From: Jim Cromie <jim.cromie@gmail.com>
Date: Mon, 12 Jul 2010 19:56:20 GMT
Subject: Re: skipping commits via commit-msg contents
Message-ID: <AANLkTikVNpFPBMO_SDIfn-5np42ALI-qGAHTOvlmONxo@mail.gmail.com>
In-Reply-To: <20100712190203.GA9365@dert.cs.uchicago.edu>

thanks guys,

Ram, not what I want, but good to know..

Jonathan, replace feels like a big hammer for a newbie..
I havent published these, since they need bisection amongst other reasons :-O
and the BUGS also require more -fu to understand than I possess


On Mon, Jul 12, 2010 at 1:02 PM, Jonathan Nieder <jrnieder@gmail.com> wrote:
> Hi Jim,
>
> Jim Cromie wrote:
>
>> if git bisect were to recognize  --skip-bisect  in the subject line
>> (or in commit-message somewhere, say top or bottom),
>> then bisection could proceed silently past such commits.
>
> In addition to Ram’s suggestion, you might want to look into
> ‘git replace’[1].  It can be useful when the broken commits
> have already been published.  It works like this:
>
>  git replace <bad commit> <bad commit>^
>
> and then ‘git bisect’ and lower-level commands like git show and
> checkout will silently substitute the parent of the broken commit
> when ever you refer to it.
>
> You can publish the resulting “replace refs” in the refs/replace/*
> namespace and anyone who explicitly chooses to fetch them will be
> able to see the same effect.
>
> Two problems:
>
>  - bisect skip is a bit more sophisticated (read: better) than just
>   substituting a parent, especially when the commit to be skipped
>   is a merge.  So it might still make sense to teach bisect to
>   respect a refs/notes/skip-bisect note that requests for it
>   to skip a specific ref.

This sounds nice.
Notionally, extending it further would be cool -

[jimc@groucho perl-git]$ git notes show 2de9db42d3e578becacbd237bf4f59b432cc82f0
good make test

with this note, Ive told myself that it passes (good), and what it
passes (make test)

I could imagine bisect knowing enough to take that as an implicit start,
unless overridden on a command line, or by a similar note on a later commit.

a note like:
skip-bisect until <tag|commit|branch>
could be cool too, though there may be deeper problems with skipping
entire subranges.

thanks again,
Jim