GM: make behavior of make_filename() more late-binding

GM: make behavior of make_filename() more late-binding

This will be needed to add the option of writing out files with
checksum-based filenames.

R=scroggo@google.com

Committed: https://code.google.com/p/skia/source/detail?r=9215

[epoger, 2013-05-19 07:02:08]

https://codereview.chromium.org/15415003/diff/1/gm/gm_expectations.h
File gm/gm_expectations.h (left):

https://codereview.chromium.org/15415003/diff/1/gm/gm_expectations.h#oldcode54
gm/gm_expectations.h:54: SkString make_filename(const char path[],
Patchset 1 just moves make_filename() from gm_expectations.[h/cpp] into
gmmain.cpp (which is the only place it is used anymore).  No behavioral change.

[epoger, 2013-05-19 09:00:43]

https://codereview.chromium.org/15415003/diff/6001/gm/gmmain.cpp
File gm/gmmain.cpp (right):

https://codereview.chromium.org/15415003/diff/6001/gm/gmmain.cpp#newcode213
gm/gmmain.cpp:213: SkString make_shortname_plus_config(const char *shortName,
const char *configName) {
Patchset 2 renames some variables, parameters, and this method to make it
clearer where we are passing around "config/shortName" vs just "shortName". 
Still no behavioral changes, as verified using the self-tests.

https://codereview.chromium.org/15415003/diff/6001/gm/gmmain.cpp#newcode1598
gm/gmmain.cpp:1598: // TODO(epoger): In this case, we call make_filename() with
Any insights on this?  I think it's worth looking into in a separate CL...

[epoger, 2013-05-19 09:49:19]

https://codereview.chromium.org/15415003/diff/9001/gm/gmmain.cpp
File gm/gmmain.cpp (right):

https://codereview.chromium.org/15415003/diff/9001/gm/gmmain.cpp#newcode218
gm/gmmain.cpp:218: const char *shortName,
Patchset 3 splits the shortNamePlusConfig parameter that had been passed into
make_filename() into two params: shortName and configName.

For the most part, there are no behavioral changes (see one note below).  But
this is an important step... it will allow us to make late-binding decisions
about how we construct the full filename.  (Specifically, when we use
checksum-based filenames, we won't include the configName within the filename.)

https://codereview.chromium.org/15415003/diff/9001/gm/gmmain.cpp#newcode1607
gm/gmmain.cpp:1607: // TODO(epoger): Make sure this still works even though the
This is the one behavioral change in this CL so far (that I know of), and
unfortunately, it isn't exercised by our self-tests).  As noted here, I think
it's actually an IMPROVEMENT, but we should try to make sure.

[epoger, 2013-05-19 10:03:56]

Ready for review at patchset 3.

A little more info about what this will enable in a future CL... the idea is, I
want to add a checksumBasedFilenames flag that will cause files to be written
out with filenames modified like this:

instead of writing out the actual 8888 selftest1 PNG to
  mismatchPath/8888/selftest1.png
or
  mismatchPath/selftest1_8888.png
write it out to
  mismatchPath/bitmap-64bitMD5/selftest1.png/1209453360120438698.png

In order to do that, we'll need to pass shortName ("selftest1") and configName
("8888") into make_filename() as separate parameters.

[scroggo, 2013-05-20 17:14:47]

lgtm

[epoger, 2013-05-21 15:45:49]

Committed patchset #4 manually as r9215 (presubmit successful).