| 03:06.19 | *** join/#asterisk-dev kpfleming (kpfleming@asterisk/digium-director-of-software-tech/kpfleming) |
| 03:06.20 | *** mode/#asterisk-dev [+o kpfleming] by ChanServ |
| 03:29.38 | *** join/#asterisk-dev mjordan (mjordan@conference/astricon/x-cfwewbcosjssbzjm) |
| 03:29.38 | *** mode/#asterisk-dev [+o mjordan] by ChanServ |
| 03:40.03 | *** join/#asterisk-dev moy (~moy@173.239.155.74) |
| 03:40.34 | *** join/#asterisk-dev kpfleming (kpfleming@asterisk/digium-director-of-software-tech/kpfleming) |
| 03:40.36 | *** mode/#asterisk-dev [+o kpfleming] by ChanServ |
| 03:50.10 | *** join/#asterisk-dev kpfleming (kpfleming@asterisk/digium-director-of-software-tech/kpfleming) |
| 03:50.10 | *** mode/#asterisk-dev [+o kpfleming] by ChanServ |
| 03:56.17 | *** join/#asterisk-dev irroot (~irroot@197.171.133.9) |
| 04:14.23 | *** join/#asterisk-dev ZX_81 (ZX_81@conference/astricon/x-dqhxcrzikghotpts) |
| 04:37.27 | *** join/#asterisk-dev moy (~moy@173.239.155.74) |
| 04:41.35 | *** join/#asterisk-dev bklang (~bklang@12.41.171.130) |
| 04:53.58 | *** join/#asterisk-dev zerohalo (~zerohalo@74.60.136.128) |
| 05:12.50 | *** join/#asterisk-dev bklang (~bklang@12.41.171.130) |
| 05:21.48 | *** join/#asterisk-dev bklang (~bklang@12.41.171.130) |
| 15:49.57 | *** join/#asterisk-dev infobot (~infobot@rikers.org) |
| 15:49.57 | *** topic/#asterisk-dev is Asterisk Development Discussion -=- http://www.asterisk.org/developers -=- Tier 2 and 3.14159265 support is in #asterisk -=- Astricon 2011 in Denver, Colorado: AstriDevCon on October 24. Be there! |
| 15:53.37 | *** join/#asterisk-dev infobot (~infobot@rikers.org) |
| 15:53.37 | *** topic/#asterisk-dev is Asterisk Development Discussion -=- http://www.asterisk.org/developers -=- Tier 2 and 3.14159265 support is in #asterisk -=- Astricon 2011 in Denver, Colorado: AstriDevCon on October 24. Be there! |
| 16:28.32 | *** join/#asterisk-dev kpfleming (kpfleming@asterisk/digium-director-of-software-tech/kpfleming) |
| 16:28.32 | *** mode/#asterisk-dev [+o kpfleming] by ChanServ |
| 16:42.28 | *** join/#asterisk-dev jrad (~jrad@173-11-144-149-houston.txt.hfc.comcastbusiness.net) |
| 16:43.36 | *** join/#asterisk-dev otherwiseguy (~otherwise@69.76.221.79) |
| 16:56.36 | *** join/#asterisk-dev irroot (~irroot@197.170.100.128) |
| 17:00.36 | BlackBishop | kpfleming: https://issues.asterisk.org/view.php?id=18703 |
| 17:00.42 | BlackBishop | this issue is since january ! :| |
| 17:04.38 | BlackBishop | am compilit acum -r1 |
| 17:07.01 | irroot | i @ home just hit a deadlock on 10 restart .... should i log it or fix it :P |
| 17:19.14 | irroot | ok not a DL |
| 17:19.20 | irroot | but ill look into it |
| 17:21.20 | *** join/#asterisk-dev anonymouz666 (~anonymouz@187-28-37-118.poolip.RJO.embratel.net.br) |
| 17:41.02 | wdoekes2 | BlackBishop: that should always be aligned, even on x86.. the few bytes it costs do not outweigh the extra #ifdef's IMHO |
| 17:41.13 | *** join/#asterisk-dev mjordan (mjordan@conference/astricon/x-igqctzpccfqzbscc) |
| 17:41.13 | *** mode/#asterisk-dev [+o mjordan] by ChanServ |
| 17:42.02 | wdoekes2 | (esp. considering that unaligned access is slower) |
| 17:49.24 | jsmith | I would tend to agree with that statement... why not just do it universally? |
| 17:49.50 | jsmith | (ignoring possible ABI breakage, etc. of course) |
