Summary: | ASTERISK-06994: monitor_format under realtime (mysql) causes crash on bridging agent to caller | ||
Reporter: | Jonathan Towne (jontow) | Labels: | |
Date Opened: | 2006-05-17 20:33:30 | Date Closed: | 2006-05-26 12:54:39 |
Priority: | Critical | Regression? | No |
Status: | Closed/Complete | Components: | Applications/app_queue |
Versions: | Frequency of Occurrence | ||
Related Issues: | |||
Environment: | Attachments: | ||
Description: | If given either a default of 'wav' (any string appears to do) or if "monitor_format" is explicitly set in the realtime database, a crash is caused, see 'additional information' for the snippet from the DEBUG log. When switching back to flat-file config, this is not reproducable; if the field "monitor_format" is NULL in the DB, it does not exhibit this behavior. With some assistance, further information can be provided (backtrace, 'strings' from corefile, whatever is necessary.) At this point, I do not have enough information to generate a patch, assistance would be greatly appreciated. Our workaround is currently to default "monitor_format" to NULL. ****** ADDITIONAL INFORMATION ****** OS: FreeBSD 6.1-RC, asterisk running on host as well as in a jail; ztdummy from SVN. MySQL versioned as such: "mysql Ver 14.7 Distrib 4.1.18, for portbld-freebsd6.1 (i386) using 5.0" The following debug log is slightly annotated where it may be useful. ------------- *** AT THIS POINT: f->subclass = 3 May 17 20:28:12 VERBOSE[58971] logger.c: -- Agent/5000 is ringing May 17 20:28:13 VERBOSE[58971] logger.c: -- SIP/5000-6b24 is ringing May 17 20:28:15 VERBOSE[58971] logger.c: -- SIP/5000-6b24 is ringing May 17 20:28:17 DEBUG[58971] chan_sip.c: Acked pending invite 102 May 17 20:28:17 DEBUG[58971] chan_sip.c: Oooh, we need to change our formats since our peer supports only 0x2 (gsm) and not 0x4 (ulaw) May 17 20:28:17 DEBUG[58971] chan_sip.c: build_route: Contact hop: <sip:5000@192.168.2.150:2057;line=tex3psas> May 17 20:28:17 VERBOSE[58971] logger.c: -- SIP/5000-6b24 answered Local/5000@default-8a5a,2 *** AT THIS POINT: f->subclass = -1 May 17 20:28:17 DEBUG[58971] app_queue.c: Dunno what to do with control type -1 *** AT THIS POINT: f->subclass = 4 May 17 20:28:17 VERBOSE[58971] logger.c: -- Agent/5000 answered IAX2/192.168.2.13:4569-1 *** Crashes here.. no more log messages, segfault and coredump on the console. ------------- | ||
Comments: | By: Jonathan Towne (jontow) 2006-05-17 21:02:55 "bt full" output for all displayed threads available at: http://ppx.slic.com/~jontow/monitor_format_core_all_threads.txt I'll upload it if someone deems it necessary. I'll retrieve the correct info with some help. By: Serge Vecher (serge-v) 2006-05-25 09:31:38 jontow: does this also occur in 1.2.7.1? By: BJ Weschke (bweschke) 2006-05-26 12:54:39 fixed yesterday in /trunk. |