diff options
Diffstat (limited to 'doc/bugs/cannot_change_locale___40__en__95__US.UTF-8__41__.mdwn')
| -rw-r--r-- | doc/bugs/cannot_change_locale___40__en__95__US.UTF-8__41__.mdwn | 19 |
1 files changed, 19 insertions, 0 deletions
diff --git a/doc/bugs/cannot_change_locale___40__en__95__US.UTF-8__41__.mdwn b/doc/bugs/cannot_change_locale___40__en__95__US.UTF-8__41__.mdwn index 178348e51..6be8bd6e6 100644 --- a/doc/bugs/cannot_change_locale___40__en__95__US.UTF-8__41__.mdwn +++ b/doc/bugs/cannot_change_locale___40__en__95__US.UTF-8__41__.mdwn @@ -26,3 +26,22 @@ he_IL he_IL.iso88598 he_IL.utf8 POSIX + +> I've made LOCPATH not be set except when git-annex is built with ghc +> older than 7.10, since the problem was fixed in ghc 7.10. +> +> Also, I loved the LOCPATH setting into the linker shim script, rather +> than in runshell, so it will only affect the programs bundled with +> git-annex (itself and git and a few other things). Which are not +> localized anyway in the bundle. So, even in builds where it's still set +> (the linux ancient build in particular), things done in the runshell +> environment won't be affected. +> +> I do wonder if there could be problems with incompatabilities between the +> bundled glibc and the system locale files, which might be for a +> newer/older libc version. Not so much random `.mo` files, which seem +> quite portable across glibc versions, but the more core locale files. +> If that turns out to be a problem, LOCPATH might have to be turned back +> on. +> +> For now, [[done]] --[[Joey]] |