhaps they were being used already?13:08
* agrier looks at linker trace to see where ega_hwinit and ega_hwterm are coming from13:08
agrierI had been assuming they were coming from nano-X, but I could be mistaken.13:09
roxanalagrier: ega_* are coming form vgainit.c13:09
roxanalwhich is a file that got into rtems a while ago13:10
roxanalbut is the same one that nanox is using13:10
agrierwe could get greg to clarify the licensing of that file13:10
DrJoeloh.. it is an old version with the license OK13:11
DrJoel * Permission is granted to use, distribute, or modify this source,13:11
DrJoel * provided that this copyright notice remains intact.13:11
DrJoelno fear13:11
roxanalright now the file has a statement that we have "Permission is granted to use, distribute, or modify this source,13:12
roxanal * provided that this copyright notice remains intact."13:12
DrJoeland that's cool.  We complied and there are no further license issues.13:12
roxanaljust to clarify: it is ok to use it then13:13
DrJoelyes.  it is ok to continue to use.13:15
roxanalThank you :)13:15
*** madrazr has quit IRC13:21
*** madrazr has joined #rtems13:21
*** cocan has quit IRC13:33
*** cocan has joined #rtems13:34
*** cocan has quit IRC13:45
DrJoeltuxmaniac: looking forward to the races (saw status on fb)13:49
tuxmaniacDrJoel: actually I am not able to read data from teh segment registers because of this news :P13:49
DrJoelLOL13:51
tuxmaniacDrJoel: I am seeing _read_DAR functions used in vector_init.c but could not find its definition. a grep doesnt turn up anything14:08
tuxmaniacDrJoel: c/src/lib/libcpu/powerpc/new-exceptions/bspsupport/vectors_init.c14:08
DrJoelty for the path :_14:08
tuxmaniacI wonder how it compiles then14:08
DrJoel;-D14:08
DrJoelSPR_RO(DAR)14:09
DrJoelinstantiates the method like I tried to show in my email14:09
*** madrazr has quit IRC14:09
DrJoel#include <libcpu/spr.h>14:09
DrJoel was near the top.. that's all that's required to instantiate the SPR methods14:09
tuxmaniacthat doesnt have the definition of _read_DAR14:10
tuxmaniaci mean spr.h14:10
DrJoelSPR_RO(DAR)14:11
tuxmaniacaah I got it14:11
DrJoelfancy schmancy macro :-D14:11
tuxmaniaci was looking for _read_DAR itself like _read_SR14:11
DrJoelsaves what you went through14:11
tuxmaniacDrJoel: true :S14:11
*** cocan has joined #rtems14:35
DrJoeldr__house: still there?15:26
dr__houseDrJoel: yep15:26
dr__housewas replying to your mail15:26
DrJoelthe mqueuedelete path was unreachable.  you have to call mq_unlink before mq_delete and it frees the memory15:27
DrJoelso 24 cases now and you get to ignore that one. :-D15:27
dr__houseDrJoel: great!15:27
DrJoeli will do another run to verify but that leaves you focused on two big chunks in mq creation15:27
dr__houseDrJoel: yeah cool.15:28
dr__houseDrJoel: will look into them tomorrow, time for me to head off for the day :). Have to wake up early tomorrow.15:28
DrJoelyour two cases should be straight forward (I hope). If you have trouble with the mq's read the opengroup.org man pages15:28
dr__houseDrJoel: Sure. Catch you tomorrow. Have a good day.15:29
*** dr__house has quit IRC15:30
DrJoeltuxmaniac: what's your code.google.com page?16:48
*** cocan has quit IRC16:50
switnickDrJoel: Still havent been able to fix the problem16:55
switnickI've sent an email to eric asking for help16:55
DrJoeli saw it.  I have no idea what's up with that.16:57
switnickIt is working but really slowly16:57
DrJoeli guess it would be time to see if you can spot something in the printk disassembly that would make it "fix" the runs16:57
DrJoeldid you calc how many instructions and how long it should ahve taken?16:58
switnick~3800016:58
switnickI set a breakpoint at boot_card16:59
switnickand got it to run successfully16:59
DrJoelthe trace time stamp appears to be nanoseconds.. I terminated at Init in ticker.exe and it took 38619000 to get there17:02
DrJoelso I think ~38.s milliseconds.  to initialize RTEMS.  Right?17:03
switnickyea17:04
switnickthats what I got aswell17:04
DrJoelso anything you call slow is wrong. :-D17:04
DrJoelmaybe real hw has a clock we have to touch17:05
DrJoelor mem controller.. max wait states by default?17:05
switnickI set the osc to 8Mhz17:06
switnickso even if its taking 4 or 5 clock cycles per instructions cycles should be running faster that ~38 ms17:07
switnickI think it might be the connection from comp to debugger17:08
DrJoeldo you have any breakpoints?17:09
switnickyes17:09
DrJoeldoes it run faster without them?17:09
switnickI got one to break at boot card17:09
switnickno, even stepping takes a long time17:10
DrJoelunix or windows?17:10
DrJoeldoes the kit come with a known working executable to try/17:12
DrJoelif not ask eric for one17:12
* DrJoel is out the door17:12
DrJoelbye all17:12
