> Only concern is FTP, which I think pretends stat is available, but I'm
> not sure what data it returns is actually reliable. (I know it does bad
> things with mode bits, for example.)
That's a really good point: if it says things that are really directories are files, we could end up not being able to merge from ftp branches, which would certainly suck - previously we would have just tried it. Similarly for http for that matter.
Perhaps we should be specifically checking "is the type known and is it not a directory" and then we can allow some transports to give noncommittal answers.
Given the only report of this bug has gone away perhaps we should reject this.
> Only concern is FTP, which I think pretends stat is available, but I'm
> not sure what data it returns is actually reliable. (I know it does bad
> things with mode bits, for example.)
That's a really good point: if it says things that are really directories are files, we could end up not being able to merge from ftp branches, which would certainly suck - previously we would have just tried it. Similarly for http for that matter.
Perhaps we should be specifically checking "is the type known and is it not a directory" and then we can allow some transports to give noncommittal answers.
Given the only report of this bug has gone away perhaps we should reject this.