Closed Rangi42 closed 7 months ago
(Even if we don't want this by default, I'd like to leave all the Q
s in so you can explicitly do make Q=@
.)
(As for when there are error messages from users' modifications, @ISSOtm 's rsgbds :crab: rewrite will be making those more readable too. :heart: )
Question: Should CI have "un-quiet" output? Again, if/when there is an error, we'd see it regardless, so I don't know if seeing the individual commands is useful or just noise to scroll through in GitHub's lazy-loading log viewer.
Personally, I'm not a fan. I'm fine with the structure being put in place, but I don't want it quiet by default.
Okay. In that case, we don't even need the $(Q)
s since make
supports -s/--silent
for opt-in quiet. Just the @echo
s so even when silent, you'll see some progress. (I would still like other devs' feedback here.)
Two middle grounds: µcity echoes merely rgbasm file.asm
(downside: that looks like a command but it's not), and kernel projects instead display shorthands like ASM file.asm
(downside: it's not very clear what those are).
Also, manually outputting commands in that way wouldn't honor -s
unless scanning MAKEFLAGS
.
Not a fan in general, especially for a project that builds from the top-level and has little command line noise like this one. Debugging any sort of build system issue will always involve re-running the entire thing, which might not trigger the same issue again.
That said, I'd consider something like this for out-of-tree builds with a more bespoke makefile, as the paths wouldn't make sense anymore.
Alright, thanks for the feedback!
This may help beginners who don't notice success/error messages among all the
make
noise. Experienced users can runmake Q=
, or modify their Makefile to do so by default.The downside is that seeing all those individual commands is a definite indicator that progress is happening; but I think just waiting quietly is also okay, since the process isn't going to infinitely loop/hang, there'd be a visible error message. (Also, even currently, there's a rather long wait period before any messages show up, I think because
scan_includes
is running.)I'm not committed to this change, just wanted feedback on a concrete implementation. Like if those explicit
echo
ed messages should be edited or have more/fewer of them.