script-src-redirect.html is flaky.

• Component: WebCore Misc.
• Status: RESOLVED FIXED

[Mike West]

From https://bugs.webkit.org/show_bug.cgi?id=86323:

(In reply to comment #22)
> (From update of attachment 141783 [details] [details])
> Clearing flags on attachment: 141783
> 
> Committed r117006: <http://trac.webkit.org/changeset/117006>

After this, script-src-redirect.html has been really flaky.
I wonder if we should roll it out.

---

Looks like `script-src-redirect.html` is another iframe ordering test. I'll rewrite it to use the same methodology as `source-list-parsing.html`.

[Mike West]

Created attachment 141971 [details]
Patch

[Mike West]

Can you take a look at this, Adam? :)

[WebKit Review Bot]

Comment on attachment 141971 [details]
Patch

Clearing flags on attachment: 141971

Committed r117123: <http://trac.webkit.org/changeset/117123>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Kent Tamura]

It looks worse.
http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=script-src-redirect.html

[Mike West]

It does look worse. (In reply to comment #5)
> It looks worse.
> http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=script-src-redirect.html

It's flaking in a very weird way:

Instead of blocking `http://127.0.0.1:8000/security/contentSecurityPolicy/resources/redir.php?url=http://localhost:8000/security/contentSecurityPolicy/resources/script.js`, CSP is blocking `http://127.0.0.1:8000/security/contentSecurityPolicy/resources/echo-script-src.pl?should_run=no&csp=%20script-src%20localhost&q=http://127.0.0.1:8000/security/contentSecurityPolicy/resources/redir.php%3furl=http://localhost:8000/security/contentSecurityPolicy/resources/script.js`, which doesn't make any sense.

I haven't been able to replicate this locally. Do you have any ideas, Adam?

[Adam Barth]

Thanks really strange.

Maybe it's some kind of strange race where we're loading about:blank but it's using the URL of the parent document for the CSP check somehow?  Another possibility is that we're using "" as the src somehow and that's getting resolve to match the URL of the parent document.

[Zoltan Arvai]

http/tests/security/contentSecurityPolicy/script-src-redirect.html
skipped on Qt in trac.webkit.org/changeset/117243 .
Please unskip the test when it works again.

[Mike West]

Arg. It's not a strange race. It's a stupid typo. :/

I'll have a patch for review in a few minutes.

[Mike West]

Created attachment 142669 [details]
Patch

Comment on attachment 142669 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=142669&action=review

> LayoutTests/http/tests/security/contentSecurityPolicy/resources/multiple-iframe-test.js:2
> +  layoutTestController.waitUntilDone();

4 spaces indenting

> LayoutTests/http/tests/security/contentSecurityPolicy/resources/multiple-iframe-test.js:11
> +    var baseURL = 'http://127.0.0.1:8000/security/contentSecurityPolicy/';

please either use ' or "

> LayoutTests/platform/chromium/test_expectations.txt:-3724
> -BUGWK86433 : http/tests/security/contentSecurityPolicy/script-src-redirect.html = TEXT PASS

why not unskip it for the other platforms as well?

[Mike West]

Created attachment 142698 [details]
Patch

[Mike West]

Thanks, Jochen. I've addressed your comments.

[WebKit Review Bot]

Comment on attachment 142698 [details]
Patch

Clearing flags on attachment: 142698

Committed r117583: <http://trac.webkit.org/changeset/117583>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Mike West]

It looks like http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=script-src-redirect.html is greening up nicely.