| 00:00.15 | X-Rob | whoever thought `cat file | md5sum ` is a good idea wins the 'useless cat' award for this week |
| 00:00.32 | X-Rob | <PROTECTED> |
| 00:00.32 | X-Rob | --disable BUILD_NATIVE --enable res_statsd --enable res_chan_stats --enable res_endpoint_stats menuselect.makeopts |
| 00:00.35 | X-Rob | building now |
| 00:01.22 | gtjoseph | X-Rob: what platform? |
| 00:01.29 | X-Rob | centos6 and centos7 |
| 00:02.02 | gtjoseph | cat | md5sum isn't useless :) some versions ONLY take from stdin |
| 00:02.13 | X-Rob | gtjoseph: Bet you're wrong. |
| 00:02.23 | gtjoseph | bet i'm not. i spent a lot of time on that |
| 00:02.30 | gtjoseph | freebsd maybe? |
| 00:02.38 | X-Rob | nah |
| 00:03.00 | X-Rob | All I Can think of is some strange cygwin environemnt? |
| 00:03.12 | gtjoseph | anyway we compile on centos7 all day |
| 00:03.20 | gtjoseph | what was the exact issue? |
| 00:03.43 | X-Rob | I got to the 'current' md5sum was wrong, even though it was right. |
| 00:03.44 | gtjoseph | wait...cygwin??? |
| 00:04.08 | X-Rob | no, you asked 'which md5sum doesn't take a param', and guessed cygwin or something dumb like that |
| 00:04.12 | gtjoseph | oh |
| 00:04.17 | gtjoseph | thank god |
| 00:04.40 | gtjoseph | i was going to redirect you to our "compiling for windows" page :) |
| 00:04.59 | gtjoseph | so exactly what version are you cimpiling? |
| 00:05.04 | gtjoseph | or compiling |
| 00:05.05 | X-Rob | I was going to rewrite it as 'md5sum /tmp/phjsip-foo.bz | cut -d\ -f1' and 'grep $FILENAME /tmp/pjsip.m5ssum | cut -d\ -f1' |
| 00:05.15 | X-Rob | rather than the messy sed |
| 00:05.32 | gtjoseph | ok but i'd rather figure out what |
| 00:05.35 | gtjoseph | s broken |
| 00:05.47 | gtjoseph | and see it for myself :) |
| 00:06.35 | X-Rob | I got the IMPRESSION that it was trying to read a file that didn't exist. The complexity is because we're doing it in mock ( because, repeatable!) it's hard to add debugging. I spent some time on it on monday, but gave up |
| 00:06.39 | X-Rob | and just removed the validate call |
| 00:07.01 | gtjoseph | ok. if you figure it out, let me know |
| 00:07.10 | X-Rob | Yeah, I will, and I'll submit a patch. |
| 00:07.16 | gtjoseph | cool |
| 00:07.20 | X-Rob | I'm going to remove the cat | md5sum tho |
| 00:07.27 | X-Rob | and you can put it back and call me a bastard |
| 00:08.10 | gtjoseph | i tested that on freebsd centos6/7 ubuntu14/16 fedora |
| 00:09.15 | X-Rob | Find me an OS that doesn't take a param to md5sum and I'll spend more time on it |
| 00:09.25 | X-Rob | rather than just doing it my way and seeing if it works |
| 00:09.40 | gtjoseph | show me that it doesn't work and I'LL spend more time on it :) |
| 00:12.11 | X-Rob | https://www.irccloud.com/pastebin/JXAdRHCu/ |
| 00:12.16 | X-Rob | It REALLY doesn't work. |
| 00:12.24 | X-Rob | But, it REALLY doesn't work in mock |
| 00:12.30 | X-Rob | maybe mock's chroot is doing strange stuff to it |
| 00:13.18 | gtjoseph | yeah you just disables the verify step |
| 00:14.03 | X-Rob | Yep, it really truly, honestly, doesn't work. But, because there's all this added complexity that we do, I didn't want to bother you guys with it yet |