| 17:52.51 | BlackBishop | beats me. :( |
| 17:52.52 | wdoekes2 | (the sizeof(x) and then +=1 is a mismatch too.. that should either be %2 or +=(sizeof(x)-(pos%sizeof(x))) |
| 17:53.03 | BlackBishop | I'm just sad it doesn't work for me by default .. |
| 17:53.14 | BlackBishop | https://issues.asterisk.org/jira/browse/ASTERISK-17310 |
| 17:53.32 | BlackBishop | feel free to improve and solve and commit patches to the tree so gentoo can get it in their repo faster |
| 17:53.59 | wdoekes2 | does everything work with the patch? |
| 17:54.18 | BlackBishop | well, testing now |
| 17:54.26 | BlackBishop | include/asterisk/compat.h .. added the same thing here too |
| 17:54.45 | BlackBishop | it's compiling now .. it'll take a while |
| 17:55.13 | jsmith | I'm in meetings at the moment, but I'd be happy to try on ARM later tonight or early tomorrow |
| 17:56.27 | BlackBishop | jsmith: I'm trying it now ! :) on a sheevaplug |
| 17:57.02 | wdoekes2 | BlackBishop: er.. did you add the || __arm__ to the endian check? is it big-endian? (they can be both, I think, and commonly little) |
| 17:58.06 | BlackBishop | yup |
| 17:59.01 | BlackBishop | altough __sparc__ is used in include/asterisk/unaligned.h and include/solaris-compat/compat.h .. |
| 17:59.05 | wdoekes2 | (oh.. that's probably only ran if the endian.h headers are broken.. I think you're safe either way) |
| 17:59.41 | BlackBishop | I only plaied with main/utils.c and include/asterisk/compat.h |
| 18:01.28 | wdoekes2 | mm.. that unaligned.h could bite you too, I see |
| 18:02.09 | BlackBishop | I'm not much of a C devel, so if anybody has patches for all 4 files, I'd be glad to test them out ! |
| 18:02.20 | BlackBishop | ( and if needed any other files ) |
| 18:03.16 | BlackBishop | #elif (defined(SOLARIS) && defined(__sparc__)) || defined(__arm__) |
| 18:03.19 | BlackBishop | is that better ? |
| 18:03.33 | BlackBishop | ( for unaligned.h ) |
| 18:04.45 | wdoekes2 | that's safer.. the worst it can do is slow things down by a fraction |
| 18:05.05 | BlackBishop | ok :) |
| 18:05.27 | BlackBishop | will wait and see how this goes... |
| 18:07.34 | BlackBishop | not to mention that from time to time .. |
| 18:07.38 | BlackBishop | the console got all weird on me |
| 18:07.47 | BlackBishop | this should fix that problem too, right ? |
| 18:13.51 | BlackBishop | Error: unrecognized option -march=armv5tel |
| 18:13.54 | BlackBishop | :| WTH ? |
| 18:15.36 | BlackBishop | I didn't specify this nowhere ! |
| 18:16.08 | wdoekes2 | Makefile, line 225? |
| 18:17.38 | BlackBishop | I guess that's the one |
| 18:17.48 | BlackBishop | lemme see if the gentoo patches this file in any way |
| 18:18.02 | tzafrir_laptop | BlackBishop, what are you trying to do? |
| 18:18.25 | BlackBishop | compile asterisk from sources instead of gentoo's repo to test a patch and get my asterisk working again ! |
| 18:19.02 | BlackBishop | wdoekes2: http://distfiles.gentoo.org/distfiles/gentoo-asterisk-patchset-1.5.tar.bz2 seems it does |
| 18:19.42 | tzafrir_laptop | BlackBishop, interesting. I'll need to try this on my Guruplug at home |
| 18:20.06 | BlackBishop | tzafrir_laptop: https://issues.asterisk.org/jira/browse/ASTERISK-17310 |
| 18:20.07 | tzafrir_laptop | IIRC qemu behaves differently with respect to alignment issues |
