diff options
| author | Ryan Hendrickson | 2015-10-27 16:39:42 -0400 |
|---|---|---|
| committer | Mike McQuaid | 2015-11-01 09:30:00 -0800 |
| commit | 91dd4d56d426ed47c4e042f7b17d6560245fd60f (patch) | |
| tree | ec48d765cf648e0cd7c6ab7e8122e9bf64237440 /Library | |
| parent | 76435ece68bc4db7dec7eb99dfc50158ac2ed729 (diff) | |
| download | brew-91dd4d56d426ed47c4e042f7b17d6560245fd60f.tar.bz2 | |
os/mac/hardware: reverse universal_archs entries.
The order ought not to matter, but GCC can fail with -arch i386 -arch
x86_64 (producing an error like "FATAL:Bad fx_size (0x8) in
fix_to_relocation_info()") but succeed with -arch x86_64 -arch i386.
Closes Homebrew/homebrew#45401.
Signed-off-by: Mike McQuaid <mike@mikemcquaid.com>
Diffstat (limited to 'Library')
| -rw-r--r-- | Library/Homebrew/os/mac/hardware.rb | 6 |
1 files changed, 5 insertions, 1 deletions
diff --git a/Library/Homebrew/os/mac/hardware.rb b/Library/Homebrew/os/mac/hardware.rb index 5643416c8..d4ce7a7e2 100644 --- a/Library/Homebrew/os/mac/hardware.rb +++ b/Library/Homebrew/os/mac/hardware.rb @@ -97,7 +97,11 @@ module MacCPUs if MacOS.version <= :leopard && !MacOS.prefer_64_bit? [arch_32_bit].extend ArchitectureListExtension else - [arch_32_bit, arch_64_bit].extend ArchitectureListExtension + # Amazingly, this order (64, then 32) matters. It shouldn't, but it + # does. GCC (some versions? some systems?) can blow up if the other + # order is used. + # http://superuser.com/questions/740563/gcc-4-8-on-macos-fails-depending-on-arch-order + [arch_64_bit, arch_32_bit].extend ArchitectureListExtension end end |