| 00:14.15 | gtjoseph | ok, no problem |
| 00:14.28 | X-Rob | OK |
| 00:14.29 | X-Rob | FINALLY |
| 00:14.32 | X-Rob | I have an asterisk built |
| 00:24.55 | *** join/#asterisk-dev newtonr (~newtonr@173-17-133-211.client.mchsi.com) |
| 00:24.55 | *** mode/#asterisk-dev [+o newtonr] by ChanServ |
| 00:27.57 | X-Rob | Aaaand it's a heisenbug |
| 00:28.16 | X-Rob | not crashing with MALLOC_DEBUG |
| 00:43.01 | gtjoseph | X-Rob: oh yeah. it wasn't that md5sum didn't take options, it was that the options weren't the same so it was better to go with no options. :) |
| 00:52.36 | *** part/#asterisk-dev kharwell (kharwell@nat/digium/x-dentktyhtsgloqfl) |
| 01:08.12 | X-Rob | Here's something odd |
| 01:08.27 | X-Rob | https://www.irccloud.com/pastebin/CBtFeLFW/13.13.1-and-13.14.0.txt |
| 01:08.53 | X-Rob | the actual asterisk binary crashed in locale stuff |
| 01:09.54 | gtjoseph | can't really tell much |
| 01:10.05 | X-Rob | I just think it's odd that that was connected to 13.13.1 |
| 01:10.14 | X-Rob | but the 13.14.0 client shit the bed, too |
| 01:11.03 | gtjoseph | check libedit. what version? |
| 01:20.19 | X-Rob | [root@ipv6 asterisk]# rpm -qa | grep edit |
| 01:20.19 | X-Rob | libedit-2.11-4.20080712cvs.1.el6.x86_64 |
| 01:20.19 | X-Rob | [root@ipv6 asterisk]# |
| 01:20.37 | X-Rob | [root@sng7 ~]# rpm -qa | grep edit |
| 01:20.37 | X-Rob | libedit-3.0-12.20121213cvs.el7.x86_64 |
| 01:20.37 | X-Rob | [root@sng7 ~]# |
| 01:21.41 | X-Rob | Ahha. IT's something to do with the way we're loading modules |
| 01:22.00 | X-Rob | https://www.irccloud.com/pastebin/6URVsZPx/modules.conf |
| 01:27.52 | X-Rob | just having autoload=yes in ther causes it to crash |
| 01:27.56 | X-Rob | so .. it's something else |
| 01:30.07 | X-Rob | you know, tm1000 said it was something to do with hints |
| 01:30.25 | X-Rob | without extensions* it doesn't crash, with extensions* it does crash |
| 01:38.15 | X-Rob | *sigh* |
| 01:38.19 | X-Rob | This is a regression |
| 01:38.41 | X-Rob | Remember how we reported the bug with the bogus extension string causing asteirsk to crash? |
| 01:38.53 | X-Rob | I left those in |
| 01:38.57 | X-Rob | https://www.irccloud.com/pastebin/2SSgjl3R/ |
| 01:39.03 | X-Rob | and they're now causing asterisk to crash. |
| 01:39.38 | X-Rob | I dunno what Qwell is seeing, but that explains why his crash is different to mine |
| 01:43.06 | X-Rob | https://issues.asterisk.org/jira/browse/ASTERISK-26668 |
| 05:07.35 | *** join/#asterisk-dev tuxd00d (~tuxd00d@ip68-106-11-24.ph.ph.cox.net) |
| 06:00.11 | *** join/#asterisk-dev Demon_VoIP (~demon@109.60.135.186) |
| 06:01.38 | Demon_VoIP | Fresh sources from git branch 13 with 0008-Re-1945-misc-Don-t-trigger-SRV-complete-callback-whe.patch. 10 hours uptime! I am happy. |
| 06:12.17 | *** join/#asterisk-dev bof22 (~Thunderbi@185.13.183.107) |
| 06:12.48 | *** join/#asterisk-dev bof22 (~Thunderbi@185.13.183.107) |
| 07:14.10 | Demon_VoIP | oops, asterisk segfaults and restarted. 11 hours uptime. Ok, it's not bad any way. I'll not rollback i think |
| 07:50.11 | *** join/#asterisk-dev tsearle (~tsearle@37.19.10.33) |
| 07:59.18 | *** join/#asterisk-dev tsearle (~tsearle@37.19.10.33) |