| 18:20.12 | BlackBishop | I think this is the problem... |
| 18:20.25 | tzafrir_laptop | What version do you try to use? |
| 18:20.26 | BlackBishop | tzafrir_laptop: if you want .. to test stuff, I'm doing it now :) |
| 18:20.31 | BlackBishop | 1.8.7.1 |
| 18:20.38 | BlackBishop | latest .. ( the only one in gentoo's repo ) |
| 18:23.11 | BlackBishop | ok, applied all 14 patches in gentoo's repo, trying to compile again |
| 18:40.54 | BlackBishop | seems I got over that problem .. ( continued compiling .. ) |
| 18:41.03 | BlackBishop | will wait for it to finish and test if it works :) |
| 18:45.23 | BlackBishop | any way I could force the compiler to align stuff without using those patches ? |
| 18:45.26 | BlackBishop | :| |
| 18:48.41 | Corydon76-home | Uh, no, you have to tell the compiler to align it in very specific ways. |
| 18:49.12 | BlackBishop | ... explain please :) |
| 18:49.22 | Corydon76-home | Kind of like asking whether you can drive a car without putting gasoline in the tank. |
| 18:49.26 | BlackBishop | and how would that affect my system if I add those to my CFLAGS .. |
| 18:49.53 | BlackBishop | ok, well, I guess that could be done, right ? |
| 18:50.01 | BlackBishop | like with some -faligned sort of stuff |
| 18:51.17 | Corydon76-home | No. |
| 18:52.44 | Corydon76-home | You haven't the foggiest idea what you're carping about. Either learn to program, or apply the patches. This conversation isn't going anywhere. |
| 18:54.10 | BlackBishop | I do have _BASIC_ ideas about programming, I just don't get the alignment part ! I'd be glad to hear explanation from someone who knows more ! |
| 18:54.40 | BlackBishop | ( I did do some php but I doubt that counts .. and some _BASIC_ C a couple of years ago ) |
| 18:55.38 | BlackBishop | anyhow, the idea behind my question was to find a way to use gentoo's package instead of compiling my own .. because I don't know when the patches will get into asterisk or gentoo . |
| 18:56.33 | Corydon76-home | Some processors don't like accessing data that isn't an even multiple of the size of that piece of data (i.e. a 4-byte datum at an address evenly divisible by 4, 8-byte by 8, etc.) |
| 18:56.59 | Corydon76-home | So alignment is about ensuring that datums are aligned with those needed boundaries |
| 18:57.37 | BlackBishop | interesting, wonder why wouldn't they like that :/ but I'll google more, don't want to disturb you ! |
| 18:58.05 | Corydon76-home | It's a limitation of the processor platform. |
| 18:58.26 | Corydon76-home | They do it because it's faster and less complicated. |
| 18:58.38 | BlackBishop | ow |
| 18:59.52 | tzafrir_laptop | (Or rather, it's x86 being not picky at all) |
| 19:00.44 | BlackBishop | 1 hour now .. compiling .. this'd better work ! |
| 19:01.20 | tzanger | BlackBishop: basically unaligned accesses add silicon since the access has to be broken up into multiple aligned accesses, and the data reassembled |
| 19:03.40 | Corydon76-home | tzanger: I think that's how you'd do it on ARM, not that processors necessarily have to do it that way. |
| 19:04.04 | BlackBishop | yeah .. but as Corydon76-home said .. in my mind ( not having enough knowledge ) all is just a stream :) same code .. and can't realize how deep those implications are at hardware level .. |
| 19:04.09 | BlackBishop | but I kinda' get it :) |
