Remove test_shell_tests dependency on the test_shell binary.

Replace test_shell_tests dependency on the test_shell with test_shell_pak dependency.
    
This fixes the missing requirement that test_shell was previously satisifying for test_shell_tests.

[Evan Martin, 2009-09-09 18:34:22]

LGTM

Here's the review where that code was introduced:
http://codereview.chromium.org/56161
it might be worth asking those people if it still applies now.

[Michael Moss, 2009-09-09 18:39:10]

Steven,
Any idea if there is a problem with removing this? test_shell_tests seems to run
fine without it, but it's not clear why it was needed (I suspect tests worked
differently 5 months ago when this was added).

[sgk, 2009-09-09 19:10:51]

When that was added, a clean rebuild of just 'test_shell_tests' would fail
unless it was accompanied by a 'test_shell' build (empirically).  I assumed
it was because it needed the binary, but there could have been something
more subtle going on.

I assume you verified this CL is okay by building just 'test_shell_tests' in
a completely clean tree.  The tests might, of course, pass if it is still
using 'test_shell' (e.g.) and there happens to be one sitting there from a
previous build.

If the 'test_shell_tests' passes now without any other built targets, then
something changed for the better in the meantime, and the code LGTM.

        --SK

On Wed, Sep 9, 2009 at 11:39 AM, <mmoss@chromium.org> wrote:

> Steven,
> Any idea if there is a problem with removing this? test_shell_tests
> seems to run fine without it, but it's not clear why it was needed (I
> suspect tests worked differently 5 months ago when this was added).
>
>
> http://codereview.chromium.org/201067
>

[Michael Moss, 2009-09-09 19:15:52]

I tested with test_shell gone, though didn't do a completely clean build to make
sure any intermediate files weren't missing (though if there are, we probably
still want this dependency gone, and the missing files added to
test_shell_common). Will try that now, thanks.

[Michael Moss, 2009-09-09 20:28:04]

It looks like the missing bit was test_shell.pak, so I made that into a separate
target and dependency.

[Michael Moss, 2009-09-10 16:54:50]

Any idea what this failure on the mac try-bot means:
http://build.chromium.org/buildbot/try-server/builders/mac/builds/16679/steps...

[Michael Moss, 2009-09-10 17:26:20]

This seems to be related to the 'process_outputs_as_mac_bundle_resources' line.
If I take that out of this patch, gyp runs fine. Any idea why gyp accepts that
as part of the original 'test_shell' target, but complains when it's brought in
as a dependency in the new 'test_shell_pak' target?

[TVL, 2009-09-10 17:32:42]

http://codereview.chromium.org/201067/diff/2003/3004
File webkit/tools/test_shell/test_shell.gyp (right):

http://codereview.chromium.org/201067/diff/2003/3004#newcode412
Line 412: # TODO(?) Would this be satisified with test_shell_pak?
no, the test load resources directly out of the TestShell.app bundle on disk, so
that needs to exist.  As to why, core code (ie-not in the unittest), need to
know where to fetch some things from disk, so it uses mac_util::MainAppBundle()
to get a bundle, the unittests can then use the override to point that where
needed.

[Michael Moss, 2009-09-10 17:36:21]

OK, I think given this and the other Mac issues that I don't understand, I'll
just cover the Linux case, unless someone thinks the 'test_shell_pak' target is
somehow useful for Mac.

[sgk, 2009-09-10 18:06:37]

latest lgtm