Make git try warn on a dirty index.

Make git try warn on a dirty index.

Many, many times this has burned me badly; if you have a dirty index,
git try will just upload your commits, and not the index.

R=nsylvain@chromium.org
BUG=None

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=151202

[gavinp, 2012-08-09 18:07:04]

nsylvain, wdyt?

[nsylvain, 2012-08-10 17:08:44]

Forwarding the cmp or szager.

[cmp, 2012-08-10 23:51:50]

lgtm with nits

http://codereview.chromium.org/10836180/diff/1/trychange.py
File trychange.py (right):

http://codereview.chromium.org/10836180/diff/1/trychange.py#newcode67
trychange.py:67: 
nit: insert an empty line before line 67

http://codereview.chromium.org/10836180/diff/1/trychange.py#newcode82
trychange.py:82: 
nit: insert an empty line before line 82

http://codereview.chromium.org/10836180/diff/1/trychange.py#newcode291
trychange.py:291: # TODO(gavinp): Can we just include the index?
safe to remove this line, we can't include the index, since our workflow assumes
that devs will be committing their changes to their repos.

[gavinp, 2012-08-12 16:52:16]

Thanks for the review cmp. I've now added unit test coverage, and it passes
presubmit, which was not true of the first upload.

http://codereview.chromium.org/10836180/diff/1/trychange.py
File trychange.py (right):

http://codereview.chromium.org/10836180/diff/1/trychange.py#newcode67
trychange.py:67: 
On 2012/08/10 23:51:50, cmp wrote:
> nit: insert an empty line before line 67

Done.

http://codereview.chromium.org/10836180/diff/1/trychange.py#newcode82
trychange.py:82: 
On 2012/08/10 23:51:50, cmp wrote:
> nit: insert an empty line before line 82

Done.

http://codereview.chromium.org/10836180/diff/1/trychange.py#newcode291
trychange.py:291: # TODO(gavinp): Can we just include the index?
On 2012/08/10 23:51:50, cmp wrote:
> safe to remove this line, we can't include the index, since our workflow
assumes
> that devs will be committing their changes to their repos.

I've removed the comment, and I'll think about this more.

But commit-first is definitely not my ideal workflow for try jobs, although the
tool currently requires this. I frequently have changes that fail on one
platform, and rather than do a local build, I send a try job to see if the
obvious fix worked, and move on to other work. The reason that I made this
change is that I frequently confused myself by running my commits only, without
my index, and I wouldn't understand why my changes had no effect.

Although perhaps the nonce commit makes sense; to change repos I need to commit
anyway, or use the unreliable stash feature. I have been known to redownload my
try diff though, and just commit that later.

So I see both sides, especially now that I realise I need to switch to another
repo anyway while i wait for the try job. Hrm.