epoupon
[Bug] Magick: abort due to signal 11 (SIGSEGV) "Segmentation Fault"
In some very specific cases (I only hit this once per month or so), the LMS server crashes with a segmentation fault. This is an example of the last log:
[2025-Feb-10 08:52:01.246] 20990 - [info] - 139671583585984 [TRANSCODING] [1273] - Transcoding file '/var/lib/syncthing/Music/Number Girl/[19991216] NUMBER GIRL - シブヤROCKTRANSFORMED状態/04. 桜のダンス.mp3'
[2025-Feb-10 08:52:01.298] 20990 - [error] "Http::ResponseContinuation: WebWriteEvent::Error"
[2025-Feb-10 08:52:01.298] 20990 - [info] "WebRequest: took 779.126 ms"
[2025-Feb-10 08:52:01.533] 20990 - [error] "Http::ResponseContinuation: WebWriteEvent::Error"
[2025-Feb-10 08:52:01.534] 20990 - [info] "WebRequest: took 1020.67 ms"
Magick: abort due to signal 11 (SIGSEGV) "Segmentation Fault"...
When I tried to play again the same file with transcoding enabled I couldn't reproduce the segmentation fault.
Hello! Thanks for reporting! You mean this crashes only on files that are transcoded? What package/distrib are you using? (Lms is using STB by default, not graphicsmagick)
I only observed this crash 2 times in the last months, so I don't have that much info right now. The two crashes happened with transcoded files.
I'm using Gentoo and I am also who is maintaining the package of LMS. I don't know why I thought STB was an alternative to graphicsmagic (in the sense the later was the default one). I would change it for next release.
If I can gather more information I will add it to this issue. For now I thought even if I don't have much to properly diagnose the issue could be helpful in case other issues hits the same isue.
Can you try to set the log level to debug?
Set in in lms.conf:
log-min-severity = "info";
I do have branch that appends more information if a crash occurs (like a callstack), but unfortunately the sig handler set by magick is likely to make it useless... It also requires to compile LMS with a special option and to keep the debug symbols.
I changed the log level to "debug" and also switched to stb instead of Magick, today LMS crashed, the loks show the next message
[2025-Mar-16 06:32:03.857] 2850 - [debug] - 140473893095104 [API_SUBSONIC] Handling request 10515 '/stream.view', continuation = true, params = {c=Symfonium}, {f=json}, {id=tr-8594}, {p=*REDACTED*}, {u=david}, {v=1.13.0},
[2025-Mar-16 06:32:03.857] 2850 - [debug] - 140473893095104 [TRANSCODING] Transcoder finished = 0, total served bytes = 2097152, mime type = audio/mpeg
[2025-Mar-16 06:32:03.857] 2850 - [debug] - 140473893095104 [TRANSCODING] Writing 262144 bytes back to client
[2025-Mar-16 06:32:03.857] 2850 - [debug] - 140473893095104 [CHILDPROC] Async read, bufferSize = 262144
[2025-Mar-16 06:32:03.857] 2850 - [debug] - 140473893095104 [API_SUBSONIC] Request 10515 '/stream' handled!
[2025-Mar-16 06:32:03.857] 2850 - [error] "Http::ResponseContinuation: WebWriteEvent::Error"
[2025-Mar-16 06:32:03.857] 2850 - [info] "WebRequest: took 953.548 ms"
[2025-Mar-16 06:32:03.857] 2850 - [debug] - 140474054596288 [CHILDPROC] Closing child process...
[2025-Mar-16 06:32:03.857] 2850 - [debug] - 140474054596288 [CHILDPROC] Killing child process...
There is the same WebWriteEvent::Error but I'm surprised that LMS seems to catch the error and instead of trying to recover it seems to shutdown¿?
Hmmm interesting! Will add some internal debug facilities to have a better idea of what is wrong, will tell you when ready
As of today, with version 3.65.0, I still get the same error from time to time. After the crash, I don't see any new debug messages. It's strange that it happens so sporadically.
Last log:
[2025-Apr-17 08:52:33.042] 26717 - [debug] - 140201531786944 [API_SUBSONIC] Handling request 60160 '/stream.view', continuation = true, params = {c=Symfonium}, {f=json}, {id=tr-4662}, {p=*REDACTED*}, {u=david}, {v=1.13.0},
[2025-Apr-17 08:52:33.042] 26717 - [debug] - 140201531786944 [TRANSCODING] Transcoder finished = 0, total served bytes = 0, mime type = audio/mpeg
[2025-Apr-17 08:52:33.042] 26717 - [debug] - 140201531786944 [TRANSCODING] Writing 262144 bytes back to client
[2025-Apr-17 08:52:33.043] 26717 - [debug] - 140201531786944 [CHILDPROC] Async read, bufferSize = 262144
[2025-Apr-17 08:52:33.043] 26717 - [debug] - 140201531786944 [API_SUBSONIC] Request 60160 '/stream' handled!
[2025-Apr-17 08:52:33.114] 26717 - [debug] - 140201523394240 [CHILDPROC] Async read cb - ec = 'Success' (0), bytesTransferred = 262144
[2025-Apr-17 08:52:33.114] 26717 - [debug] - 140201523394240 [TRANSCODING] Have 262144 more bytes to send back
[2025-Apr-17 08:52:33.114] 26717 - [debug] - 140201523394240 [API_SUBSONIC] Handling request 60161 '/stream.view', continuation = true, params = {c=Symfonium}, {f=json}, {format=mp3}, {id=tr-16154}, {maxBitRate=256}, {p=*REDACTED*}, {u=david}, {v=1.13.0},
[2025-Apr-17 08:52:33.114] 26717 - [debug] - 140201523394240 [TRANSCODING] Transcoder finished = 0, total served bytes = 262144, mime type = audio/mpeg
[2025-Apr-17 08:52:33.115] 26717 - [debug] - 140201523394240 [TRANSCODING] Writing 262144 bytes back to client
[2025-Apr-17 08:52:33.115] 26717 - [debug] - 140201523394240 [CHILDPROC] Async read, bufferSize = 262144
[2025-Apr-17 08:52:33.115] 26717 - [debug] - 140201523394240 [API_SUBSONIC] Request 60161 '/stream' handled!
[2025-Apr-17 08:52:33.115] 26717 - [error] "Http::ResponseContinuation: WebWriteEvent::Error"
[2025-Apr-17 08:52:33.115] 26717 - [info] "WebRequest: took 381.093 ms"
[2025-Apr-17 08:52:33.115] 26717 - [debug] - 140201670198976 [CHILDPROC] Closing child process...
[2025-Apr-17 08:52:33.115] 26717 - [debug] - 140201670198976 [CHILDPROC] Killing child process...
Hello! Thanks for the update. Indeed I didn't add any crash analysis yet , will do!
Using version 3.69.0 I noticed the crash with the message: Magick: abort due to signal 11 (SIGSEGV) "Segmentation Fault"... but I found it curious since it was using STB (it showed Using stb (resize version 2) when compiling)
Ah then maybe this is because Wt is built using magick support. Could you check using ldd on wt libraries?