Bug 58402 - old-run-webkit-tests/new-run-webkit-tests shouldn't read or write .checksum files
Summary: old-run-webkit-tests/new-run-webkit-tests shouldn't read or write .checksum f...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: 528+ (Nightly build)
Hardware: Other OS X 10.5
: P2 Normal
Assignee: Tony Chang
URL:
Keywords:
: 57575 (view as bug list)
Depends on:
Blocks: 56286
  Show dependency treegraph
 
Reported: 2011-04-12 17:21 PDT by Tony Chang
Modified: 2011-04-21 14:10 PDT (History)
3 users (show)

See Also:


Attachments
Patch (2.80 KB, patch)
2011-04-12 17:24 PDT, Tony Chang
no flags Details | Formatted Diff | Diff
Patch (10.63 KB, patch)
2011-04-21 11:49 PDT, Tony Chang
no flags Details | Formatted Diff | Diff
add comment back to ORWT (10.73 KB, patch)
2011-04-21 11:53 PDT, Tony Chang
ojan: review+
Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Tony Chang 2011-04-12 17:21:38 PDT
old-run-webkit-tests shouldn't write .checksum files
Comment 1 Tony Chang 2011-04-12 17:24:10 PDT
Created attachment 89313 [details]
Patch
Comment 2 Tony Chang 2011-04-12 17:24:44 PDT
I'll send a note to webkit-dev before landing this.
Comment 3 Tony Chang 2011-04-13 16:58:12 PDT
Feel free to suggest other reviewers.
Comment 4 Adam Roben (:aroben) 2011-04-14 06:16:15 PDT
Comment on attachment 89313 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=89313&action=review

> Tools/Scripts/old-run-webkit-tests:936
> +            if (-f $expectedChecksumPath) {
> +                unlink $expectedChecksumPath;
> +            }

This seems strange. People are going to have files disappearing from their working trees without any explanation. I think it would be a lot better to just do a one-time delete of all .checksum files in the WebKit tree and commit that separately.
Comment 5 Tony Chang 2011-04-14 10:13:51 PDT
(In reply to comment #4)
> This seems strange. People are going to have files disappearing from their working trees without any explanation. I think it would be a lot better to just do a one-time delete of all .checksum files in the WebKit tree and commit that separately.

I said I would send a note to webkit-dev before landing this :)

But if you prefer, I can update the existing pngs and remove the .checksum files first.  The downside is there will be a window of time between the patches (or patches created before the .checksum delete and landed after) where people will be checking in .checksums after I've deleted all of them.  I guess I can just do a second pass a week later.
Comment 6 Adam Roben (:aroben) 2011-04-14 10:47:58 PDT
Seems like the steps could be:

1) Embed checksums in all existing PNGs
2) Change ORWT not to read or write .checksum files
3) Delete all .checksum files

These could each be done separately.
Comment 7 Tony Chang 2011-04-14 10:56:00 PDT
(In reply to comment #6)
> Seems like the steps could be:
> 
> 1) Embed checksums in all existing PNGs
> 2) Change ORWT not to read or write .checksum files
> 3) Delete all .checksum files
> 
> These could each be done separately.

That sounds fine to me.  This is part of step 2 (the read part is done).  I'll come back and r? this once step 1 is done.
Comment 8 Tony Chang 2011-04-21 10:53:53 PDT
*** Bug 57575 has been marked as a duplicate of this bug. ***
Comment 9 Tony Chang 2011-04-21 11:49:13 PDT
Created attachment 90566 [details]
Patch
Comment 10 Adam Roben (:aroben) 2011-04-21 11:50:36 PDT
Comment on attachment 90566 [details]
Patch

The ORWT changes look good to me. Someone more familiar with NRWT should review that part.
Comment 11 Tony Chang 2011-04-21 11:50:50 PDT
(In reply to comment #6)
> 1) Embed checksums in all existing PNGs
> 2) Change ORWT not to read or write .checksum files
> 3) Delete all .checksum files


(1) is done.  This is (2).  I'll send a note to webkit-dev after landing this.  It'll be harmless if people continue to check in .checksum files for a few days.  They'll just be ignored.
Comment 12 Tony Chang 2011-04-21 11:53:26 PDT
Created attachment 90567 [details]
add comment back to ORWT
Comment 13 Ojan Vafai 2011-04-21 11:58:30 PDT
Comment on attachment 90567 [details]
add comment back to ORWT

View in context: https://bugs.webkit.org/attachment.cgi?id=90567&action=review

> Tools/Scripts/webkitpy/layout_tests/layout_package/single_test_runner.py:149
>          if self._options.pixel_tests and driver_output.image_hash:

maybe this check should be driver_output.image instead of driver_output.image_hash now?
Comment 14 Tony Chang 2011-04-21 14:09:14 PDT
Committed r84544: <http://trac.webkit.org/changeset/84544>
Comment 15 Tony Chang 2011-04-21 14:10:26 PDT
(In reply to comment #13)
> (From update of attachment 90567 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=90567&action=review
> 
> > Tools/Scripts/webkitpy/layout_tests/layout_package/single_test_runner.py:149
> >          if self._options.pixel_tests and driver_output.image_hash:
> 
> maybe this check should be driver_output.image instead of driver_output.image_hash now?

I switched this before committing.  There's a lot of code in NRWT that we can update/delete now, but I'll do that in a follow up patch.