Is output-sync in CI worth it? It's super spammy with "entering/leaving directory" messages for some reason.
I think one can avoid that with -s
I don't want -s though
Ah, I never bothered to see what other (useful) output it removes.
Does --no-print-directory
help?
I want it to print directory, it helps emacs jump to error locations, but output-sync messes with it somehow
Indeed for me output of CI is just unreadable, I'm a big fan of fully silent builds.
Silent build output is easier to read but it is less likely that you will find relevant information in it. I prefer to search a bit in a log over having to rerun CI with manipulated options. So I am a fan of verbose builds.
What is relevant information? I am interested only in warnings and errors, the rest I don't think it is relevant, and can be saved to _build/log
as we do in the silent builds for further debugging if needed.
if something hangs we don't get _build/log
I have seen logs where I did see error messages, but it was quite hard to tell where they came from.
I don't say that one shouldn't try to organize logs in a reasonable way. My requirement for such an organized log would be that the relative frequency for having to restart CI cause of insufficient logs should be less than 1/100, better less than 1/1000. The shortest log which meets this requirement is what I really want. I guess we agree on this, we are just approaching this goal from different ends.
if something hangs we don't get _build/log
that's a bug on our CI as it should upload the _build/log
in case of timeout; in any case running a build with --display=verbose
is very easy if needed, but still the silent default is much much better, and in fact we already have filters to eliminate all the noise the CI does.
I have seen many times people spending minutes to locate an error in a make -j 8
regular Coq build
I think I added --output-sync
because otherwise we occasionally get garbled lines and the pretty-timing table gets confused. (I can't recall whether the python script ever errors on garbled logs, it's possible that it does and should be made robust to such garbling)
Last updated: Oct 13 2024 at 01:02 UTC