Fix webkit build failure reported by evan@

Fix webkit build failure reported by evan@

When fetching svn URL via 'git svn info --url' command
git-svn.perl script may print its progress log to stdout instead of stderr: as a result we end up using all this junk as svn repository URL.
Fix it by filtering output.

BUG=None
TEST=None

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=76803

[Denis Lagno, 2011-03-03 19:33:25]

please try and take a look

[Evan Martin, 2011-03-03 19:49:31]

Can you make the review description describe the change?  (Someone looking at
the svn history won't know what your change is doing)

http://codereview.chromium.org/6603022/diff/6001/build/util/lastchange.py
File build/util/lastchange.py (right):

http://codereview.chromium.org/6603022/diff/6001/build/util/lastchange.py#new...
build/util/lastchange.py:138: match = re.search('^.*://.*$', output, re.M)
Maybe use
  r'^\w+://.*$'
just to make it harder to accidentally match.  (What you have searches for any
line that has a "://" in it...)

http://codereview.chromium.org/6603022/diff/6001/build/util/lastchange.py#new...
build/util/lastchange.py:146: Fetch root of SVN repository bound to git.
This part of the change seems unrelated to the bug (?)

[Denis Lagno, 2011-03-03 20:18:39]

http://codereview.chromium.org/6603022/diff/6001/build/util/lastchange.py
File build/util/lastchange.py (right):

http://codereview.chromium.org/6603022/diff/6001/build/util/lastchange.py#new...
build/util/lastchange.py:138: match = re.search('^.*://.*$', output, re.M)
On 2011/03/03 19:49:31, Evan Martin wrote:
> Maybe use
>   r'^\w+://.*$'
> just to make it harder to accidentally match.  (What you have searches for any
> line that has a "://" in it...)

Done.

should I use those 'raw string literals'?  I do not observe
any difference during execution.

http://codereview.chromium.org/6603022/diff/6001/build/util/lastchange.py#new...
build/util/lastchange.py:146: Fetch root of SVN repository bound to git.
On 2011/03/03 19:49:31, Evan Martin wrote:
> This part of the change seems unrelated to the bug (?)

yes.  But only now I realized how URL and Root are used.
Normally root is expected to be a prefix of URL.
Should I send it as a separate patch?

[Evan Martin, 2011-03-03 20:26:46]

On 2011/03/03 20:18:39, Denis Lagno wrote:
> > Maybe use
> >   r'^\w+://.*$'
> > just to make it harder to accidentally match.  (What you have searches for
any
> > line that has a "://" in it...)
> 
> Done.
> 
> should I use those 'raw string literals'?  I do not observe
> any difference during execution.

It prevents \ from being interpreted in the string.  I guess it's just good
style.

> build/util/lastchange.py:146: Fetch root of SVN repository bound to git.
> On 2011/03/03 19:49:31, Evan Martin wrote:
> > This part of the change seems unrelated to the bug (?)
> 
> yes.  But only now I realized how URL and Root are used.
> Normally root is expected to be a prefix of URL.
> Should I send it as a separate patch?

I'd prefer that.  Other people encountered this build breakage, and it was
suggested we revert your previous, but I figured you could fix it quickly.  It
would be nice to get to a known-compiling state before adding more logic so if
this new logic causes new regressions we can revert to a known-good state.

[Denis Lagno, 2011-03-03 20:32:22]

On 2011/03/03 20:26:46, Evan Martin wrote:
> > Should I send it as a separate patch?
> 
> I'd prefer that.  Other people encountered this build breakage, and it was
> suggested we revert your previous, but I figured you could fix it quickly.  It
> would be nice to get to a known-compiling state before adding more logic so if
> this new logic causes new regressions we can revert to a known-good state.

Ok, extras removed.

[Evan Martin, 2011-03-03 20:50:16]

LGTM