| 08:04.26 | *** join/#asterisk-dev pchero_work (~pchero@2a00:c80:1072::10de) |
| 08:55.52 | *** join/#asterisk-dev tuxian (~tuxian@igilmour.plus.com) |
| 08:59.38 | *** join/#asterisk-dev hehol (~hehol@gatekeeper.loca.net) |
| 10:19.32 | *** join/#asterisk-dev jkroon (~jkroon@41.13.124.224) |
| 11:50.28 | *** join/#asterisk-dev oej (~oej@194.15.212.5) |
| 11:50.28 | *** mode/#asterisk-dev [+o oej] by ChanServ |
| 12:44.41 | *** join/#asterisk-dev tuxian (~tuxian@igilmour.plus.com) |
| 13:54.20 | file | gtjoseph: https://issues.asterisk.org/jira/browse/ASTERISK-26802 |
| 13:54.50 | gtjoseph | hmmm. |
| 13:55.15 | gtjoseph | X-Rob reported issues here last night as well |
| 13:55.20 | gtjoseph | with md5sum |
| 13:56.06 | gtjoseph | it's on my "fix while on vacation next week" list. |
| 13:56.19 | gtjoseph | i'll copy it |
| 13:57.45 | gtjoseph | do you see the '[patch]'? |
| 13:58.28 | file | it'll probably go up on gerrit |
| 13:58.41 | file | once elguero sorts it out |
| 13:59.21 | gtjoseph | i'm not going to copy the issue then |
| 14:23.27 | *** join/#asterisk-dev tuxd00d (~tuxd00d@ip68-106-11-24.ph.ph.cox.net) |
| 14:26.03 | *** join/#asterisk-dev jkroon (~jkroon@uls-154-73-35-201.wall.uls.co.za) |
| 15:01.48 | *** join/#asterisk-dev kharwell (kharwell@nat/digium/x-wmavqmyetlvozsod) |
| 15:01.48 | *** mode/#asterisk-dev [+o kharwell] by ChanServ |
| 15:38.28 | *** join/#asterisk-dev cresl1n (Adium@asterisk/libpri-and-libss7-expert/Cresl1n) |
| 15:38.28 | *** mode/#asterisk-dev [+o cresl1n] by ChanServ |
| 15:51.29 | *** join/#asterisk-dev rmudgett (rmudgett@nat/digium/x-erltpzzodrsiemdv) |
| 15:51.29 | *** mode/#asterisk-dev [+o rmudgett] by ChanServ |
| 15:52.32 | *** join/#asterisk-dev kharwell (kharwell@nat/digium/x-xeuxlvrwbjulkfnf) |
| 15:52.32 | *** mode/#asterisk-dev [+o kharwell] by ChanServ |
| 16:59.47 | *** join/#asterisk-dev kharwell (kharwell@nat/digium/x-llnbjjgslnazqulj) |
| 16:59.47 | *** mode/#asterisk-dev [+o kharwell] by ChanServ |
| 17:14.39 | kharwell | in pjsip_options at startup it appears we loop through all aors and for each of those we loop through each endpoint. Then for each of those endpoints (that had a matching aor) it loops over its aors and then updates the endpoint's persistent state. |
| 17:15.10 | kharwell | Is there a reason to loop over all the aors initially? |
| 17:15.23 | gtjoseph | yes. that whole logic is quite bizxarre |
| 17:15.54 | gtjoseph | it's been hacked on so many times to "improve performance" it's not fully |
| 17:15.57 | gtjoseph | funny |
| 17:16.00 | kharwell | Instead just loop over all endpoints and if it has at least one aor set the state to online |
| 17:16.28 | file | I'd say before any changes are made to it, define exactly what it needs and the behavior |
| 17:16.30 | gtjoseph | the problem is that in order to do that, you have to retrive all endpoints from the db |
| 17:16.34 | kharwell | I think the performance was improved for realtime, but now config files are suffering |
| 17:16.45 | gtjoseph | could be |
| 17:16.54 | gtjoseph | well actually |
| 17:17.21 | gtjoseph | not sure how config files could be worse than realtime |
