Make gclient cleanup do a 'correcter' thing for git

Make gclient cleanup a no-op for git

Patch-contributed-by: Nasser Grainawi <nasser@codeaurora.org>;

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=37021

[Nasser Grainawi, 2010-01-14 04:00:39]

This is based off http://codereview.chromium.org/546045

[M-A Ruel, 2010-01-14 15:14:26]

I don't understand why you remove fsck?

[Nasser Grainawi, 2010-01-18 16:43:52]

git fsck takes over 6 minutes to run on large projects (like the Linux kernel).
Maybe cleanup is supposed to run for that amount of time?

Also, should 'gclient cleanup' do a 'git clean'? Given the nested repo structure
of cr-os, it's a potentially dangerous command to run if the .gitignore files
aren't up to date. But it's also incredibly useful for getting back to a clean
slate.

[M-A Ruel, 2010-01-18 21:53:43]

On 2010/01/18 16:43:52, Nasser Grainawi wrote:
> git fsck takes over 6 minutes to run on large projects (like the Linux
kernel).
> Maybe cleanup is supposed to run for that amount of time?
> 
> Also, should 'gclient cleanup' do a 'git clean'? Given the nested repo
structure
> of cr-os, it's a potentially dangerous command to run if the .gitignore files
> aren't up to date. But it's also incredibly useful for getting back to a clean
> slate.

Some context:
In svn, when a svn update is interupted, the svn lock is left held and the next
svn update will fail until a svn cleanup is done.

With git, the problem is not as severe and probably not as necessary (if
necessary at all)

So cleanup() != git clean. cleanup() should just do whatever it takes to resume
an interrupted git command with kill -9.

I'm not very intimate with git, fsck is probably not necessary (?)

[Nasser Grainawi, 2010-01-19 17:38:17]

On 2010/01/18 21:53:43, Marc-Antoine Ruel wrote:
> On 2010/01/18 16:43:52, Nasser Grainawi wrote:
> > git fsck takes over 6 minutes to run on large projects (like the Linux
> kernel).
> > Maybe cleanup is supposed to run for that amount of time?
> > 
> > Also, should 'gclient cleanup' do a 'git clean'? Given the nested repo
> structure
> > of cr-os, it's a potentially dangerous command to run if the .gitignore
files
> > aren't up to date. But it's also incredibly useful for getting back to a
clean
> > slate.
> 
> Some context:
> In svn, when a svn update is interupted, the svn lock is left held and the
next
> svn update will fail until a svn cleanup is done.
> 
> With git, the problem is not as severe and probably not as necessary (if
> necessary at all)
> 
> So cleanup() != git clean. cleanup() should just do whatever it takes to
resume
> an interrupted git command with kill -9.
> 
> I'm not very intimate with git, fsck is probably not necessary (?)

Thanks for the context, it helps quite a bit. I don't believe there's an
equivalent
situation in git. Interrupting most git commands has no ill effect, but for
those where
it does leave you in an imperfect state, there's no one magic command to fix it.
:-/

Should we make cleanup a no-op for git repos? The gc --auto is invoked by
various
commands anyway, so it shouldn't be necessary to manually invoke it ever.

[M-A Ruel, 2010-01-22 02:01:24]

On 2010/01/19 17:38:17, Nasser Grainawi wrote:
> Should we make cleanup a no-op for git repos? The gc --auto is invoked by
> various
> commands anyway, so it shouldn't be necessary to manually invoke it ever.

Yes. I agree with a no-op.

[M-A Ruel, 2010-01-22 18:16:19]

lgtm

[Mandeep Singh Baines, 2010-01-25 17:47:25]

On 2010/01/14 03:54:15, Nasser Grainawi wrote:
> 

Applied. Thanks!