A build failure with "Active GCC too old! Must have at least GCC4.3" is a bit late in my eyes. How about a "!<sys-devel/gcc-4.3" blocker, instead?
The blocker shouldn't be used since you should be able to have older versions of gcc installed at the same time.
Good point. I guess we cannot require GCC as people may be using a different compiler? Anything we could that dependency make conditional of? Too bad there is no "gcc" use flag. Other ideas?
inherit toolchain src_setup() { if tc_version_is_at_least 4.3 || die "Active GCC too old! Must have at least GCC4.3" } Or is that too ugly? I think we do not really need to think about requiring gcc, as IIRC portage does not really handle anything else especially good.
Looking at the ebuild I find the line DEPEND="|| ( >=sys-devel/gcc-4 >=sys-devel/gcc-apple-4 )" If that's in there anyway why not bump it to gcc-4.3? Btw why is GCC 4.3 required anyway?
(In reply to comment #3) > src_setup() { ... > } The existing check is ALREADY in a pkg_setup block in the eclass. We can't have it any earlier. (In reply to comment #4) > DEPEND="|| ( >=sys-devel/gcc-4 >=sys-devel/gcc-apple-4 )" > If that's in there anyway why not bump it to gcc-4.3? > Btw why is GCC 4.3 required anyway? I've just bumped the DEPEND, but that doesn't help the check, as the ACTIVE gcc still needs to be >=4.3. C++ template use is why the newer GCC is required. 4.2 can work depending on which parts of the server you omit, but 4.3 is needed in general.
So the check was good, and the DEPEND now breaks HPPA.
DEPEND.bad 11 dev-db/mysql/mysql-5.0.83.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.0.84.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.0.84-r1.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.0.87.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.0.88.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.0.90.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.0.90-r1.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.0.90-r2.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.1.39-r1.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.1.44.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3'] dev-db/mysql/mysql-5.1.44-r1.ebuild: ~hppa(default/linux/hppa/10.0) ['>=sys-devel/gcc-4.3']
if it's checked at configure time then that's early enough.
HPPA: What should we do regarding you on the MySQL ebuilds? Either we trace down the parts that actually need 4.3 and "fix" them to work with GCC 4.2, or we see about dropping the ~hppa keyword on the ebuilds that need a newer GCC. What's the holdup for HPPA support on the newer GCC?
(In reply to comment #9) > What's the holdup for HPPA support on the newer GCC? bug #272645
ok, that mask is in place, and i've also removed the ~hppa keywords to make repoman really happy as it was being cranky at me.
Ok, following a discussion with upstream https://lists.launchpad.net/maria-discuss/msg00116.html the actual code that was broken in 5.0.8[3-6] with older GCC versions was replaced in 5.0.87, and I have confirmed that it compiles even with hardened gcc 3.4.6, so I'm closing this as invalid now.
I removed the package.mask entries in arch/hppa profile.