| 19:04.10 | tzanger | Corydon76-home: well unaligned accesses almost exclusively end up being broken into individual aligned accesses |
| 19:04.38 | Corydon76-home | tzanger: on x86, too? |
| 19:04.57 | *** join/#asterisk-dev irroot (~irroot@197.110.219.70) |
| 19:05.03 | tzanger | Corydon76-home: I haven't done much low-level x86... not since the 80386 |
| 19:05.32 | tzanger | but if you want a 32-bit value that's at an address that does not have the low 2 address bits '0'.. how else would you do it? |
| 19:05.45 | tzanger | youd' have to do 2 reads or writes |
| 19:06.02 | tzanger | mask off the appropriate bytes |
| 19:06.05 | tzanger | etc |
| 19:07.50 | Corydon76-home | Admittedly, my asm skills are rusty, but I've never had to deal with any of those alignment issues until dealing with SPARC, ARM, and other processors. I don't know if microcode hid the issue from me or not. |
| 19:08.13 | BlackBishop | done compiling, /me crosses fingers |
| 19:09.18 | tzanger | well what I mean is that ARM et al just do not "fix" this themselves. they throw an exception |
| 19:09.35 | tzanger | x86 will fulfill the request by creating multiple accesses |
| 19:09.56 | tzanger | moment I am going to see if I can test it |
| 19:10.52 | BlackBishop | YUP, IT WORKS ! |
| 19:10.54 | BlackBishop | THIS WAS IT ! |
| 19:14.07 | tzanger | http://pastebin.com/ia3uDAYD |
| 19:14.12 | tzanger | gotta check the asm out though |
| 19:15.10 | *** join/#asterisk-dev mjordan (mjordan@conference/astricon/x-odsvbicyxsnbvswn) |
| 19:15.10 | *** mode/#asterisk-dev [+o mjordan] by ChanServ |
| 19:16.08 | BlackBishop | tzanger: need ssh access to the box to test stuff ? |
| 19:17.59 | tzanger | no I am testing on x86-64 |
| 19:18.17 | tzanger | the assembly output looks like it just does an unaligned access without blinking |
| 19:18.45 | tzanger | %ebx has the unaligned address, it pushes it, then lea's %ebx with 4 for the next int and away it goes |
| 19:19.38 | tzanger | I haven't got ARM super-handy right now |
| 19:19.41 | tzanger | actually that's a lie |
| 19:19.44 | tzanger | moment |
| 19:20.20 | BlackBishop | well, I do have an arm around which I could of provided access to so you could test stuff on :) |
| 19:20.44 | BlackBishop | now that this is known and tested, wonder when will the patches get rolled in the tree... |
| 19:21.50 | tzanger | yeah the arm version is identical |
| 19:21.59 | tzanger | the C compiler doesn't protect you from yourself |
| 19:22.08 | tzanger | although I bet there's a warning for unaligned accesses |
| 19:23.41 | tzanger | BlackBishop: http://lwn.net/Articles/260456/ goes into way more detail than you probably want to know |
| 19:24.36 | BlackBishop | added to the toread list :) |
| 19:24.54 | BlackBishop | now I'm glad I have a working asterisk ( FINALY ! ) after days lost trying to figure out what the heck is wrong ! |
| 19:25.45 | BlackBishop | whom do I have to talk to about the patches and release dates ? |
| 19:28.02 | tzanger | BlackBishop: I'm REALLY surprised that this was a problem. Asterisk runs on ARM just fine |
| 19:28.09 | tzanger | ARM, Blackfin... |
| 19:28.24 | BlackBishop | tzanger: well, it does run .. but I had problems auth-ing |
| 19:28.45 | BlackBishop | here: http://forums.asterisk.org/viewtopic.php?f=1&t=80420&sid=d856cb5dde59b0e90af8bbacc221390c |