*** DrJoel has quit IRC17:12
*** switnick has quit IRC17:14
*** cocan has joined #rtems17:45
*** DrJoel has joined #rtems19:13
*** ChanServ sets mode: +o DrJoel19:13
*** chrisjohns has joined #rtems19:32
*** cocan has quit IRC19:43
*** cocan has joined #rtems19:44
*** DrJoel has quit IRC19:54
*** cocan has quit IRC20:57
*** DrJoel has joined #rtems21:18
*** ChanServ sets mode: +o DrJoel21:18
*** DrJoel has quit IRC21:30
*** chrisjohns has quit IRC23:46
*** chrisjohns has joined #rtems23:46
*** madrazr has joined #rtems00:43
*** sebhub has joined #rtems01:54
sebhubgood morning02:07
kiwichrishi seb02:07
sebhubchris: why is 'extern const char* bsp_boot_cmdline;' conditionally declared in confdefs.h?02:08
kiwichrisI will have a look.02:08
kiwichrisit is only conditional on confdefs.h providing an Init table02:09
kiwichrisor am I missing something ?02:09
sebhubyes, but this is the only place where it is declared02:10
sebhubi will add it also to bootcard.h02:10
kiwichrisyes you are correct. Hmmm I will need to think about this a little more.02:10
kiwichrisI think bootcard.h is a good place.02:11
kiwichrisI am fighting confdefs.h and dummy.o at the moment.02:11
kiwichrisI think there are problems with them.02:11
kiwichrisDid you manage to take a look at my email about the bdbuf code I sent ?02:12
sebhubi was on holiday half of this week02:12
kiwichrisNice.02:12
kiwichrisLucky you.02:13
kiwichrisI have completed coding and in testing. Seems to break down with a large disk.02:13
kiwichrisI have not tested the changing of buffer sizes but will.02:13
sebhubi will have a look at it now02:13
kiwichrisOk. I have implemented worker threads for the swap out task.02:14
kiwichrisI have no time to test this code. It also need a couple of async type output devices like SPI.02:14
kiwichrisThere is a design issue with worker threads and driver we need to think about.02:15
tuxmaniacI am a bit curious on some OT stuff. Hope its ok to ask here. sebhub and kiwichris are you both also part of OAR or ou do RTEMS work as part of some other firm which also uses RTEMS?02:26
sebhubi work for embedded brains GmbH in munich02:27
kiwichrisI am not part of OAR but do consult to them and other.02:27
tuxmaniacsebhub: aah that I know. So you work for clients who have RTEMS based systems right?02:28
sebhubyes02:29
kiwichristuxmaniac, it would be fair to say all contributors work for people with RTEMS based systems :)02:30
kiwichristuxmaniac, paid or not.02:30
tuxmaniackiwichris: ofcourse :)02:30
tuxmaniacnah I was just curious about your works. Because when I was working, I used to do Open source work only during the nights. My day job was different. It must be interesting to match your day job with your passion. hence asked02:32
tuxmaniachence I quit and became a student  :P02:32
*** madrazr has quit IRC02:34
kiwichrissebhub, have you played with new default Init I added ?02:48
sebhubwhat is new with the Init? :-)02:52
sebhubI did not notice a change02:52
kiwichrisIf you define CONFIGURE_RTEMS_INIT_TASKS_TABLE you do not need to define an Init, rather main will be called.02:52
kiwichrisThe Init is hidden away for you if you do not want it.02:53
sebhubah02:53
kiwichrisIt has a little issue which I would to figure out. I currently initialise the networking if RTEMS is configured with networking.02:54
*** madrazr has joined #rtems05:23
*** madrazr has quit IRC05:23
*** madrazr has joined #rtems05:40
*** madrazr has quit IRC06:11
*** madrazr has joined #rtems06:13
*** madrazr has quit IRC07:10
*** madrazr has joined #rtems07:11
*** DrJoel has joined #rtems07:14
*** ChanServ sets mode: +o DrJoel07:14
kiwichrisHi Joel07:17
*** cocan has joined #rtems07:17
*** chrisjohns has quit IRC07:18
DrJoelkiwichris: hi07:18
kiwichrisI have split the dummy.c file in 2.07:18
DrJoelgood.  I know it causes problems sometimes being one lump07:19
kiwichrisI have created dummy-networking.c07:19
sebhubhi07:19
DrJoelsebhub: hello07:19
kiwichrisMy new Init initialises networking and I think we may need something else here.07:20
kiwichrisThis is the Init that calls main for you if your app does not contain an Init.07:20
DrJoelok.. i am tinkering with standard helpers for the bsp cmdline you added07:20
kiwichrisNice.07:21
kiwichrisI have the cache code building but have a problem. My bdbuf tester is showing some problems.07:21
kiwichrisWill hut them down tomorrow. After that no need to pre-allocate buffer pools.07:22
kiwichrisWill hunt them down tomorrow. After that no need to pre-allocate buffer pools.07:22
* DrJoel was going to ask if one of you could try to figure out a test case to get to the heap allocate aligned case. It is now ~25% of the remaining instructions to cover. :(07:22
DrJoelanyone who volunteers will get an email. :-D07:22
DrJoelbrb07:22
kiwichrisWhat do you mean by " heap allocate aligned case" ?07:22
sebhubsry, i am very busy until the usb mass storage works07:22
sebhubbtw: what happend to the RTEMS_MALLOC_BOUNDARY_HELPERS?07:24
