CI and Nightlies: sporadic build failures on `make`: mv: cannot stat 't-zh_CN.gmo': No such file or directory

[eht16]

Since some time, we got build failures sporadically.

Errors are like: mv: cannot stat 't-zh_CN.gmo': No such file or directory

The language code in the error message varies a bit, I've seen 'zh_CN' and 'uk' so far. It happened at least once in the Github CI, I think and it happened at least twice in the nightly build setup. Unfortunately, I wasn't able to find the failed CI build.

I'm pretty sure this already happened before merging #1434.

Build logs: 2025-06-09: https://gist.github.com/eht16/d6e8e739e3d515e14501e023343e165f 2025-07-06: https://gist.github.com/eht16/ee48ee4ecc7ab577b4b5af3cbaee2970 2025-08-17: https://gist.github.com/eht16/912367c5ac0bf3b5eb6c6b7fbf51f211 (will add more the next time it happens)

What's interesting in the log above is that zh_CN and uk seem to appear twice in the log:

[...]
rm -f zh_CN.gmo && /usr/bin/msgfmt -c --statistics --verbose -o zh_CN.gmo zh_CN.po
rm -f uk.gmo && /usr/bin/msgfmt -c --statistics --verbose -o uk.gmo uk.po
[...]
rm -f zh_CN.gmo && /usr/bin/msgfmt -c --statistics --verbose -o zh_CN.gmo zh_CN.po
rm -f uk.gmo && /usr/bin/msgfmt -c --statistics --verbose -o uk.gmo uk.po
zh_CN.po: 351 translated messages, 565 fuzzy translations, 932 untranslated messages.
zh_CN.po: 351 translated messages, 565 fuzzy translations, 932 untranslated messages.
mv: cannot stat 't-zh_CN.gmo': No such file or directory
make[4]: *** [Makefile:392: zh_CN.gmo] Error 1
make[4]: *** Waiting for unfinished jobs....
uk.po: 1543 translated messages, 83 fuzzy translations, 222 untranslated messages.
uk.po: 1543 translated messages, 83 fuzzy translations, 222 untranslated messages.
[...]

Maybe it's related to calling make with four parallel jobs (make -j 4), just a wild guess. Otherwise I didn't see anything special with those two languages yet.

Most of the times, the very same setup builds fine. You actually find this error on the net a couple times but mostly it's related to not having the gettext package installed, which is not the case in here.

[eht16]

I don't consider this release critical, it might be a bit annoying but triggering the build again should work usually.

[b4n]

We can possibly try and (temporarily) revert d2d95bb823f05b6cca8d62cb7e7d98a73248ccea if we want to see whether it still happens or not. We have the same rule in Geany though, so it might be unrelated.