chromite: Switch from execve to import.

chromite: Switch from execve to import.

BUG=None
TEST=Entered shell from various places in the tree, and from inside the chroot.
TEST='chromite workon x86-mario list --all' outside repo and chroot with CROS_WORKON_SRCROOT set 
TEST='chromite workon x86-mario list --all' inside repo and outside chroot with CROS_WORKON_SRCROOT set 
TEST='chromite workon x86-mario list --all' outside repo and chroot with CROS_WORKON_SRCROOT not set 
TEST='chromite workon x86-mario list --all' inside repo and outside chroot with CROS_WORKON_SRCROOT not set 
TEST='chromite workon x86-mario list --all' inside chroot

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=15175cf

[jrbarnette, 2011-02-02 01:02:42]

LGTM!

I've downloaded the patch, and run the tests described here:
    http://codereview.chromium.org/6312068

[diandersAtChromium, 2011-02-03 00:44:15]

Thanks for all of your work on this!

...please address comments and re-upload.  At the moment, this script will break
if the user tries to run it with src/scripts as the CWD.  We also don't want
SRCROOT in the sys.path inside the chroot so that a devs don't accidentally
start relying on code in SRCROOT that isn't part of chromite.

[diandersAtChromium, 2011-02-03 00:44:22]

http://codereview.chromium.org/6334039/diff/1003/bin/chromite
File bin/chromite (right):

http://codereview.chromium.org/6334039/diff/1003/bin/chromite#newcode32
bin/chromite:32: 
Please make it so that if chromite is already in the sys.path that we use that
one.  Don't always add CROS_WORKON_SRCROOT before trying the first import.  That
way, if importing chromite works w/out any path modifications then we don't add
SRCROOT to the sys.path.  This make it so that we can't treat other dirs
directly in SRCROOT as python libraries.

http://codereview.chromium.org/6334039/diff/1003/bin/chromite#newcode44
bin/chromite:44: sys.path.insert(0, '')
Please don't add ever '' into the path.  This makes things fail if the CWD of
the user happens to have a folder called 'chromite'.

Since we still have an old version of 'chromite' in 'src/scripts', it means that
this script will fail if the user's CWD happens to be src/scripts.

http://codereview.chromium.org/6334039/diff/1003/bin/chromite#newcode64
bin/chromite:64: sys.exit(0)
Can we remove this sys.exit(0)?  ...if not, can you add a comment about why it's
needed?

[Kenneth Waters, 2011-02-03 18:16:41]

ptal.

http://codereview.chromium.org/6334039/diff/1003/bin/chromite
File bin/chromite (right):

http://codereview.chromium.org/6334039/diff/1003/bin/chromite#newcode32
bin/chromite:32: 
jrbarnette wanted it this way.

http://codereview.chromium.org/6334039/diff/1003/bin/chromite#newcode44
bin/chromite:44: sys.path.insert(0, '')
I was partially confused by by what interactive python does.

It's worse that that because of the way search works it fails if your CWD is
anywhere below scripts/.  I've made it more robust.

http://codereview.chromium.org/6334039/diff/1003/bin/chromite#newcode64
bin/chromite:64: sys.exit(0)
legacy.  removed.

[diandersAtChromium, 2011-02-03 18:36:00]

LGTM.  Feel free to push and I will do a future (minor) patch for the chroot
case.

http://codereview.chromium.org/6334039/diff/1003/bin/chromite
File bin/chromite (right):

http://codereview.chromium.org/6334039/diff/1003/bin/chromite#newcode32
bin/chromite:32: 
Talked to jrbarnette.  You can push this CL as is.  

I will submit a future CL that will detect if we're in the chroot (via existence
of /etc/debian_chroot).  If we're in the chroot, we'll do the "yield []" as the
first thing.  Else we won't do the "yield []".