kiwichrisDrJoel, I did notice warnings in functions like _Protected_heap_Resize_block (pheapresizeblock.c). Why is intptr_t being used rather than uintptr_t ?07:24
DrJoelkiwichris: i don't know why it is signed.07:26
DrJoelsometimes it is a +/- thing for offsets but not likely there07:26
kiwichrisDrJoel, there is a bunch of these warnings.07:26
kiwichrisDrJoel, yeah with size of memory a -ve does not make sense.07:26
sebhubthe types for the task stack allocator should also match the heap types07:26
DrJoelkiwichris: we have 24 cases for 260 bytes (45 instructions) left to cover for the profile (score, rtems, sapi and posix dirs) we have been examining on the erc3207:27
*** cocan has quit IRC07:27
kiwichrisDrJoel, nice. Ah this is the heap traces you sent me.07:28
DrJoelthere is a 12 instruction (48 byte) sequence in _Heap_Allocate_aligned that I can' t figure out how to get to -- even going directly to the heap API and using a fresh heap07:28
sebhubinteresting07:28
kiwichrisWhat would break if you removed them ?07:29
DrJoelsantosh is working on a new test that will cover another 60 bytes. :-D07:29
DrJoelI have no idea.  I can't get my head around the logic there enough to know if it is impossible to reach or not07:29
DrJoelI emailed Sergei Organov (who submitted it a few years ago) a couple of weeks ago but no reply.  Email address may no longer be good .. who knows07:30
kiwichrisSure07:30
* kiwichris off for the night07:31
* kiwichris is away: I'm away but logging07:31
DrJoelsebhub: how's the usb going?07:32
sebhubi did not work with it the last two weeks07:32
sebhubi hope that i can start next week07:33
sebhubthe board initialization with boot from external flash was pretty nasty07:33
DrJoelok.. i hope you have a magic break through07:33
sebhubregarding the simple binary semaphores: are non recursive mutexes available via the classic API?07:34
DrJoelit's early.. what do you mean by non-recursive?07:36
sebhuba task blocks as soon as it tries to obtain the mutex more than once07:37
DrJoelthat's what a simple binary semaphore does07:37
DrJoel      the_mutex_attr.lock_nesting_behavior = CORE_MUTEX_NESTING_BLOCKS;07:38
DrJoel      the_mutex_attr.only_owner_release = false;07:38
DrJoelso the assumption is that we don't track the thread that locks the simple bin sem so it can be released by any other thread/ISR.  If unavailable, you block07:38
sebhubyes, but then priority inheritance is not available07:39
sebhuband i guess also priority ceilling07:40
DrJoelright.. you have to track the holder to have the priority inversion avoidance protocols.  This combination of attributes slipped through the error checks but wasn't there.  I recently improved the error checking07:40
sebhubok, so there is a difference to the FreeBSD mutex API here. this should not be a big problem07:42
DrJoelif you like, email me a link to a description of them and I will take the time to think when I am really awake07:42
* DrJoel needs to head to the office .. 07:42
DrJoelbye for now07:43
*** DrJoel has quit IRC07:43
*** cocan has joined #rtems07:43
*** cocan_ has joined #rtems08:05
*** cocan has quit IRC08:07
*** cocan_ is now known as cocan08:07
*** cocan_ has joined #rtems08:31
*** cocan has quit IRC08:32
*** cocan_ is now known as cocan08:32
*** DrJoel has joined #rtems08:57
*** ChanServ sets mode: +o DrJoel08:57
* DrJoel is now at the office.. as if anyone notices .. LOL09:01
sebhubi am nearly finished for today ;-)09:01
* tuxmaniac notices09:01
tuxmaniacDrJoel: I think PSIM also oes a pte search as loading a out of memory range (a address out of the one in linkcmds) seem to result in a unmapped address is trying to be accessed error09:02
tuxmaniacdoes*09:02
tuxmaniacloading into SDR1 register I mean09:03
tuxmaniactrying to check if its me or psim and have been doing it the whole day now :)09:03
*** madrazr has left #rtems09:08
DrJoeltuxmaniac: psim has some tracing options  which might help you and you have the source :-D09:09
*** cocan_ has joined #rtems09:09
*** cocan has quit IRC09:10
*** cocan_ is now known as cocan09:10
tuxmaniacDrJoel: yes. checking the source09:10
tuxmaniac:)09:10
*** cocan has quit IRC09:12
*** cocan has joined #rtems09:27
sebhubis there a special reason why static char ttyname_buf[sizeof (_PATH_DEV) + MAXNAMLEN] = _PATH_DEV; is initialized this way in ttyname.c?09:57
DrJoelthat's BSD code.  I have no idea.  On the head, it is actually disabled and we use the implementation in newlib.  It is the static buffer used by ttyname which wraps ttyname_r.  I would ask on the newlib list but suggest that ttyname.c be split into ttyname.c (w/static buf) and ttyname_r.c10:08
sebhubah ok, so the file in libcsupport is actually dead code on the head10:11
sebhubhm, I have /opt/rtems-4.10/arm-rtems4.10/lpc24xx_ncs_ram/lib/librtemscpu.a(libcsupport_a-ttyname.o) this in my linker map file10:12