| 17:17.52 | kharwell | I think it is spending a lot of time down in ast_variable_lists_match which is not used by realtime |
| 17:18.01 | gtjoseph | ah possibly |
| 17:18.35 | gtjoseph | one of the things on my database improvement list is to use sqlite3 as the internal store instead of ao2 containers |
| 17:18.55 | kharwell | 3000+ aors and for each of those 3000+ endpoints. For each of those it tries to match the aor using that variable list match |
| 17:19.04 | gtjoseph | this way we can use sql to find the aors with qualify_timeout > 0 and do a join with endpoint |
| 17:19.20 | gtjoseph | yeah |
| 17:19.45 | kharwell | if I just retrieve all endpoints (pass in NULL instead of a var) it's still a bit slow, but loads up everything under 30 seconds maybe?? where as with the "var" it takes 10+ minutes |
| 17:20.07 | gtjoseph | 10 minutes? |
| 17:20.41 | gtjoseph | i have a stress test pjsip.conf with 2k extensions and i don't remember it taking that long |
| 17:20.48 | gtjoseph | tests again |
| 17:21.11 | kharwell | actually that was after I modified it to match exact on aors instead of a "like". With a "like" I am not sure how long it took. I eventually just killed it |
| 17:21.51 | kharwell | 2k extensions doesn't take as long |
| 17:22.29 | gtjoseph | i just did it in under 3 seconds |
| 17:22.41 | kharwell | hrm |
| 17:22.59 | gtjoseph | let me make sure i have qualify tuened on |
| 17:23.45 | kharwell | I shouldn't actually is the weird thing. I have "qualify_frequency=0" |
| 17:24.13 | gtjoseph | yeah you have something else weird going on |
| 17:24.14 | rmudgett | gtjoseph: That liquid nitrogen quantum computer of yours will make short work of most anything. |
| 17:24.30 | gtjoseph | i just set qf=60 and it loaded almost immediately |
| 17:24.32 | file | cloud threads. |
| 17:24.35 | gtjoseph | ha! |
| 17:24.55 | gtjoseph | how are you measuring "done"? |
| 17:25.09 | rmudgett | When the Asterisk Ready prompt appears. |
| 17:25.22 | gtjoseph | i'm asking @kharwell |
| 17:26.41 | kharwell | yeah when the prompt appears |
| 17:26.56 | kharwell | I can also put logging in to count as it's looping and it slow |
| 17:27.22 | gtjoseph | something else is going on |
| 17:27.34 | gtjoseph | what compile options? |
| 17:30.13 | *** join/#asterisk-dev cresl1n (Adium@asterisk/libpri-and-libss7-expert/Cresl1n) |
| 17:30.13 | *** mode/#asterisk-dev [+o cresl1n] by ChanServ |
| 17:30.54 | kharwell | don't optimize, compile double, loadable better backtraces, build native, optional api, do crash and test framework |
| 17:31.05 | kharwell | dev mode enabled |
| 17:35.48 | *** join/#asterisk-dev cresl1n (Adium@asterisk/libpri-and-libss7-expert/Cresl1n) |
| 17:35.48 | *** mode/#asterisk-dev [+o cresl1n] by ChanServ |
| 17:55.53 | *** join/#asterisk-dev luckman212 (~luckman21@unaffiliated/luckman212) |
| 18:02.33 | *** join/#asterisk-dev cresl1n (Adium@asterisk/libpri-and-libss7-expert/Cresl1n) |
| 18:02.33 | *** mode/#asterisk-dev [+o cresl1n] by ChanServ |
| 18:51.18 | seanbright | can someone tell me what the following line does: |
| 18:51.19 | seanbright | struct digestkeys *i = i; |
| 18:51.41 | seanbright | i assume a whole lot of nothin' |
| 18:52.00 | gtjoseph | cause a segfault? |
