Summary: | CFLAGs -O0 is ignored by some packages, for those packages -O2 or -O3 was used; I have reason to believe that by negating -O entirely I found a bug at one point | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Trevor A. Thompson <trevorat> |
Component: | Current packages | Assignee: | Gentoo Linux bug wranglers <bug-wranglers> |
Status: | RESOLVED CANTFIX | ||
Severity: | normal | CC: | vapier |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
URL: | http://gcc.gnu.org/onlinedocs/gcc-4.1.2/gcc/Optimize-Options.html#Optimize-Options | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Trevor A. Thompson
2007-09-10 03:24:16 UTC
not a portage bug ... you'll probably have to file individual bugs for each package rather than talking in broad terms as this will need to be fixed for each package if a package has timing issues that crop up when using -O0 rather than -O2, that package is a pile of crap and has race conditions in its code that need to be fixed, not ignored. anyone who tells you differently probably has no idea what's going on. @vapier - feel free to fix the mplayer "pile of crap" e.g.. Until then, maintainers will just kill any bugs about "ignored CFLAGS" and the ebuild will stay as is. Closing this bug because it can't be assigned. why dont you read my comment again instead of making up your own i made no such statement that said "any package forcing optimizations is broken" packages which do force CFLAGS should have a local USE flag 'custom-cflags' like mplayer already does Just two technical remarks which are widely unknown: 1. Using no optimization does not necessarily produce the most stable code. In fact, it can produce broken code, see bug 146300 or bug 187946. 2. The -O? flags enable also optimizations which cannot be enabled individually (or at least not individually in exactly the same manner). An example are the -falign-options which can be only used to tune *some* of the processor-specific alignment defaults set by -O2. both bugs you cite are with an older hardened compiler ... -O0 there is not even close to the same as -O0 on a vanilla compiler |