Bug: Server does not exit gracefully

[dimitry-ishenko]

Observed Behavior

For example, starting server in an empty directory (without casparcg.config), I get this:

dimitry@laptop:~/tmp$ casparcg 
Type "q" to close application.
Aborted

If there is an earlier backtrace in the directory, I will get this:

dimitry@laptop:~/tmp$ casparcg 
[2024-09-19 23:12:11.593643] [0x000079c055fba700] [error]    0# 0x00005D8FCAB0C9B4
 1# 0x0000708F77A45320
 2# 0x0000708F77A9EB1C
 3# 0x0000708F77A4526E
 4# 0x0000708F77A45320
 5# 0x00005D8FCAB850BC
 6# 0x00005D8FCAB80955
 7# 0x00005D8FCAB85E25
 8# 0x0000708F86CD797F
 9# 0x0000708F77A9CA94
10# 0x0000708F77B29C3C

Type "q" to close application.
Aborted

If the config file is corrupt, I may get one of these:

dimitry@laptop:~/tmp/casparcg$ casparcg-server-2.4 
Type "q" to close application.

dimitry@laptop:~/tmp/casparcg$ casparcg-server-2.4 
Type "q" to close application.
Aborted

dimitry@laptop:~/tmp/casparcg$ casparcg-server-2.4 
[2024-09-19 23:15:11.469021] [0x000070382a705700] [error]    0# 0x000059878C21D9B4
 1# 0x000071F27B845320
 2# 0x000071F27B89EB1C
 3# 0x000071F27B84526E
 4# 0x000071F27B845320
 5# 0x000059878C2960BC
 6# 0x000059878C291955
 7# 0x000059878C296E25
 8# 0x000071F28C19A97F
 9# 0x000071F27B89CA94
10# 0x000071F27B929C3C

Type "q" to close application.

Similar thing happens if I try to launch it with, eg casparcg -h or casparcg --help or any other invalid path to the config file.

Expected behaviour

Looking at the code, I should be getting an error message:

 ### Invalid configuration file. ###

But, I don't. Running in Debug mode, it seems to abort on this line:

server/src/common/log.cpp

Line 152 in 804b8de

append_timestamp(pre_message_stream, boost::log::extract<boost::posix_time::ptime>("TimestampMillis", rec).get());

There appears to be some sort of a race condition, because if I step through the code it does sometimes display the message...

Steps to reproduce

See above.

Environment

• Server version: v2.4.0
• Operating system: Ubuntu Linux 24.04

[alevecchio]

I have a similar exit when I try to start CasparCG without a mounted drive. I didn't check the error code to see if is the same.
Config file is ok, but the media drive "is not there".

[dimitry-ishenko]

I have a similar exit when I try to start CasparCG without a mounted drive. I didn't check the error code to see if is the same. Config file is ok, but the media drive "is not there".

I suspect it's the same problem. The server tried to access the media path, an exception gets thrown and it aborts during logging.

[dimitry-ishenko]

Just confirmed that if I set media path to a non-existent directory and the server doesn't have write access to create it, it will fail in the same way.

[Julusian]

This should be better now.

The root problem looks to have been that the main method and thread exited before the logs had been flushed. Adding a simple flush call appears to have fixed this.

I've also put a bit of effort into improving the logging when it fails to find a config file, or encounters an invalid config file, so errors here should be a bit more descriptive and not just a long stack trace