Created attachment 290535 [details] eprefixified mysql-v2.eclass dev-db/mysql-5.1.56-r1 introduced a new set of eclasses for mysql that prevented proper building on systems running Gentoo prefix. The following attachements are based on the old mysql.eclass, thus I hope for fast inclusion. ;) mysql-cmake.eclass was not eprefixied (bc it wasn't needed on my system), but can be done if it is wished. On a side-note: I'll be adding #385001 as a blocker for this, because I'm still getting procps as a dependency for mysql.
Created attachment 290537 [details, diff] eprefixified mysql-autotools.eclass
Created attachment 290539 [details, diff] eprefixified mysql_fx.eclass
let's get mysql team in the loop
I was about to apply the patches when I recalled that the mysql/mariadb ebuilds are still EAPI-2 and EROOT is only set on EAPI-3. @Robin, are you ok with bumping the EAPI to 3 or 4?
I applied your patches and bumped the EAPI in the ebuilds to 4. I've also did some work to "eprefixify" mysql-cmake.eclass. Please test the changes on prefix. Please check the changes done on bug 348788 as well.
Sorry, I forgot to mention that I committed the changes to the mysql overlay[1]. Thanks to Jeremy for calling my attention. [1] - http://git.overlays.gentoo.org/gitweb/?p=proj/mysql.git;a=summary
were these changes already forwarded to gx86? If not, would explain bug #388869
(In reply to comment #7) > were these changes already forwarded to gx86? If not, would explain bug > #388869 No, they weren't. I was waiting for some testing before moving them to gentoo-x86.
-I/usr//include '--with-plugindir=/usr/lib/mysql/plugin' libtool: link: x86_64-apple-darwin11-gcc -O3 -march=nocona -pipe -DHAVE_ERRNO_AS_DEFINE=1 -fno-strict-aliasing -D_P1003_1B_VISIBLE -DSIGNAL_WITH_VIO_CLOSE -DSIGNALS_DONT_BREAK_READ -DIGNORE_SIGHUP_SIGQUIT -DDONT_DECLARE_CXA_PURE_VIRTUAL -Wl,-dead_strip_dylibs -o mi_test3 mi_test3.o /Library/Gentoo/usr/lib/gcc/x86_64-apple-darwin11/4.2.1/libstdc++.dylib ./.libs/libmyisam.a ../../mysys/.libs/libmysys.a ../../dbug/.libs/libdbug.a ../../strings/.libs/libmystrings.a -L/usr//lib -lz -lm * die "Aborting due to QA concerns: there are files installed outside the prefix" I guess that means no. I'll check what (if any) was done. I checked the git.o.g.o/mysql.git versions.
ok, mysql.eclass has no prefix changes (prefix overlay has)
local screwup, didn't test the mysql-overlay eclasses. Emerge fine when using them. Not sure MY_DATADIR is set correctly.
jmbsvicetto: please check http://overlays.gentoo.org/proj/alt/changeset/60167 and mind http://overlays.gentoo.org/proj/alt/changeset/60168 If you prefer, I'll give you a diff against git HEAD in mysql-overlay.
I also needed http://overlays.gentoo.org/proj/alt/changeset/60169 I'll diff against mysql-overlay
Created attachment 291977 [details, diff] prefix fixes for mysql eclasses These patches are tested on mysql-5.1.59 with Darwin and Solaris hosts. This included running emerge --config step to successfully initialise the database.
Created attachment 292037 [details, diff] prefix fixes for mysql eclasses (v2) I forgot to make the general libname.dylib symlink, updated patches
(In reply to comment #12) > and mind > > http://overlays.gentoo.org/proj/alt/changeset/60168 This was tackled (slightly different) in bug #388673
@grobian: Your patch is backwards, and the first set was already applied to the mysql overlay. Can you please rediff against that to show what is needed still?
Created attachment 296349 [details, diff] prefix fixes for mysql eclasses (v3) (In reply to comment #17) > @grobian: > Your patch is backwards, and the first set was already applied to the mysql > overlay. Can you please rediff against that to show what is needed still? My git diff was not reversed :) I've been adding prefixes where necessary. So, I "undo" your first set, to get them properly fixed. (e.g MY_LOCALSTATEDIR has EPREFIX in it) Here is the updated patchset against the current tip of default (or how that is called in git land).
I've moved the eclasses to the tree, so this bug can be closed as fixed. As the bug is assigned to prefix I'm not going to close it, unless I get an OK from you.
I assume fixed, since I tested it myself a lot when I created the patches.