| 19:35.24 | BlackBishop | it does run just fine now ! YEEY :) |
| 19:39.23 | *** join/#asterisk-dev blitzrage (Leif@asterisk/documenteur-extraordinaire/blitzrage) |
| 19:39.23 | *** mode/#asterisk-dev [+o blitzrage] by ChanServ |
| 19:40.39 | BlackBishop | mhmm, no weird chars appearing on the console either ! |
| 20:09.26 | *** join/#asterisk-dev mjordan (mjordan@conference/astricon/x-dnwdxtdfqwoexxkl) |
| 20:09.26 | *** mode/#asterisk-dev [+o mjordan] by ChanServ |
| 20:12.24 | *** join/#asterisk-dev bklang (bklang@conference/astricon/x-yptsohtuzyhfwarn) |
| 20:12.55 | bklang | Corydon76-home: are you around for a question about the build system? I'm sitting here with David and he thinks you might be able to explain a crash I'm seeing in 10-svn |
| 20:13.12 | Corydon76-home | I am around |
| 20:13.25 | bklang | This is the ticket: https://issues.asterisk.org/jira/browse/ASTERISK-18728 |
| 20:14.08 | bklang | This was working as of about 1.5 months ago, and broke when I updated for the recent Asterisk security announcement |
| 20:14.08 | bklang | It's crashing on ast_agi_register() |
| 20:14.29 | bklang | David was looking through it and was thinking that it might be related to some build system changes you made recently |
| 20:14.37 | Corydon76-home | Are you running on a redhat-derived system? |
| 20:14.40 | bklang | no, Solaris |
| 20:15.00 | Corydon76-home | Was res_agi.so already loaded when this crash occurred? |
| 20:15.04 | bklang | yes |
| 20:15.16 | bklang | I did noload => app_stack.so |
| 20:15.27 | bklang | then after Asterisk was fully booted and taking calls I ran "module load app_stack.so" and got the crash |
| 20:15.40 | bklang | (we use a lot of AGI so I know it was working) |
| 20:15.59 | Corydon76-home | The change I made that I suspect he thinks is relevant is to optional_api.h |
| 20:16.15 | Corydon76-home | but the change was to support OS X Lion |
| 20:16.16 | bklang | he agrees |
| 20:16.32 | Corydon76-home | Pastebin your autoconfig.h |
| 20:16.36 | bklang | sure |
| 20:16.52 | bklang | (sec, slow wifi here at AstriCon) |
| 20:16.55 | BlackBishop | tzanger: tried it on another arm ( netgear wndr3700 router )[C[C |
| 20:16.59 | BlackBishop | it worked out of the box |
| 20:17.14 | BlackBishop | then noticed in their trunk, they already have that patch in for their packages |
| 20:17.17 | BlackBishop | ( openwrt ) |
| 20:17.30 | Corydon76-home | bklang: by-the-by, what version of gcc are you running? |
| 20:17.52 | bklang | gcc version 3.4.3 (csl-sol210-3_4-20050802) |
| 20:18.07 | Corydon76-home | Oh, ancient |
| 20:19.32 | Corydon76-home | One suggestion, before I see your autoconfig.h, is to look at either gcc 4.0 or gcc 4.2 or later (not gcc 4.1) |
| 20:19.43 | bklang | on this version of Solaris that's not an easy option |
| 20:19.58 | Corydon76-home | Not an easy option on sunfreeware? |
| 20:20.31 | bklang | not sure, I'd have to see. So far I've been building Asterisk entirely without sunfreeware |
| 20:20.59 | Corydon76-home | Ah. It's what I've used in the past when getting a Solaris instance up to a build environment |
| 20:25.51 | Corydon76-home | Taking a while to post, isn't it? |
| 20:32.20 | BlackBishop | there's another patch in the openwrt repo, I wonder what it does .. |
| 20:32.29 | BlackBishop | - (redirect = iks_find_cdata(traversenodes, "redirect")) && |
