Add ConfigSettings.relative_cwd.

Add ConfigSettings.relative_cwd.

This will permits defining .relative_cwd in the .isolated file to be
deterministic on the actual .isolate that definied the command, not the root
one. This is important, because the command could be defined in an important
.isolate file.

R=vadimsh@chromium.org
BUG=swarming:98

Committed: https://code.google.com/p/swarming/source/detail?repo=client&r=b53d0c1

[M-A, 2014-03-27 15:31:40]

This currently depends on https://codereview.appspot.com/81320044. There isn't
much actual dependence but it happens they touch the same lines so I decided to
pipeline them.

Even if not obvious, this CL is towards fixing
https://code.google.com/p/chromium/issues/detail?id=333473. The reason is that
the command and relative_cwd would mismatch when using 'includes'. Fixing this
will make the actual code for the component builds much simpler to write by
simply using an includes but currently, this breaks relative_cwd, which makes
the command unusable.

Note that the .isolated relative_cwd value is not changed in this CL. This CL
only add bookkeeping that is not yet used but tested with the unit test. I'll do
the actual work towards that in a separate CL, I preferred to not get a too
large CL.

[vadimsh, 2014-03-27 18:30:49]

https://codereview.appspot.com/81470043/diff/20001/isolate_format.py
File isolate_format.py (right):

https://codereview.appspot.com/81470043/diff/20001/isolate_format.py#newcode516
isolate_format.py:516: name, it must be an absolute path, it's the path whre to
start the command
Why is it called "relative_cwd" then?.. Relative to what? Maybe just "cwd" or
"command_cwd".

Also typo: "the path whre"

https://codereview.appspot.com/81470043/diff/20001/isolate_format.py#newcode598
isolate_format.py:598: This class doesn't hold relative_cwd, since it is
dependent on the finaly
nit: "on the final"?

[M-A, 2014-03-27 20:56:04]

https://codereview.appspot.com/81470043/diff/20001/isolate_format.py
File isolate_format.py (right):

https://codereview.appspot.com/81470043/diff/20001/isolate_format.py#newcode516
isolate_format.py:516: name, it must be an absolute path, it's the path whre to
start the command
On 2014/03/27 18:30:49, vadimsh wrote:
> Why is it called "relative_cwd" then?.. Relative to what? Maybe just "cwd" or
> "command_cwd".

The idea is that all the dependencies and the command are relative to this
directory.
I could use base_dir or isolate_dir, would that be ok?

> Also typo: "the path whre"

Done.

https://codereview.appspot.com/81470043/diff/20001/isolate_format.py#newcode598
isolate_format.py:598: This class doesn't hold relative_cwd, since it is
dependent on the finaly
On 2014/03/27 18:30:49, vadimsh wrote:
> nit: "on the final"?

Done.

[vadimsh, 2014-03-27 21:06:10]

https://codereview.appspot.com/81470043/diff/20001/isolate_format.py
File isolate_format.py (right):

https://codereview.appspot.com/81470043/diff/20001/isolate_format.py#newcode516
isolate_format.py:516: name, it must be an absolute path, it's the path whre to
start the command
On 2014/03/27 20:56:04, M-A wrote:
> On 2014/03/27 18:30:49, vadimsh wrote:
> > Why is it called "relative_cwd" then?.. Relative to what? Maybe just "cwd"
or
> > "command_cwd".
> 
> The idea is that all the dependencies and the command are relative to this
> directory.
> I could use base_dir or isolate_dir, would that be ok?
> 
> > Also typo: "the path whre"
> 
> Done.

base_dir sounds better to me

[vadimsh, 2014-03-27 21:18:37]

On 2014/03/27 21:06:10, vadimsh wrote:
> https://codereview.appspot.com/81470043/diff/20001/isolate_format.py
> File isolate_format.py (right):
> 
>
https://codereview.appspot.com/81470043/diff/20001/isolate_format.py#newcode516
> isolate_format.py:516: name, it must be an absolute path, it's the path whre
to
> start the command
> On 2014/03/27 20:56:04, M-A wrote:
> > On 2014/03/27 18:30:49, vadimsh wrote:
> > > Why is it called "relative_cwd" then?.. Relative to what? Maybe just "cwd"
> or
> > > "command_cwd".
> > 
> > The idea is that all the dependencies and the command are relative to this
> > directory.
> > I could use base_dir or isolate_dir, would that be ok?
> > 
> > > Also typo: "the path whre"
> > 
> > Done.
> 
> base_dir sounds better to me

Just looked at next CL. It uses isolate_dir in couple places. So maybe
'isolate_dir' is ok. As you prefer. I just don't like relative_cwd that is an
absolute path.

[M-A, 2014-03-28 13:32:09]

On 2014/03/27 21:18:37, vadimsh wrote:
> Just looked at next CL. It uses isolate_dir in couple places. So maybe
> 'isolate_dir' is ok. As you prefer. I just don't like relative_cwd that is an
> absolute path.


Used isolate_dir. Also fixed one unit test that used an ill formed isolate file
to unit test pretty_print, which was missing the point.

[vadimsh, 2014-03-28 17:41:35]

lgtm

[M-A, 2014-03-28 17:46:51]

Committed patchset #4 manually as rb53d0c1 (presubmit successful).