sebhubso I don't use the newlib version?10:13
DrJoelit should have 0 .text.  It is compiled all the time but see HAVE_TTYNAME or something in there10:13
DrJoelbut the implementation is the same.  We just got rid of all cases in libcsupport where code havd been copied from newlib over the years10:14
sebhub./arm-rtems4.10/c/lpc24xx_ncs_ram/cpukit/config.h:/* #undef HAVE_TTYNAME */10:14
sebhubI have the ttyname stuff from libcsupport10:15
DrJoelhmm.. I dont10:17
DrJoelHmmm.. I have the same thing10:17
DrJoelAsk on the list.  I think we need to coordinate this via Ralf.  We can split the file in libcsupport as I suggested but newlib needs to split it also and we want to move to their implementation..  So 3 things and I wonder why it isn't turned on... so best to get his feedback.  He was trackign this and might have missed it.  Or there is a reason10:18
sebhubi asked some time ago an the newlib list and it is already fixed in newlib10:19
sebhubin the cpukit configure.ac is a FIXME comment around ttyname10:22
DrJoelLOL10:22
DrJoelask Ralf what's broken with ttyname configure.ac.  it is his FIXME.  looks like 1/3 of the job is done, we just need to split it in RTEMS and then properly handle turning it off10:23
sebhubin the cpukit doxygen documentation these handler modules belong all to the supercore?10:29
DrJoelhuh?  i don't understand/10:30
sebhubwe have a group (= module) hierarchy in the doxygen cpukit documentation with various group names that have a handler in them10:31
sebhubmaybe it is better to put all of them in a super group "Super Core" or so10:32
tuxmaniacshouldnt the psim trace be enabled when I put the --enable-sim-trace in the rtems configure command itself?10:32
tuxmaniac[tuxmaniac@yoda psim]$ ../rtems/configure --target=powerpc-rtems4.10 --disable-multiprocessing --enable-cxx --disable-rdbg --enable-maintainer-mode --enable-tests --enable-networking --enable-posix --disable-itron --disable-deprecated --disable-ada --enable-expada --enable-rtemsbsp=psim CLOCK_DRIVER_USE_FAST_IDLE=010:33
tuxmaniac--enable-sim-trace10:33
DrJoelEach group should indicate a logical handler/object/package .. like chain, thread, thread queue.  Not the entire subsystem.  If there is a way to collect them together into a higher level that would be good.  Then you would have a subsystem called Classic, POSIX, SuperCore and individual objects/groups within there10:34
DrJoeltuxmaniac: no.. --enable-sim-trace is a gdb build option10:34
sebhubyes, that is how i have done it with the Classic in the latest patch10:35
DrJoellike nested groups?10:35
*** switnick has joined #rtems10:36
sebhubgoups may contain groups10:36
sebhuband one group may be contained in different groups10:36
DrJoelthen we should have a SuperCore Group and a Classic API group, etc10:37
DrJoelwith subgroups under them10:37
sebhubyes, see my latest patch10:37
DrJoelcool .. neatly reflects intended organization10:37
DrJoelit is in the queue.. :)10:37
sebhub*g*10:37
sebhubi want to structure it a little bit before the trainee gets hand on it10:38
DrJoelI try to convert a .h to Doxygen everytime I edit one now.10:38
sebhubthat good, but we should agree on a standard Doxygen comment style10:39
DrJoelagreed.10:40
DrJoelwhen i do this i am not rewriting existing comments, just reformatting10:40
sebhubI personally don't like the @param directives, I think the in text parameter description like in man pages, Java API and Qt is better10:40
DrJoelit's a lot of monkey work to change them10:41
tuxmaniacDrJoel: so i rebuild gdb 4.16 with this option to verify whats really happening inside psim?10:42
sebhubyes, thats it10:42
tuxmaniacsebhub: is that for my question?10:42
sebhubno its for the monkey work ;-)10:43
tuxmaniacok :)10:43
DrJoeltuxmaniac: why 4.16?  gdb is at 6.8?  you want a 10+ year old gdb LOL10:43
DrJoelit looks like the --enable-sim-trace option is turned on when the RPMs are built.  It is up to you to hack on the device tree built by "psim" or psim-gdb to enable it10:44
tuxmaniacaah ok10:45
DrJoelyou should be able to add a -t option to the "tar sim" command argumetns in the gdb script and add the trace options.  Do a powerpc-rtems4.10-run --help and see the arguments10:47
switnickDrJoel: Hello10:54
DrJoelswitnick: hi10:59
switnickDrJoel: Just an update, talking to Eric11:01
DrJoeli saw that it probably works better on windows11:01
switnickyea11:01
switnickis there some way to compile in windows?11:01
DrJoelyes but easier if you just use a virtual machine and Linux sometimes and Wndows sometimes .. share a folder and put exe in it.  You would have to rebuild tools to windows host and that creates its own set of issues.  Better to just use windows for debugger11:03
switnickok11:04
DrJoelif you need a VMWare image, we have one available for torrent.  I also have a Fedora 10 VirtualBox image but I don't know if it is finished yet.11:05