| 18:52.12 | gtjoseph | where do you see it? |
| 18:52.15 | seanbright | it's just an assignment |
| 18:52.20 | seanbright | sip_digest_parser in chan_sip.c |
| 18:52.32 | gtjoseph | oh. i'm not looking in there. |
| 18:52.36 | seanbright | heh |
| 18:52.37 | gtjoseph | i just ate |
| 18:52.39 | seanbright | i think it's a noop |
| 18:52.55 | gtjoseph | depends on what i is |
| 18:53.01 | seanbright | it's a point |
| 18:53.06 | seanbright | s/point/pointer/ |
| 18:53.18 | seanbright | it's self-assignment |
| 18:53.45 | gtjoseph | well, it's setting the contents of the location i points to to the location |
| 18:54.03 | seanbright | right, but it never deferences it |
| 18:54.17 | gtjoseph | weird |
| 18:54.21 | seanbright | anyway, good times. |
| 18:54.29 | gtjoseph | do a git blame and see who wrote it. |
| 18:54.33 | seanbright | meh |
| 18:54.47 | Qwell | git blame on chan_sip? See you in a week. |
| 18:54.52 | seanbright | for realz |
| 18:56.11 | *** join/#asterisk-dev oej (~oej@h87-96-134-129.cust.se.alltele.net) |
| 18:56.12 | *** mode/#asterisk-dev [+o oej] by ChanServ |
| 18:57.59 | gtjoseph | (Jonathan Rose 2011-09-22 16:35:20 +0000 17066) struct digestkeys *i = i; |
| 18:58.09 | seanbright | ugh. jrose. so typical. |
| 18:58.16 | seanbright | i kid. |
| 19:06.27 | *** join/#asterisk-dev oej (~oej@h87-96-134-129.cust.se.alltele.net) |
| 19:06.27 | *** mode/#asterisk-dev [+o oej] by ChanServ |
| 19:39.59 | *** join/#asterisk-dev cresl1n (Adium@asterisk/libpri-and-libss7-expert/Cresl1n) |
| 19:39.59 | *** mode/#asterisk-dev [+o cresl1n] by ChanServ |
| 20:07.39 | *** join/#asterisk-dev cresl1n (Adium@asterisk/libpri-and-libss7-expert/Cresl1n) |
| 20:07.39 | *** mode/#asterisk-dev [+o cresl1n] by ChanServ |
| 20:15.51 | *** join/#asterisk-dev cresl1n (Adium@asterisk/libpri-and-libss7-expert/Cresl1n) |
| 20:15.51 | *** mode/#asterisk-dev [+o cresl1n] by ChanServ |
| 20:23.53 | rmudgett | Good thing the results of that assignment aren't used. |
| 20:53.09 | *** join/#asterisk-dev newtonr (RustyNewto@nat/digium/x-hirfgbuuxtrakbat) |
| 20:53.09 | *** mode/#asterisk-dev [+o newtonr] by ChanServ |
| 21:07.56 | *** join/#asterisk-dev pchero (~pchero@109.70.54.56) |
| 21:40.15 | *** join/#asterisk-dev newtonr (RustyNewto@nat/digium/x-wqpsddituxhotcsd) |
| 21:40.15 | *** mode/#asterisk-dev [+o newtonr] by ChanServ |
| 21:48.49 | *** join/#asterisk-dev tuxd00d (~tuxd00d@wsip-184-176-140-167.ph.ph.cox.net) |
| 22:25.50 | *** join/#asterisk-dev pchero (~pchero@109.70.54.56) |
| 22:33.14 | *** join/#asterisk-dev tuxd00d (~tuxd00d@wsip-184-176-140-167.ph.ph.cox.net) |
| 22:46.57 | *** join/#asterisk-dev tuxd00d (~tuxd00d@wsip-184-176-140-167.ph.ph.cox.net) |
| 22:59.30 | X-Rob | FYI file, I pulled https://gerrit.asterisk.org/#/c/4960/ into the FreePBX asterisk RPMs and confirmed the problem is fixed. Hence https://community.freepbx.org/t/asterisk-restarts-after-clicking-apply-config-changes/39577/8?u=xrobau |
| 22:59.36 | X-Rob | so that's a +1 from me |
| 23:00.37 | *** part/#asterisk-dev kharwell (kharwell@nat/digium/x-llnbjjgslnazqulj) |
| 23:19.37 | *** join/#asterisk-dev tuxd00d (~tuxd00d@ip68-106-11-24.ph.ph.cox.net) |