Make sure branch-heads is configured whenever requested.

Make sure branch-heads is configured whenever requested.

This allows the branch-heads to be pulled even if not configured on the
original clone.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=194382

[M-A Ruel, 2013-04-12 20:28:12]

https://codereview.chromium.org/13951009/diff/2001/gclient_scm.py
File gclient_scm.py (right):

https://codereview.chromium.org/13951009/diff/2001/gclient_scm.py#newcode855
gclient_scm.py:855: """Add, and optionally fetch, the "branch-heads" refspecs if
requested."""
Adds

https://codereview.chromium.org/13951009/diff/2001/gclient_scm.py#newcode860
gclient_scm.py:860: '+refs/branch-heads/*:refs/remotes/branch-heads/*',
I still don't know what branch-heads is supposed to mean. It's a meta-way of
keeping multiple branches? Why not tags?

[Michael Moss, 2013-04-12 20:58:36]

https://codereview.chromium.org/13951009/diff/2001/gclient_scm.py
File gclient_scm.py (right):

https://codereview.chromium.org/13951009/diff/2001/gclient_scm.py#newcode855
gclient_scm.py:855: """Add, and optionally fetch, the "branch-heads" refspecs if
requested."""
On 2013/04/12 20:28:12, Marc-Antoine Ruel wrote:
> Adds

Done.

https://codereview.chromium.org/13951009/diff/2001/gclient_scm.py#newcode860
gclient_scm.py:860: '+refs/branch-heads/*:refs/remotes/branch-heads/*',
On 2013/04/12 20:28:12, Marc-Antoine Ruel wrote:
> I still don't know what branch-heads is supposed to mean. It's a meta-way of
> keeping multiple branches? Why not tags?

branch-heads is specifically the heads of branches that are mirrored from svn
repositories. It's named that way to help distinguish them from "native" git
branches, and to help isolate them for people who don't ever need them, and thus
don't want to sync that extra data (i.e. only 'refs/heads/*' is fetched by
default).

[M-A Ruel, 2013-04-16 14:44:09]

https://codereview.chromium.org/13951009/diff/2001/gclient_scm.py
File gclient_scm.py (right):

https://codereview.chromium.org/13951009/diff/2001/gclient_scm.py#newcode860
gclient_scm.py:860: '+refs/branch-heads/*:refs/remotes/branch-heads/*',
On 2013/04/12 20:58:36, Michael Moss wrote:
> On 2013/04/12 20:28:12, Marc-Antoine Ruel wrote:
> > I still don't know what branch-heads is supposed to mean. It's a meta-way of
> > keeping multiple branches? Why not tags?
> 
> branch-heads is specifically the heads of branches that are mirrored from svn
> repositories. It's named that way to help distinguish them from "native" git
> branches, and to help isolate them for people who don't ever need them, and
thus
> don't want to sync that extra data (i.e. only 'refs/heads/*' is fetched by
> default).

What you are describing is kinda what lightweight tags are for. So I just want
to know what advantages the custom branch-heads has over using git native
lightweight tags and using "git fetch --tags".

[Michael Moss, 2013-04-16 16:02:17]

> What you are describing is kinda what lightweight tags are for. So I just want
> to know what advantages the custom branch-heads has over using git native
> lightweight tags and using "git fetch --tags".

No technical advantage really, more just following the traditional sense of tags
vs. branches, where tags are static, but branches follow an active line of
development, which these do. Also, I don't want to confuse them with mirrored
svn tags, since that's not what they represent (and there could be naming
collisions if we did). Not a big deal either way, but I think tags make less
sense semantically, and frankly, it's been like this for a while now and doesn't
seem worth the effort to change for little or no impact (especially since I can
no longer access the mirroring host to quickly modify the repo configs).

[M-A Ruel, 2013-04-16 16:07:38]

thanks for the explanation, lgtm

https://codereview.chromium.org/13951009/diff/5001/gclient_scm.py
File gclient_scm.py (right):

https://codereview.chromium.org/13951009/diff/5001/gclient_scm.py#newcode871
gclient_scm.py:871: print('Retrying...')
I'd prefer it to time.sleep(5) between retries.

[Michael Moss, 2013-04-16 16:20:18]

https://codereview.chromium.org/13951009/diff/5001/gclient_scm.py
File gclient_scm.py (right):

https://codereview.chromium.org/13951009/diff/5001/gclient_scm.py#newcode871
gclient_scm.py:871: print('Retrying...')
On 2013/04/16 16:07:38, Marc-Antoine Ruel wrote:
> I'd prefer it to time.sleep(5) between retries.

Done.

[Michael Moss, 2013-04-16 17:28:59]

Committed patchset #3 manually as r194382 (presubmit successful).