switnickok11:10
switnickI think I should be ok11:10
switnickbut thanks11:12
*** cocan has quit IRC11:14
*** sebhub has quit IRC11:47
*** madrazr has joined #rtems11:58
*** switnick has quit IRC12:02
*** roxanal has quit IRC12:07
*** roxanal has joined #rtems12:26
*** switnick has joined #rtems12:38
tuxmaniacDrJoel: changed the device tree ram settings and it seems it works. But still it doesnt answer my doubt of hardware pte search13:35
tuxmaniactried changing the model that is simulated to be 603 which does not do hardware tlb search itself.13:35
tuxmaniacstill the same issue13:35
tuxmaniaclooking into it now13:35
DrJoelsorry.. sometimes simulators are good.. other times they require work.13:36
DrJoelqemu has a powerpc simulation but we don't have a BSP for it.  But it might make a good reference13:36
agrierapparently mozilla has stacks of nokia phones to test fennec on13:38
agrierthe simulator nokia provided them is buggy in different ways than the actual hardware13:39
DrJoelcool.. wish we had a big sugar daddy ;)13:39
agrierback in January they were looking for symbian experience.  now it's shifted to linux...13:40
tuxmaniacagrier: yes. I remember the mozilla folks displaying some cool nokia phones at fosdem last year. some that are just prototypes with all black plastic around etc13:40
tuxmaniac:)13:40
agrierI have images in my mind of cellphones hot-glued to rackmount shelves13:41
*** switnick has quit IRC13:44
*** switnick has joined #rtems13:44
switnickDrJoel: Hello sorry to bother you again13:52
DrJoelswitnick:  no problem13:53
switnickI think I need to add some flags for the asm compilation13:53
switnicknot sure where to add them13:54
switnickASMFLAGS = $(CFLAGS) -Wa,-gstabs13:54
switnicksomething like this13:54
tuxmaniacany idea what generates the psim_tree.  I ahve modified the linkcmds as well as the c/src/lib/libbsp/powerpc/psim/startup/device_tree to increase the oea memory size. still the psim_tree.tuxmaniac that gets generated has only 8 MB as oea memory13:55
tuxmaniacthis file gets generated when I run the tests/psim-gdb13:55
tuxmaniacimodified the linkcmds and device_tree to have 64 MB13:56
tuxmaniacrebuilt it from scratch13:56
DrJoeltuxmaniac: are you using the sim-scripts from the gcc-testing cvs module?13:58
DrJoelswitnick: you should be able to add those to CPU_CFLAGS in make/custom/avrtest.cfg... just add the Wa,...part13:59
switnickok13:59
switnick will try that thanks13:59
tuxmaniacno. the ones inside the psim directory that gets populated in make14:00
DrJoeltuxmaniac: the scripts in that cvs module are newer and smarter .. derived from what you are looking at but better14:01
DrJoelif you check that out, see psim.in and *gdb*in .. everything is magically put together to share the infrastructure so all simulators are run by the same commands14:01
tuxmaniacyeah looking into those now14:01
DrJoelin the new scripts, there is a method called something like bspGenerateDeviceTree and "GetSimArgs" which should be what you are looking for14:02
tuxmaniacthank you DrJoel14:03
DrJoelbspGenerateGDBCommands in psim.in14:03
DrJoelthese scripts are (IMHO as author) very neat.  ABout 20 simulators all invoked with the same arguments.  THe script is named BSP and it makes building test infrastructre on top of it a LOT easier since orthogonality is the name of the game14:04
DrJoelpatches are gladly accepted.14:04
DrJoelthe next step is to intewgrate the RTEMS GDB macros into the gdb side so they are always included for you14:04
tuxmaniacnice14:04
tuxmaniacDrJoel: :D fixed. *phew*14:19
DrJoeltuxmaniac: :D14:19
DrJoelIs there anything to add to the psim scripts14:19
tuxmaniacDrJoel: i dont think so. it worked. But I have another conceptual issue with psim.14:20
tuxmaniacgrrr.14:20
tuxmaniacpsim keeps calling the exception handler without completely executing the funcntion.14:21
tuxmaniaci mean the handler14:21
tuxmaniacso all it does is repeated calling of the exception handler and whenever it tries to execute a few statement, it again calls the exception handler and this goes into a vicious loop14:21
tuxmaniacI am trying to fix that now14:22
tuxmaniac:)14:22
roxanalDrJoel: the documentation is almost ready. I am waiting to hear back from Aaron if he wants me to change some things before submitting it to you15:02
DrJoelroxanal: :-D15:03
*** lee21 has quit IRC15:21
* agrier reads docs15:21
agrierJoel, is there any guideline for moving code into libchip?  I am thinking of vgainit.c, specifically15:23
roxanalI was just about to ask that also15:24
DrJoel(1) the code has to deal with a peripheral that is not unique to a single board15:24
DrJoel(2) it should provide a way to configure whatever board specific hooks are needed to touch hardware15:24
DrJoelthat's it.  I think vga hits (1).  What about (2)?  Does it use IO instructions that might be memory mapped on another architecture?15:25
