Created attachment 366178 [details, diff] readline-6.2_p4.patch This is my take on adding multilib support for readline. It's needed for multilib sqlite. Please have a look at my changes :)
FYI, base-system packages are not supposed to be EAPI-5
Comment on attachment 366178 [details, diff] readline-6.2_p4.patch >+++ sys-libs/readline/readline-6.2_p4.ebuild update to 6.2_p5 >+ multilib_copy_sources fix the ebuild instead to build out of tree
I'll work on it.
Created attachment 366224 [details, diff] readline-6.2_p5.ebuild patch
Created attachment 366226 [details, diff] readline-6.2_p5.ebuild patch, r1 - added missing einstalldocs call, - fixed HTML docs to be installed in $docdir/html rather than $docdir/ps/html/doc :).
Created attachment 366242 [details, diff] readline-6.2_p5.ebuild patch, r2 This one respects AR properly.
Comment on attachment 366242 [details, diff] readline-6.2_p5.ebuild patch, r2 >+multilib_src_configure() { >+ # Make sure configure picks a better ar than `ar`. #484866 >+ export ac_cv_prog_AR=$(tc-getAR) just do it in the main configure func ? >-src_install() { >- default >+multilib_src_install() { >+ emake DESTDIR="${D}" install why the change from default ? don't want to `dodoc` more than once ? i'm somewhat inclined to just live with that. >+multilib_src_install_all() { >+ einstalldocs hrm, seems you committed this to the eclass w/out proper documentation. can you fix that please ?
(In reply to SpanKY from comment #7) > >+multilib_src_configure() { > >+ # Make sure configure picks a better ar than `ar`. #484866 > >+ export ac_cv_prog_AR=$(tc-getAR) > > just do it in the main configure func ? Well, it supposedly differs per ABI. Another thing is that it looks for i686-*-ar while we have only i386-*-ar, and both ar instances are the same anyway. > >-src_install() { > >- default > >+multilib_src_install() { > >+ emake DESTDIR="${D}" install > > why the change from default ? don't want to `dodoc` more than once ? i'm > somewhat inclined to just live with that. Sure, either is fine for me. > >+multilib_src_install_all() { > >+ einstalldocs > > hrm, seems you committed this to the eclass w/out proper documentation. can > you fix that please ? Oh, I wonder why I did that. Will do later today, thanks for noticing.
Created attachment 366292 [details, diff] readline-6.2_p5.ebuild patch, r3 Fixed what you pointed out, also committed einstalldocs docs :). Ok to commit now?
(In reply to Michał Górny from comment #8) > Well, it supposedly differs per ABI. sure, but so far, ebuilds haven't been doing this. in fact, they've been doing `tc-export` which exports to the env, and then later calls re-use that (unless any calls happen to be in subshells). maybe the multilib code always uses subshells as a feature rather than incidentally via multiprocessing. > Another thing is that it looks for i686-*-ar while we have only i386-*-ar, > and both ar instances are the same anyway. it looks for $CHOST-ar. i don't know what you mean by i386-* ... those are only installed if you have that as your CHOST or you installed a cross-compiler. i think it's reasonable to expect people who are doing multilib to be using the same toolchain. i'm not sure it's really feasible to hand maintain your own i686-* toolchain in parallel to a non-multilib x86_64-* toolchain. it's not like we have variables even available to set e.g. CC on a per-ABI basis.
(In reply to Michał Górny from comment #9) lgtm; feel free to commit
(In reply to SpanKY from comment #10) > (In reply to Michał Górny from comment #8) > > Another thing is that it looks for i686-*-ar while we have only i386-*-ar, > > and both ar instances are the same anyway. > > it looks for $CHOST-ar. i don't know what you mean by i386-* ... those are > only installed if you have that as your CHOST or you installed a > cross-compiler. Well, I have no idea how those files got onto my system. Curious enough, they all point to x86_64 binutils. I guess it's not worth wondering about. (In reply to SpanKY from comment #11) > (In reply to Michał Górny from comment #9) > > lgtm; feel free to commit +*readline-6.2_p5-r1 (27 Dec 2013) + + 27 Dec 2013; Michał Górny <mgorny@gentoo.org> +readline-6.2_p5-r1.ebuild: + Add multilib support, bug #496006.
