Closed pierrewillenbrockdfki closed 9 months ago
I don't mind per se ... but can you explain the need ? Anything that catches exceptions generically should really output the message already.
I don't mind per se ... but can you explain the need ? Anything that catches exceptions generically should really output the message already.
The messages are more for the benefit of the person that has to deal with tasks that don't catch the exception. Not sure if orogen/rtt does anything about exceptions leaking from the hooks, though. If that were the case, i would wonder why the messages seem to not make it to the log file.
I am wondering exactly that. The message of exceptions that are caught "leaking" off the hooks are displayed
Maybe the transformer/aggregator isn't ... in which case the best would be to fix that.
I see the discussion in https://github.com/rock-core/drivers-transformer/pull/20. I'll try to figure out what path is failing to log, maybe drop the extra logging part.
My current theory is that for all deployments, RTT ends up logging to the same orocos.log file, while base::logging is always logging to stderr, which is redirected at a separate file. This would be entirely our(dfki) own fault for not setting ORO_LOGFILE to a sensible value in all cases.
The what() is output also to stdout/stderr. At least, these error messages do end up in the "normal" log file.
True, but only if ORO_LOGLEVEL >= 3(Error); we run at ORO_LOGLEVEL=2(Critical); the file gets everything up to 5(Info)(Edit, or higher if ORO_LOGLEVEL is higher).
Dropped the changes to the NonAligningTransfomer
. Once we have fixed our process startup no extra messages are needed. Kept the one where the assertion was replaced for consistency.
This adds some more error messages and changes one assert to a message+throw because the assertion is not backed by documentation or code structure, so aborting the process seems a bit harsh.