roxanalThe IO (ioctls) structures and actions are not defined in vgainit.c15:28
tuxmaniachaha. looks like its printk which is the culprit15:29
tuxmaniacprintk seems to cause a protected memory access which my MMU doesnt like15:29
DrJoelroxanal: vga_init has out_word which is PC specific.  That would have to be configured in a way that register accesses on another board would be handled properly15:30
DrJoeland writeregs..15:30
DrJoeldoes that make sense?15:31
roxanalyes15:31
DrJoeltuxmaniac: so are you passing bad data to printk?15:31
tuxmaniaclooks like bt I dont know what bad data is . All I do is printk("DSI Exception hit") and this blocked any furhter execution of the exception handler15:33
tuxmaniacand MMU caused another exception when this statement was executed15:33
tuxmaniacand the same happened down the function with another printk15:33
tuxmaniacso now removing both and seeing whether life becomes much better15:33
*** cocan has joined #rtems15:48
DrJoelcocan: how are things going on your projects?16:01
switnickDrJoel: I think the reason I was getting the error was because of the wrong option flag16:02
switnickseems -g was correct16:02
switnickwith the debug format to follow16:02
cocanDrJoel: great...I am working now on the decoding..had to go back and change some things on the rtems_trace tool..should finish the decoding part this week16:03
DrJoelswitnick: good16:03
switnickDrJoel: the reason for all this is i'm having some problems with avr studio16:04
DrJoelcocan: wonderful.. be sure to start a wiki page on how to do this16:04
DrJoelswitnick: well eric w is the man for that.  :)16:04
switnickyea16:04
switnickwas wondering...16:04
switnickcan a generate an asm file?16:04
cocanDrJoel: ok...will do that16:05
switnickthat I can assemble with avr studio?16:05
DrJoelsure .. if nothing else look in the logs for a build of a single .S file.  Then cut and paste that to a script file.. use -E instead of -o XXX.o and then save the preprocessed output.16:08
DrJoelThat will at least give you a preprocessed .s file to use as input to their assembler16:09
DrJoelnight all .. heading home16:30
*** DrJoel has quit IRC16:30
*** roxanal has quit IRC16:54
*** switnick has quit IRC16:55
*** madrazr has left #rtems17:46
*** madrazr has joined #rtems17:47
*** cocan has quit IRC17:55
*** roxanal has joined #rtems18:54
*** chrisjohns has joined #rtems20:32
*** chrisjohns has quit IRC23:42
* kiwichris is back (gone 17:07:11)00:38
* kiwichris is away: I'm away but logging01:19
*** sebhub has joined #rtems01:54
*** madrazr has joined #rtems02:12
* kiwichris is back (gone 01:11:36)02:30
kiwichrissebhub, hi02:30
sebhubgood morning02:31
kiwichrisI am testing a new version of the cache code. Chasing a bug I have introduced. I need someone to test the swapout worker thread change and I do not really a real set up that can use it.02:32
kiwichris I do not have a  real set up that can use it.02:33
kiwichrisYou said you have multidisk set ups.02:33
sebhubi said that the usb may introduce a multi disk setup02:33
sebhubi hope i can start with the mass storage driver next week02:34
kiwichrisAh ok. Any SPI based systems with more than one disk02:34
kiwichrisUSB would be a really good test.02:34
sebhubmy usb deadline is end of august02:35
kiwichrisI will finish my tested and the code can sit in the cache until you say more than 0 worker threads.02:35
kiwichrisThe design issue worker threads brings, is getting more than one thread into a driver at once.02:36
kiwichrisShould we assume drivers have protection for this, ie a mutex ?02:36
kiwichrisor queue ?02:36
sebhubhm, good question02:37
kiwichrisThis happens if more than one discontinuous blocks exists in the cache for the same device02:38
kiwichrisThe other solution is to only allow a single transaction to a single device at a time. This load shares the threads better02:39
sebhubis it not possible to assosiacte the worker threads with the devices02:39
sebhubassociate02:40
kiwichrisYou could have that model and I suspect the changes to implement are not big, but you may only have 2 device active at once with 4 devices in a system.02:40
kiwichrisI will not work on this any more so we have some time to ponder the issue.02:41
kiwichrisWould you be able to chat to Thomas about it and if you come to a preferred solution let me know via email or IRC ?02:41
sebhubin FreeBSD the device driver specifies how many transactions it can handle concurrently02:42
sebhubok02:42
*** madrazr has left #rtems02:42
kiwichrisHmm, ok. I suspect that is a DOS type function.02:42
kiwichrisor a load share.02:43
kiwichrisWhat this does mean is device can handle more than one thread.02:43
kiwichrisFor an RTOS we need to think if making the drivers more complex is worth it.02:43
kiwichrisI suppose they are already thinking about it more. On the read path we have as many threads in the driver as there are readers.02:44
