Add basic support to dump pixel results

Add basic support to dump pixel results

BUG=111316
TEST="webkit/tools/layout_tests/run_webkit_tests.py --chromium --debug --driver-name=content_shell --additional-drt-flag=--dump-render-tree css3/css3-modsel-33.html" should pass

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=144436

[marja, 2012-06-26 13:32:43]

https://chromiumcodereview.appspot.com/10656044/diff/5002/content/shell/layou...
File content/shell/layout_test_controller_host.cc (right):

https://chromiumcodereview.appspot.com/10656044/diff/5002/content/shell/layou...
content/shell/layout_test_controller_host.cc:34: expected_pixel_hash_ =
expected_pixel_hash_;
This NOOP line is probably not what you want (extra _).

[jochen (gone - plz use gerrit), 2012-06-26 13:38:55]

John, can you plz review this CL

https://chromiumcodereview.appspot.com/10656044/diff/5002/content/shell/layou...
File content/shell/layout_test_controller_host.cc (right):

https://chromiumcodereview.appspot.com/10656044/diff/5002/content/shell/layou...
content/shell/layout_test_controller_host.cc:34: expected_pixel_hash_ =
expected_pixel_hash_;
On 2012/06/26 13:32:44, marja wrote:
> This NOOP line is probably not what you want (extra _).

Done.

[jochen (gone - plz use gerrit), 2012-06-26 13:40:32]

https://chromiumcodereview.appspot.com/10656044/diff/7007/content/shell/layou...
File content/shell/layout_test_controller.cc (right):

https://chromiumcodereview.appspot.com/10656044/diff/7007/content/shell/layou...
content/shell/layout_test_controller.cc:106: void MakeBitmapOpaque(SkBitmap*
bitmap) {
compare to makeCanvasOpaque in TestShell.cpp

https://chromiumcodereview.appspot.com/10656044/diff/7007/content/shell/layou...
content/shell/layout_test_controller.cc:177: void
LayoutTestController::OnCaptureImageDump() {
Compare to ChromeRenderViewObserver::OnCaptureSnapshot

https://chromiumcodereview.appspot.com/10656044/diff/7007/content/shell/layou...
File content/shell/layout_test_controller_host.cc (right):

https://chromiumcodereview.appspot.com/10656044/diff/7007/content/shell/layou...
content/shell/layout_test_controller_host.cc:102: void
LayoutTestControllerHost::OnImageDump(const SkBitmap& image) {
compare to TestShell::dumpImage

[jam, 2012-06-26 15:03:10]

lgtm

have you timed how long it takes to send the image over IPC? i thought that
MacIPC is slow at sending large data. not urgent, but if it turns out to be slow
enough to impact test run performance, you might want to send it over shared
memory.

http://codereview.chromium.org/10656044/diff/7007/content/shell/layout_test_c...
File content/shell/layout_test_controller_host.cc (right):

http://codereview.chromium.org/10656044/diff/7007/content/shell/layout_test_c...
content/shell/layout_test_controller_host.cc:110:
base::MD5Sum(reinterpret_cast<void*>(image.getPixels()),
nit: every pointer is a void*, so shouldn't need a cast

http://codereview.chromium.org/10656044/diff/7007/content/shell/shell_messages.h
File content/shell/shell_messages.h (right):

http://codereview.chromium.org/10656044/diff/7007/content/shell/shell_message...
content/shell/shell_messages.h:22: // responds with a
ShelLViewHostMsg_ImageDump.
nit ShelL

[jochen (gone - plz use gerrit), 2012-06-26 17:36:42]

On 2012/06/26 15:03:10, John Abd-El-Malek wrote:
> lgtm
> 
> have you timed how long it takes to send the image over IPC? i thought that
> MacIPC is slow at sending large data. not urgent, but if it turns out to be
slow
> enough to impact test run performance, you might want to send it over shared
> memory.

No, I didn't time it yet. The pictures are 800x600, so it's not that much data.

However, I could just send the pixel hash to the renderer as well, and only send
the image back if it differs from the expected result.