Bug 216100

Summary:

run-webkit-tests spends 20-30 seconds "building" ImageDiff

Product:

WebKit

Reporter:

Jonathan Bedard <jbedard>

Component:

Tools / Tests

Assignee:

Jonathan Bedard <jbedard>

Status:

RESOLVED DUPLICATE

Severity:

Normal

CC:

ap, darin, ews-watchlist, glenn, thorton, webkit-bug-importer

Priority:

Keywords:

InRadar

Version:

WebKit Nightly Build

Hardware:

Unspecified

OS:

Unspecified

Attachments:

Description	Flags
Patch	thorton: review-

Description Jonathan Bedard 2020-09-02 14:52:04 PDT

Tools/Scripts/build-imagediff --release

takes an inordinate amount of time for a null build (20-30 seconds), and run-webkit-tests calls it on every invocation, even if you pass --no-build.

This is 40-50% of the time it takes to run-webkit-tests a single test in the iOS simulator, and thus would wildly increase the turnaround rate when developing a test.

Comment 1 Jonathan Bedard 2020-09-02 14:52:29 PDT

<rdar://problem/68232690>

Comment 2 Jonathan Bedard 2020-09-02 14:57:33 PDT

Created attachment 407819 [details]
Patch

Comment 3 Tim Horton 2020-09-02 15:11:32 PDT

Comment on attachment 407819 [details]
Patch

Won't _path_to_image_diff (which is @memoize'd) store the failure to find it in check_image_diff, and return that after we build it (thus making ImageDiff not found if it got built)?

Comment 4 Jonathan Bedard 2020-09-02 15:38:36 PDT

(In reply to Tim Horton from comment #3)
> Comment on attachment 407819 [details]
> Patch
> 
> Won't _path_to_image_diff (which is @memoize'd) store the failure to find it
> in check_image_diff, and return that after we build it (thus making
> ImageDiff not found if it got built)?

Likely yes, the right solution is probably to clear out the memoization, but our current @memoize doesn't support that

Comment 5 Alexey Proskuryakov 2020-09-02 18:18:15 PDT

Isn't bug 204420 the root cause here?

Comment 6 Tim Horton 2020-09-02 18:22:12 PDT

(In reply to Alexey Proskuryakov from comment #5)
> Isn't bug 204420 the root cause here?

Do you mean "isn't this a dupe of that"? I think the answer is "yes"!

Comment 7 Darin Adler 2020-09-02 21:20:44 PDT


*** This bug has been marked as a duplicate of bug 204420 ***