kiwichrisThe ATA driver has that silly queue and malloc stuff to handle this.02:44
sebhubyes, if we make it to sophisticated we could have used CAM from FreeBSD02:44
kiwichrisCAM ?02:44
sebhubCommand Access Method, the storage subsystem in FreeBSD02:45
sebhubone of the biggest subsystems02:45
kiwichrisI will look it up.02:45
kiwichrisis it Common Access Method  or Command Access Method ?02:46
sebhuboh its Common, sry02:47
kiwichrisno prob02:47
sebhubfor the usb mass storage driver i need to implement parts of the interface to access the standard umass driver02:48
kiwichrisIf I need a file system that needs that sort of complication I would use FreeBSD, Solaris, Linux etc.02:48
kiwichrisI am sure you do as the SCSI command set is almost a standard these days.02:49
sebhubyes, we should keep it simple02:49
kiwichrisDoes it include the command queuing and scheduling logic to get the best out of the buses ?02:49
sebhubthe FreeBSD CAM yes, but my umass interface no02:50
kiwichrisAnd that I think is fair.02:50
kiwichrisThis is where the system architect needs to think about the cross over between realtime and file system importance.02:50
sebhubyes, real time and average performance are two different goals02:52
kiwichrisWith my cache changes I have removed all pools from the cache. There is now  just a cache and no pools. You specify the min and max size of buffers.02:52
kiwichrisThe benefit is a single cache is the static struct is global and all accesses are direct so I think the compiler and more to work with to get better code.02:53
sebhubare the buffers a whole multiple of the min size?02:53
kiwichrisYes.02:53
kiwichriseg 512 and 409602:54
sebhubok02:54
kiwichrisand allocation is a multiple of 202:54
kiwichrisbrb02:54
kiwichrisfor 512 and 4096 you have 512, 1024, 2048 and 4096. You can say any size you like but the allocation is the above list.02:55
kiwichrisbrb02:55
*** rokka has quit IRC03:15
*** rokka has joined #rtems03:17
*** madrazr has joined #rtems06:12
*** dr__house has joined #rtems06:16
*** cocan has joined #rtems06:40
*** bazinski has quit IRC08:01
*** bazinski has joined #rtems09:47
*** dr__house has quit IRC09:50
*** madrazr has quit IRC09:57
*** switnick has joined #rtems11:09
*** sebhub has quit IRC11:27
*** madrazr has joined #rtems11:28
*** dr__house has joined #rtems11:29
*** bazinski has quit IRC11:41
*** switnick has quit IRC12:39
*** roxanal has quit IRC12:51
*** cocan has quit IRC14:00
*** cocan has joined #rtems14:15
*** switnick has joined #rtems14:18
*** dr__house has quit IRC14:28
*** switnick has quit IRC15:57
*** switnick has joined #rtems16:46
*** switnick has quit IRC17:00
*** cocan has quit IRC17:03
*** madrazr has quit IRC17:35
*** cocan has joined #rtems17:41
*** cocan has quit IRC18:10
*** cocan has joined #rtems18:11
*** cocan has quit IRC18:56
*** switnick has joined #rtems18:59
*** DrJoel has joined #rtems20:01
*** ChanServ sets mode: +o DrJoel20:01
*** switnick has quit IRC22:01
* kiwichris is away: I'm away but logging01:08
*** madrazr has joined #rtems02:24
*** madrazr1 has joined #rtems02:24
*** madrazr has quit IRC02:24
*** madrazr has joined #rtems02:25
*** madrazr1 has quit IRC02:25
*** madrazr has quit IRC05:00
*** madrazr has joined #rtems05:06
*** madrazr has quit IRC06:02
*** madrazr has joined #rtems07:59
* DrJoel is at home.. still drippy nose and stuffed up head .. logging .. walking by computer when I am up08:52
*** switnick has joined #rtems09:10
*** madrazr1 has joined #rtems09:12
*** madrazr has quit IRC09:12
*** dr__house has joined #rtems09:13
switnickDrJoel: Hello, I have encountered some difficuties09:16
DrJoelswitnick: what kind?09:16
switnickI have been talkin with Eric09:16
switnickhe is saying to use the avr studio debugger we will need to build on windows09:17
DrJoeli followed some of it.09:17
DrJoelwhy doesn't the debugger in their understand the debug information in our executables?09:17
DrJoelthey are built from the same sources09:17
switnickit does but it needs the source09:17
switnickand wont debug without it09:17
DrJoelthen copy the source so it can find it09:18
switnickhmm09:18
switnickvery smart09:18
switnickhadn't thought of that09:18
DrJoelIf it lets you get far enough, you can use the gdb "directory" command to fix his search path09:18
switnickok I will look into that09:18
DrJoelI have been through this before.  Lots of tools are windows specific and require weird hw setup .. so build where you want and move what you have to.  You can even publish your source directory using Samba and mount it using Windows so you aren't copying it.09:19
DrJoelJust access it across the network09:19
DrJoelplay with that.. i am still stopped up.. dripping.. sneezing.. but will walk by the computer09:20
DrJoelneed to take a steamy shower and hope it helps this morning09:20
