Closed danielchasehooper closed 10 months ago
Hmm, the original behaviour is actually intended. Both errors and warnings are not supposed to terminate execution, only the panic log level.
For instance, in the validation layer, there's a single panic at the end if any validation had failed (and before that the validation layer may output multiple error log messages):
My apologies, I thought it was expected that errors would cause an abort. I misread the sokol_log implementation.
call _SG_ERROR/_SG_WARN after _SG_LOGMSG so that if the logging callback panics on errors/warnings, the log message will have made it to the console before the panic. Really helps with debugging.