[BUG?] retrying with "am -3" doesn't work anymore

[BUG?] retrying with "am -3" doesn't work anymore

[Jeff King]

I noticed that I could not get a patch from Junio to apply earlier
today, and I think it is a regression in the builtin git-am
implementation.  I had trouble reproducing with a basic test case,
though.

Basically, I picked up the three patches from this sub-thread:

  http://thread.gmane.org/gmane.comp.version-control.git/288987/focus=290222

and tried to apply them on top of v2.8.0.

Doing it with "git am -3 patches" works. But doing it with:

  git am patches
  git am -3

doesn't. Bisecting shows that it did work before 783d7e8 (builtin-am:
remove redirection to git-am.sh, 2015-08-04).

I tried to tweak t4150 to show this by replacing a "git am -3 foo" with
"test_must_fail git am foo && git am -3", but it failed even on older
git, which made me think I fumbled the test somehow.

-Peff

Re: [BUG?] retrying with "am -3" doesn't work anymore

[Paul Tan]

Hi Jeff,

On Wed, Mar 30, 2016 at 10:15 AM, Jeff King <peff@peff.net> wrote:
> I noticed that I could not get a patch from Junio to apply earlier
> today, and I think it is a regression in the builtin git-am
> implementation.  I had trouble reproducing with a basic test case,
> though.
>
> Basically, I picked up the three patches from this sub-thread:
>
>   http://thread.gmane.org/gmane.comp.version-control.git/288987/focus=290222
>
> and tried to apply them on top of v2.8.0.
>
> Doing it with "git am -3 patches" works. But doing it with:
>
>   git am patches
>   git am -3
>
> doesn't. Bisecting shows that it did work before 783d7e8 (builtin-am:
> remove redirection to git-am.sh, 2015-08-04).

Yeah, with "git am -3" when resuming, the -3 will only affect the
current conflicting patch since 852a171 (am: let command-line options
override saved options, 2015-08-04).

Regards,
Paul

Re: [BUG?] retrying with "am -3" doesn't work anymore

[Jeff King]

On Wed, Mar 30, 2016 at 12:18:30PM +0800, Paul Tan wrote:

> On Wed, Mar 30, 2016 at 10:15 AM, Jeff King <peff@peff.net> wrote:
> > I noticed that I could not get a patch from Junio to apply earlier
> > today, and I think it is a regression in the builtin git-am
> > implementation.  I had trouble reproducing with a basic test case,
> > though.
> >
> > Basically, I picked up the three patches from this sub-thread:
> >
> >   http://thread.gmane.org/gmane.comp.version-control.git/288987/focus=290222
> >
> > and tried to apply them on top of v2.8.0.
> >
> > Doing it with "git am -3 patches" works. But doing it with:
> >
> >   git am patches
> >   git am -3
> >
> > doesn't. Bisecting shows that it did work before 783d7e8 (builtin-am:
> > remove redirection to git-am.sh, 2015-08-04).
> 
> Yeah, with "git am -3" when resuming, the -3 will only affect the
> current conflicting patch since 852a171 (am: let command-line options
> override saved options, 2015-08-04).

Ah, right. I had a nagging feeling that we had discussed this, and
indeed, I already participated in the discussion last July[1]. I even
apparently argued in favor of the new behavior[2]. Yikes. There goes my
mind.

Running "git am -3" for each patch which needs it does indeed
successfully apply the series.

Thanks for a quick response, and sorry for the noise.

-Peff

[1] http://thread.gmane.org/gmane.comp.version-control.git/274574

[2] In http://thread.gmane.org/gmane.comp.version-control.git/274574/focus=274635,
    though I think there I more meant that in:

       git am
       git am -3
       git am

    the third one would not use "-3" again. So I was mostly confused
    that in:

       git am
       git am -3

    we would not use "-3" for the subsequent patches applied by that
    second invocation. So maybe that is a bug. I dunno. I could see
    arguments either way.