*** dr__house has quit IRC09:21
DrJoeldidn't even see dr__house here.. :(09:21
switnickOk09:22
switnickI will try with that09:22
switnickthanks09:22
*** madrazr has joined #rtems09:34
*** madrazr1 has quit IRC09:34
* DrJoel is back.. steamed from shower .. still not feeling great though. :(10:10
*** switnick_ has joined #rtems10:17
*** switnick has quit IRC10:19
*** switnick has joined #rtems10:19
*** switnick_ has quit IRC10:35
*** madrazr has quit IRC12:03
*** madrazr1 has joined #rtems12:03
*** switnick has quit IRC12:14
*** switnick has joined #rtems12:15
*** switnick has quit IRC12:23
*** madrazr has joined #rtems12:46
*** madrazr1 has quit IRC12:46
*** dr__house has joined #rtems12:51
dr__houseDrJoel: Good afternoon12:55
* dr__house had a bad headache, but I feel slightly better now.12:57
DrJoeldr__house: hello.. how's the testing going?13:54
dr__houseDrJoel: kind of slow today :(13:55
DrJoelnot feeling well or unlucky on the tests?13:55
dr__houseDrJoel: not feeling well13:55
DrJoeli'm not either.. messed up a couple of attempts to hit a path through heapallocatealigned.c myself13:56
dr__houseDrJoel: hmm.. get well soon to both of us :)13:57
*** switnick has joined #rtems13:58
*** madrazr has quit IRC15:14
*** madrazr1 has joined #rtems15:14
*** dr__house has quit IRC16:11
*** madrazr1 has quit IRC17:55
*** switnick has quit IRC18:22
*** madrazr has joined #rtems01:01
*** madrazr1 has joined #rtems03:02
*** madrazr has quit IRC03:03
*** madrazr has joined #rtems03:44
*** madrazr1 has quit IRC03:44
*** madrazr has left #rtems06:14
*** madrazr has joined #rtems08:11
*** lee21 has joined #rtems08:38
*** dr__house has joined #rtems09:43
dr__houseDrJoel: good morning :)09:43
DrJoeldr__house: good morning!  is your headache better after a good night's sleep?09:44
dr__houseDrJoel: yep09:44
DrJoeli'm still stopped up but bought some new medicine which seemed to dry me up for a while.  Need to be passable for a funeral today. :(09:44
dr__houseDrJoel: oh ok. :( About that case, I don't think I am hitting the required block of code. The control isn't entering _POSIX_Message_queue_Create_support subroutine09:45
DrJoelhmmm.. send me your patch and let me take a look before I hop in the shower09:46
* DrJoel is worried that the big heap case showed up in a run yesterday. I am wondering if it varies based upon the initial alignment of the heap. :(09:48
DrJoelheap memory I mean09:48
dr__houseDrJoel: Sent the patch09:59
DrJoeldr__house: is the code at mqueuecreatesupp.c:101 impossible to trigger because of the attempt to see if the message queue already exists and we are just attaching to it in mqueueopen.c?  Just a thought.. let me look at the patch and think09:59
dr__housethats what I too thought09:59
dr__houseDrJoel: ^^09:59
dr__houseDrJoel: _POSIX_Message_queue_Create_support() is called only after all the checks are done and hence I think that piece of code is never triggered10:00
DrJoelwe allocate an fd in open at line 73 to see if it exists.10:00
DrJoelagreed.  We need to mark that check as RTEMS_DEBUG.10:01
DrJoelWrap it in a "#if defined(RTEMS_DEBUG)" with an explanation comment about why it can't happen.10:01
DrJoelTo me.. these are some of the hardest cases :)10:01
DrJoelthen focus on the memory allocation errors using the same testing pattern used in the other tests I showed you (sp18 and psxkey03?)10:03
dr__houseDrJoel: ok, cool.10:04
DrJoelfwiw i went after the psignal.c case.  turned out it was unreachable and i had to restructure the code.  <sigh> that turned one 12 byte thing into two 4 byte so I am adding a test10:04
dr__houseDrJoel: oh ok10:04
DrJoeldr__house: the memory allocation one should be fairly straight forward using those tests as templates .. actually if you base it on sp18, it will be easy.. that is using stack size.. use the number going down as msg size and reserve one of that size... keep going until it allocates it successfully and place a large (~64 byte) name on the queue. :)  Then break out of the loop, delete it and the test checks that the heap is still ok10:06
* DrJoel needs to take a shower for the steam :)10:06
DrJoelbrb10:06
dr__housesure10:06
dr__houseDrJoel: there's mistake in that patch, it has to be assert(errno == ENFILE) instead of assert(sec_Queue==ENFILE) :P10:11
* dr__house is away: afk10:40
DrJoelyeah.. i spotted that .. corrected in email to you.  the "rtems_set_errno_and_return_minus_one" is a very verbose hint at how a routine returns. :)10:48
*** lee21 has quit IRC11:09
*** lee21 has joined #rtems11:17
* dr__house is back.11:49
dr__houseDrJoel: heh, yeah it is :)11:49
*** lee21 has left #rtems12:11
*** madrazr has quit IRC15:24
* dr__house is leaving for the day. Good day/night/evening to everyone.16:59
*** dr__house has quit IRC16:59
*** chrisjohns has joined #rtems18:34
*** chrisjohns has quit IRC23:44

Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!