On 05/03/2025 13:10, Jonathan Wakely via Gcc wrote:
> While onboarding somebody today we noticed an error in the
> customization script if you use a non-default value for the local
> prefix.
>
> I reproduced it with bash -x to show where it happens. In the output
> below I entered "jw" as the local prefix, instead of the default "me":
>
> + echo 'Setting up tracking for personal namespace redi in remotes/users/jw'
> Setting up tracking for personal namespace redi in remotes/users/jw
> + git config remote.users/jw.url git+ssh://r...@gcc.gnu.org/git/gcc.git
> + '[' x '!=' x ']'
> + git config --replace-all remote.users/jw.fetch
> '+refs/users/redi/heads/*:refs/remotes/users/jw/*'
> refs/users/redi/heads/
> + git config --replace-all remote.users/jw.fetch
> '+refs/users/redi/tags/*:refs/tags/users/jw/*' refs/users/redi/tags/
> + git config --replace-all remote.users/jw.push
> 'refs/heads/jw/*:refs/users/redi/heads/*' refs/users/redi
> + '[' me '!=' jw -a me '!=' origin ']'
> + git config --remove-section remote.me
> fatal: no such section: remote.me
> + git config --unset-all remote.origin.fetch refs/users/redi/
> + git config --unset-all remote.origin.push refs/users/redi/
> + git fetch users/jw
>
> The script finishes successfully, but because the last line that is
> printed out is "fatal: no such section: remote.me" it makes it look
> like it failed and exited. But it's only fatal to the 'git config'
> sub-process, not the script that the user is actually running.
>
> Should we just add a 2>/dev/null redirect to that command, or do we
> want to check if it exists before trying to remove it? i.e.
> if git config get --regexp remote.me >/dev/null; then
You'd want to match ^remote.${old_pfx}, for safety. You could also use 'git
config --get-regexp', but perhaps that's what you mean anyway.
> git config --remove-section remote.me
> fi
>
> Or should we just remove that part entirely?
> I doubt anybody used the original version of that script prior to
> January 2020, and hasn't updated to the new structure yet. See
> r10-6086-g24b178184f260a which introduced that part.
But this might be easier. I agree, that was very much a transitional bit of
code as we settled on our new workflows.
I'll trust your decision on this one.
R.
