I got feedback due to gentoo-commits that I shouldn't set variables to their default value, e.g. S="${WORKDIR}/${P}". This could be checked by repoman, other candidates are probably RDEPEND=${DEPEND}, src_compile which does nothing but econf/emake, and probably others.
(In reply to comment #0) > > This could be checked by repoman, other candidates are probably > RDEPEND=${DEPEND}, src_compile which does nothing but econf/emake, and probably > others. > RDEPEND=${DEPEND} probably should not be checked for as I think some future EAPI version is meant to remove the automation here.
so checking for S=SOMETHING is easy. Checking what the contents of src_compile is...Honestly thats not something I want repoman doing. Its not bash, it doesn't have a bash parser; most of the stuff in repoman is fairly pragmatic, but even the quoting code is pretty complicated (but at least misquoted leads to actual breakage, having a default src_compile doesn't break anything). Maybe have a subset of tests that run during a commit; I think there are a set of issues we are looking for that we want covered so that we don't break the tree. There are also a set of issues we are looking for that are not the cause of errors but could be improved. Should we separate these out?
Do we really need checks for cosmetic stuff? (as none of this affects functionality of ebuilds)
(In reply to comment #0) > I got feedback due to gentoo-commits that I shouldn't set variables to their > default value, e.g. S="${WORKDIR}/${P}". > I do sometimes need this for vdr-plugin ebuilds. The eclass sets S to "${WORKDIR}/${P#vdr-}" but sometimes "${WORKDIR}/${P}" is needed. Or do I need to changed this to setting S before inherit - and let the eclass only set S if it is empty?
I don't see any point in this.