Refactor with fixes for msys-git.

Refactor with fixes for msys-git.

- Check for -h|--help before anything is modified.
- Discard stderr from 'git submodule update'.
- Be careful not to overwrite script while it's executing
  on win.
- Run hooks by invoking 'git runhooks'.
- Fix --dry-run option.

R=iannucci@chromium.org
CC=jar@chromium.org,scottmg@chromium.org

BUG=


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=189926

[szager1, 2013-03-19 21:51:54]

This should fix some of the lingering issues with the submodule flow on windows,
including the 'script disappears when updated by git' problem.

[iannucci, 2013-03-19 22:11:52]

https://codereview.chromium.org/12865010/diff/2001/crup-runner.sh
File crup-runner.sh (right):

https://codereview.chromium.org/12865010/diff/2001/crup-runner.sh#newcode116
crup-runner.sh:116: ( grep -v '^Skipping submodule' || true ) | sed "s|^|[$1]
|g" 2>/dev/null
Can this lead to silent failures?

https://codereview.chromium.org/12865010/diff/2001/git-crup
File git-crup (right):

https://codereview.chromium.org/12865010/diff/2001/git-crup#newcode92
git-crup:92: if [ "${kernel_name:0:5}" = "MINGW" -o "${kernel_name:0:6}" =
"CYGWIN" ]; then
Why not do `which git` and fall back to git.exe?

https://codereview.chromium.org/12865010/diff/2001/git-crup#newcode112
git-crup:112: GIT_CRUP_REINVOKE=1 exec bash "$0" $orig_args
Why not just do the windows-path on all platforms? Less likely to have a
platform specific breakage and it can't hurt anything...

[szager1, 2013-03-20 21:47:01]

https://codereview.chromium.org/12865010/diff/2001/crup-runner.sh
File crup-runner.sh (right):

https://codereview.chromium.org/12865010/diff/2001/crup-runner.sh#newcode116
crup-runner.sh:116: ( grep -v '^Skipping submodule' || true ) | sed "s|^|[$1]
|g" 2>/dev/null
On 2013/03/19 22:11:52, iannucci wrote:
> Can this lead to silent failures?

The lines below that check $? should prevent completely silent failures.  But
there's no way to separate progress messages (which print to stderr) from error
messages, and the progress messages get hopelessly scrambled on the terminal.

https://codereview.chromium.org/12865010/diff/2001/git-crup
File git-crup (right):

https://codereview.chromium.org/12865010/diff/2001/git-crup#newcode92
git-crup:92: if [ "${kernel_name:0:5}" = "MINGW" -o "${kernel_name:0:6}" =
"CYGWIN" ]; then
On 2013/03/19 22:11:52, iannucci wrote:
> Why not do `which git` and fall back to git.exe?

On further examination, I don't think GIT_EXE is even necessary.  I was
concerned before that running 'git' on Windows would go through the whole
wrapper stack, but that doesn't appear to be the case.  So, I have just
s/GIT_EXE/git/ everywhere.

https://codereview.chromium.org/12865010/diff/2001/git-crup#newcode112
git-crup:112: GIT_CRUP_REINVOKE=1 exec bash "$0" $orig_args
On 2013/03/19 22:11:52, iannucci wrote:
> Why not just do the windows-path on all platforms? Less likely to have a
> platform specific breakage and it can't hurt anything...

I separated out copying off upate_depot_tools.bat to this CL:

https://codereview.chromium.org/12755033/

This script will now run from a temporary bash script on all platforms.

[szager1, 2013-03-20 23:05:49]

ping

[iannucci, 2013-03-20 23:41:19]

https://codereview.chromium.org/12865010/diff/7001/git-crup
File git-crup (right):

https://codereview.chromium.org/12865010/diff/7001/git-crup#newcode93
git-crup:93: # runs.  So, we re-exec from a temporary script.
Are we absolutely sure that this happens? Doesn't bash read+buffer the script
contents (even on windows)?

[szager1, 2013-03-21 00:06:08]

https://codereview.chromium.org/12865010/diff/7001/git-crup
File git-crup (right):

https://codereview.chromium.org/12865010/diff/7001/git-crup#newcode93
git-crup:93: # runs.  So, we re-exec from a temporary script.
On 2013/03/20 23:41:19, iannucci wrote:
> Are we absolutely sure that this happens? Doesn't bash read+buffer the script
> contents (even on windows)?

By golly, you're right!  I'll nuke this code.

[iannucci, 2013-03-22 22:19:00]

lgtm (assuming you tested on windows/non-windows?)

[szager1, 2013-03-22 22:24:08]

On 2013/03/22 22:19:00, iannucci wrote:
> lgtm (assuming you tested on windows/non-windows?)

Yes!

[commit-bot: I haz the power, 2013-03-22 22:24:15]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/szager@chromium.org/12865010/13001

[commit-bot: I haz the power, 2013-03-22 22:26:05]

Change committed as 189926