Open thesamesam opened 1 year ago
I suppose the problem is due to the combination of failglob
and the subshell.
Basically, we catch two problems:
[]
in global scope because it causes failglob
to exit while sourcing ebuild.${PYTHON_USEDEP}
directly because it causes junk to land in the variable.The problem is that unescaped []
inside the subshell causes the subshell to exit immediately with the error, and $()
evaluates to an empty string. However, we never catch that error (I'm not even sure if it's possible) and empty string is valid in this context, so the user never sees a problem (except for noise in output).
Steps to reproduce:
git checkout 72a1fbf3592c8ceee6f797e36fe474c2d251ee47~1
caja-dropbox-1.26.0.ebuild clean install
The bad *DEPEND snippet:
I feel like something should be dying fatally here? Apparently
egencache
shows it too:pkgcheck scan
isn't bothered by it.