net-libs/libetpan links to liblockfile.so when it's available, despite it not being an explicit dep in the ebuild. Because of that, "emerge depclean" may break libetpan and dependent packages (claws-mail, etc.). The attached patches fix that by adding an option to the configure script, and a local USE flag to the ebuild. I suggest something like "net-libs/libetpan:liblockfile - Use liblockfile for locking mailboxes" for the use.local.desc of this flag. Also, i will submit the configure.in patch upstream.
Created attachment 109175 [details, diff] files/libetpan-0.49-configure_liblockfile.patch
Created attachment 109177 [details, diff] libetpan-0.49-ebuild--IUSE_liblockfile.patch
The patch is not acceptable, since it doesn't allow libetpan to *not* use liblockfile. Current libetpan is able to work without liblockfile.
(In reply to comment #3) > The patch is not acceptable, since it doesn't allow libetpan to *not* use > liblockfile. > Well, maybe i'm missing something obvious, but i don't see why. If you don't set the liblockfile USE flag, then you'll give --disable-lockfile to ./configure, thus get $enable_lockfile=no and skip the AC_CHECK_LIB block, meaning you won't use liblockfile.
(In reply to comment #4) > (In reply to comment #3) > > The patch is not acceptable, since it doesn't allow libetpan to *not* use > > liblockfile. > > > > Well, maybe i'm missing something obvious, but i don't see why. If you don't > set the liblockfile USE flag, then you'll give --disable-lockfile to > ./configure, thus get $enable_lockfile=no and skip the AC_CHECK_LIB block, > meaning you won't use liblockfile. > Ah, sorry, I was too quick in my judgment - I misunderstood the logic of your if statement. Anyway, let's wait for libetpan upstream decision. Hoa is usually quick. I'll as k him about this tomorrow on IRC.
(In reply to comment #5) > Ah, sorry, I was too quick in my judgment - I misunderstood the logic of your > if statement. Ah, ok, no problem. > Anyway, let's wait for libetpan upstream decision. Hoa is usually quick. Seems so, yes. He has already answered me that he is in vacation, and will merge the patch when he's back. But we can wait for his actual commit to be sure, there's no hurry.
Depclean has just reminded me this bug report by unmerging liblockfile again. Hoa has merged my patch in April, so the configure option is available in versions 0.50 and above. Coud you add the corresponding USE flag in the 0.52 ebuild? Thanks.
Created attachment 129848 [details, diff] libetpan-0.52-ebuild--IUSE_liblockfile.patch
Fixed in 0.52-r1, thanks.