| 20:32.34 | BlackBishop | + (redirect = iks_find_cdata(traversenodes, "sta:redirect")) && |
| 20:32.44 | BlackBishop | in channels/chan_gtalk.c |
| 20:34.45 | bklang | Corydon76-home: sorry, lost internet for a few minutes there. Here is my autoconfig.h: http://pastebin.com/0mk2s3Fb |
| 20:36.47 | wdoekes2 | tzanger: IA64 throws warnings but succeeds, ARM fails silently (at least those I've tested) |
| 20:37.50 | tzanger | fails silently? the CPU throws an exception. Unless your kernel is trapping it and fixing it up (which is possible, but not on Linux I don't think) then it should explode |
| 20:41.51 | Corydon76-home | bklang: your autoconfig.h does not show that optional_api.h changed in effect for you (or should not have, unless configure.ac somehow is no longer detecting something that it once was) |
| 20:42.45 | Corydon76-home | The added code in that commit only takes effect in a circumstance where weak_import is detected, which it is not. |
| 20:43.32 | Corydon76-home | Or hang on, maybe not |
| 20:44.26 | bklang | I'm not running autoconf, just the configure that comes with an svn checkout |
| 20:44.32 | Corydon76-home | Okay, we'll have to come up with a patch and verify on multiple platforms |
| 20:45.04 | Corydon76-home | Well, you are, you're just not regenerating the code. You run autoconf generated code when you run the configure script |
| 20:45.10 | bklang | oh ok |
| 20:46.09 | Corydon76-home | Does it also crash on 1.8? If consistent, it should. |
| 20:46.23 | Corydon76-home | Fix went into all branches |
| 20:48.33 | wdoekes2 | tzanger: no, no exceptions, but wrong values |
| 20:49.11 | tzanger | wdoekes2: that's not been my experience with ARM |
| 20:49.30 | tzanger | in fact I've had to write the exception handler to reset the system in the case of an unaligned access |
| 20:51.00 | wdoekes2 | http://tthsum.devs.nu/svn/trunk/tthsum/CHANGES.txt <-- 2 arm issues, both giving wrong values instead of exploding |
| 20:57.21 | bklang | I haven't tried 1.8 recently, but I can |
| 20:57.27 | bklang | 1.8 was working fine until I updated to 10-beta |
| 20:57.35 | bklang | then 10-beta was working fine as of late August |
| 20:57.49 | bklang | then broke when I built last week |
| 20:59.01 | bklang | I'll have to try 1.8 when I can get to a stable internet connection. I'm getting 15 second(!!) latency |
| 21:00.45 | *** join/#asterisk-dev oron_laptop (~oron_lapt@line25-214.adsl.actcom.co.il) |
| 21:02.27 | *** join/#asterisk-dev otherwiseguy (~otherwise@12.156.16.105) |
| 21:07.21 | Corydon76-home | bklang: on line 106 of optional_api.h, change "HAVE_ATTRIBUTE_weak" to "__APPLE__" and recompile |
| 21:11.16 | Corydon76-home | I just realized that on Snow Leopard, at least, __Darwin__ is no longer defined. |
| 21:14.28 | jsmith | blinks |
| 21:27.59 | Corydon76-home | jsmith: wasn't on Tiger, either. I wonder how we ever depended upon that define. |
| 21:42.27 | *** join/#asterisk-dev hardwire (~spencersr@2610:100:4fff:23::2) |
| 22:01.18 | jsmith | Corydon76-home: That's a good question... |
| 22:30.43 | *** join/#asterisk-dev marquis42 (bfhbmw0@conference/astricon/x-yjpzhjpjssjkrnrx) |
| 23:00.58 | *** join/#asterisk-dev zerohalo (~zerohalo@74.60.136.128) |
| 23:49.52 | *** join/#asterisk-dev otherwiseguy (~otherwise@69.76.221.79) |