threads / rfc / 30718

[RFC] git bisect old/new

Subject: [RFC] git bisect old/new

## tl;dr

2 messages between Jun 5, 2012 and Jun 5, 2012.

replies: 1people: 2as markdown or json

duperrav@minatec.inpg.fr· Jun 5, 2012, 16:45 UTC · lore
Hi,

We would like to work on modifying git bisect so it can be used easily when looking for a fix instead of a regression. Indeed, when you want to do so you have to mentally switch "good" and "bad" to mark commit which are still unfixed as good and those fixed as bad.

There have been discussions and patchs on how it could be implemented:
- http://thread.gmane.org/gmane.comp.version-control.git/86063:
   introduced git fixed and git unfixed which could be use instead of
   git good and git bad when looking for a fix.
- http://thread.gmane.org/gmane.comp.version-control.git/182398:
   created the option --reverse for git bisect start to switch good and
   bad.

But both solutions did not address the issue of how "git bisect visualize" show the review of the commits because in both cases the refs of the commit are not changed, the older commits are in revs/good and the new ones are in revs/bad. To prevent that, the idea raised in the second thread was to use old/new (or yes/no) and ask the user what he is looking for. With this we could look not only for regression or fix but for whatever property used to be true and isn't anymore without flipping the revs following the cases. We will work on that last solution but we wanted to ask for comments before coding too much for nothing.

Valentin DUPERRAY Franck JONAS Lucien KONG Thomas NGUY Hyuhn Khoi Nguyen NGUYEN

Grenoble INP ENSIMAG
Junio C Hamano· Jun 5, 2012, 17:23 UTC · re: duperrav@minatec.inpg.fr · lore

Re: [RFC] git bisect old/new

duperrav@minatec.inpg.fr writes:
> We will work on that last solution but we wanted to ask for comments
> before coding too much for nothing.
After re-reading the thread that ends at Peff's
    http://thread.gmane.org/gmane.comp.version-control.git/182398/focus=183410
I have nothing to add to the conclusion of the thread.

← back to recent threads