Re: safe.directory warnings for root-owned repositories

From: Michael Orlitzky <michael@orlitzky.com>
Date: Oct 31, 2024, 12:15 UTC
Message-ID: <ZyN03D8os53XJGDo@mertle>
In-Reply-To: <D59QZ1P3KSNB.ZXDHUA5HGC97@pm.me>

On 2024-10-31 05:38:00, Caleb White wrote:

Show 9 quoted lines

> 
> The dubious ownership check simply reports that the directory is owned by
> someone other than the user running the command, with no special handling
> for the root user. While the error might not make the most sense in this
> context, I'm not sure that it's worth special-casing the root user
> (really the user with id = 0 as it might not be named `root`) in the
> implementation.
> 
> Why would you initialize a repository as `root` in the first place?

To avoid the dubious ownership warning, obviously :)

These are shared repositories that I and my coworkers push to over SSH. Write access is granted via ACLs, with ownership being mostly irrelevant. (This is still "unsafe," but not for the stated reason.)

I don't necessarily have a problem with adding O(m*n) safe.directory entries, but every once in a while someone will ask me about it, and I don't have a good answer for why it's not safe to push to a repository that's owned by root. I guess it's just more annoying to have to override the warning when the warning is wrong. Though if it was changed to "dubious repository writability," I wouldn't be able to complain any more.