Summary: | >=dev-db/mysql-5.0.83 does not build with <sys-devel/gcc-4 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Emmanuel Andry <eandry> |
Component: | Hardened | Assignee: | Gentoo Linux MySQL bugs team <mysql-bugs> |
Status: | RESOLVED FIXED | ||
Severity: | major | CC: | hardened, prefix, selinux |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | x86 | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Emmanuel Andry
2009-09-14 15:18:20 UTC
Seen this some months ago. It's some defines they changed up in mysql 5.1.x series, looks like they've made their way back to mysql-5.0.x series. Not sure if simply changing up the defines is safe or good enough though and didn't bother to test. If I recall I just installed ~arch hardened gcc and built it with that instead. hardened/gengor: Even with the defines changed, there's some other code I don't trust in newer versions of MySQL with GCC3 anymore. How close is hardened to having a stable GCC4? I can state that the hardened GCC4 runs fine with MySQL, I'm using it on some infra production boxes. (In reply to comment #2) > hardened/gengor: > Even with the defines changed, there's some other code I don't trust in newer > versions of MySQL with GCC3 anymore. > Yeah, that's what I recall seeing too. Didn't bother to try to figure out a gcc-3.4-safe patch for it and also just used ~arch hardened gcc-4.3.x. I doubt you're interested in such a venture either. We're pretty close to stable-marked gcc-4.x on hardened. Not sure exactly when, but think days and weeks, not months. Assuming there's not a patch (and quick), the ebuilds should probably be adjusted to die gracefully if <gcc-4 is in use and inform the user to use a newer gcc. If you're planning to stabilize >=dev-db/mysql{,-community}-5.0.83, please package.mask on hardened so stable users don't get bothered. Otherwise it can remain unmasked and hardened users trying to compile with gcc-3.4.x will just hit the error. gengor: I wouldn't be surprised to see a sec vuln soon for MySQL (again, a timescale of weeks), which will necessitate stabilization, and be hard to solve on hardened if masked. What's still blocking gcc-4 from being stable for hardened? deps updated. (In reply to comment #5) > deps updated. > Not appropriate. sys-devel/gcc-apple works fine too. So, this change to protect hardened users just broke osx users. re-fixed for prefix now. Also, selinux needs to update their hardened profiles to stop masking gcc-4 ASAP. Just a long-term followup on this: https://lists.launchpad.net/maria-discuss/msg00116.html Followup on bug #284946, #293801, #310615, #307251: In 5.0.87, the mutex locking code was changed a lot again, so that while 5.0.84 required a newer GCC due to the HAVE_ATOMIC_BUILTINS error, newer releases do NOT need those GCC versions anymore. This also frees up ~hppa to do more stabilization. |