Question: Effect of LOG4CPLUS_REQUIRE_EXPLICIT_INITIALIZATION on Windows? about log4cplus HOT 5 CLOSED

If anyone sets the AsyncAppend=true option and the thread pool has been disabled at compile time, the code will silently use the synchronous logging instead: https://github.com/log4cplus/log4cplus/blob/2.1.x/src/appender.cxx#L296-L320

from log4cplus.

See global-init.cxx for the gory details of initialisation and deinitialisation. The LOG4CPLUS_REQUIRE_EXPLICIT_INITIALIZATION does something for initialisation only. It does not do anything for deinitialisation. The deinitialisation is necessary mainly for the thread pool to be shut down. If you are not going to use the asynchronous logging feature, you can set LOG4CPLUS_ENABLE_THREAD_POOL=OFF at compile time and the thread pool will not be created. I have not tested this but I think you should not get stuck threads when you exit main() even without the Initializer

from log4cplus.

Thanks a log @wilx for your swift response. It makes the matter totally clear to me.
We are using the configuration file to configure the framework, so I cannot rule out that someone would use the AsyncAppender. Hence, I cannot disable the thread pool (correct me if I am wrong).

I just wanted to make sure not to miss any built-in option to solve the issue. I will now implement a viable workaround in my shared library.

Shall I close this issue now?

from log4cplus.

This is incredible @wilx ! I am really baffled by your support! Thank you so much. I'll try this ASAP and give you feedback here. But I have no doubt that your patch is going to work.

Thanks again, this project is amazing.

from log4cplus.

@wilx The feedback as promised: The patch works nicely on windows (and is not needed on Linux). To me this issue has now been entirely solved within a single day :-D

Thanks a lot for your great support. I'll close the issue now.

from log4cplus.