Reland 'Add path utils, plus a test for it.'

Reland 'Add path utils, plus a test for it.'

Build SkPathJoin and SkBasename on windows also.

Previous CL did not build on Windows because the two functions were
accidentally placed inside an ifdef that did not include windows.
Move the functions to the top of the file, and add a comment by the
endif for clarity.

Previously reviewed at https://codereview.chromium.org/15747004/

Committed: https://code.google.com/p/skia/source/detail?r=9295

[scroggo, 2013-05-28 16:44:39]

Elliot, this is basically the same patch as
https://codereview.chromium.org/15747004/, which you already approved. Just
sending you the code review FYI.

https://codereview.chromium.org/15740024/diff/1/src/utils/SkOSFile.cpp
File src/utils/SkOSFile.cpp (right):

https://codereview.chromium.org/15740024/diff/1/src/utils/SkOSFile.cpp#newcode9
src/utils/SkOSFile.cpp:9: SkString SkOSPath::SkPathJoin(const char *rootPath,
const char *relativePath) {
This file is the only one that has changed from the prior patch. These are moved
to the top of the file, outside the platform specific sections.

[scroggo, 2013-05-28 16:45:22]

Committed manually as r9295 (presubmit successful).

[epoger, 2013-05-28 17:48:11]

LGTM

https://codereview.chromium.org/15740024/diff/1/gm/factory.cpp
File gm/factory.cpp (right):

https://codereview.chromium.org/15740024/diff/1/gm/factory.cpp#newcode1
gm/factory.cpp:1: /*
Would using the Commit Queue (or its trybots [1]) have prevented
https://code.google.com/p/skia/source/detail?r=9277 from landing in such a way
that it broke the Windows build?

If not, please file a bug.

[1] - by "or its trybots", I mean running "tools/submit_try $CHANGENAME --bot
cq", which runs tryjobs on the same set of trybots that the commit queue would
use

https://codereview.chromium.org/15740024/diff/1/gm/factory.cpp#newcode3
gm/factory.cpp:3: *
IMHO, when relanding patches I think it's preferable to upload it as two
patchsets: patchset 1 which just reapplies the previous patch, and patchset 2
which fixes the bug.  That way it's easy to see what changed from the previous
attempt.

For example: https://codereview.appspot.com/7132060 ('re-land r7258 with fixes
and tests')

[scroggo, 2013-05-28 17:56:36]

https://codereview.chromium.org/15740024/diff/1/gm/factory.cpp
File gm/factory.cpp (right):

https://codereview.chromium.org/15740024/diff/1/gm/factory.cpp#newcode1
gm/factory.cpp:1: /*
On 2013/05/28 17:48:11, epoger wrote:
> Would using the Commit Queue (or its trybots [1]) have prevented
> https://code.google.com/p/skia/source/detail?r=9277 from landing in such a way
> that it broke the Windows build?
> 
> If not, please file a bug.
> 
> [1] - by "or its trybots", I mean running "tools/submit_try $CHANGENAME --bot
> cq", which runs tryjobs on the same set of trybots that the commit queue would
> use

I suspect that it would have, although I did not suspect there was an issue of
this not building on other platforms.

https://codereview.chromium.org/15740024/diff/1/gm/factory.cpp#newcode3
gm/factory.cpp:3: *
On 2013/05/28 17:48:11, epoger wrote:
> IMHO, when relanding patches I think it's preferable to upload it as two
> patchsets: patchset 1 which just reapplies the previous patch, and patchset 2
> which fixes the bug.  That way it's easy to see what changed from the previous
> attempt.
> 
> For example: https://codereview.appspot.com/7132060 ('re-land r7258 with fixes
> and tests')

I'll make a mental note to do that next time. That would have been much more
helpful, but I didn't think about it :(