Hi Junio,
On Sun, 21 Oct 2018, Junio C Hamano wrote:
> Alban Gruin <alban.gruin@xxxxxxxxx> writes:
>
> > The error comes from the call to `git stash apply $stash_id' in
> > builtin/rebase.c:261. When $stash_id only contains decimals and no
> > letters, git-stash tries to apply stash@{$stash_id}[0][1]. Thas was not
> > a real problem with the shell script, because it did not abbreviate the
> > object id of the stashed commit, so it was very unlikely that the oid
> > would contain only digits. builtin/rebase.c shortens the oid[2], making
> > this problem more likely to occur.
>
> OK, so make "rebase in C" a faithful conversion of the original, the
> code that lead to builtin/rebase.c:261 must be fixed not to pass a
> shortened oid. I suspect that internally it has a full object name,
> so not shortening would be the right thing anyway, so regaredless of
> this bug, it probably makes sense to do the fix.
>
> But as you said, even the scripted version that passed the full
> object name had a (10/16^40) chance of using a 40-hex that only has
> [0-9] and caused the same breakage, so such a change to "rebase in
> C" is sweeping the age old bug under the same rug, in the context of
> discussing this particular bug.
>
> I agree with you that it is a better fix to the problem to allow the
> caller to say "I am giving an oid---it may (or may not---I do not
> even bother to check) consist of only digits, but do not treat it as
> an index to the stash reflog."
We already have a way to say "I am giving you an oid": append `^0` to the
hash.
I implemented a fix, pushed it up to GitGitGadget, and will submit it
tomorrow (after a fresh look, and after the build finished):
https://github.com/gitgitgadget/git/pull/52
Ciao,
Dscho
