Bug 182286 – [WPE] Update jhbuild dependencies

• Component: WPE WebKit
• Status: RESOLVED FIXED

[Charlie Turner]

[WPE] Update jhbuild dependencies

[Charlie Turner]

Created attachment 332648 [details]
Patch

[Charlie Turner]

Created attachment 332649 [details]
Patch

[Michael Catanzaro]

Comment on attachment 332649 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=332649&action=review

> Tools/wpe/jhbuild.modules:61
> +    <branch module="/pub/GNOME/sources/glib/2.54/glib-2.54.3.tar.xz" version="2.54.3"

This will break RunLoop, see https://bugzilla.gnome.org/show_bug.cgi?id=761102. It looks OK if you add the patch for this.

[Charlie Turner]

Created attachment 332668 [details]
Patch

[Carlos Alberto Lopez Perez]

Comment on attachment 332668 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=332668&action=review

> Tools/ChangeLog:25
> +
> +        The initial reason for performing these upgrades was that when
> +        visiting https://youtube.com, WPE was getting TLS certificate
> +        errors. After upgrading glib-networking, these were fixed, but the
> +        upgrade introduced dependencies on newer versions of the other
> +        packages upgraded in this commit.
> +
> +        The upgrade to glib caused a linking error in gstreamer, the
> +        following errors were being logged during linking,
> +
> +        //usr/lib/x86_64-linux-gnu/libpangoft2-1.0.so.0: undefined reference to `hb_glib_script_from_script'
> +        //usr/lib/x86_64-linux-gnu/libpangoft2-1.0.so.0: undefined reference to `hb_glib_get_unicode_funcs'
> +        //usr/lib/x86_64-linux-gnu/libpangoft2-1.0.so.0: undefined reference to `hb_glib_script_to_script'
> +
> +        This was fixed by removing the --with-glib=no from the harfbuzz
> +        package.

....

> Tools/wpe/jhbuild.modules:122
>    <autotools id="harfbuzz" autogen-sh="configure"
> -        autogenargs="--with-cairo=no --with-glib=no --with-freetype=yes --with-fontconfig=yes">
> +        autogenargs="--with-cairo=no --with-freetype=yes --with-fontconfig=yes">
>      <dependencies>
>        <dep package="freetype6"/>
>        <dep package="fontconfig"/>

Why is better passing "--with-glib=no" to the configure instead of adding a '<dep package="glib">' entry in the module build dependencies ?

[Charlie Turner]

Created attachment 332678 [details]
Patch

[Charlie Turner]

(In reply to Carlos Alberto Lopez Perez from comment #5)
> Comment on attachment 332668 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=332668&action=review
> 
> > Tools/ChangeLog:25
> > +
> > +        The initial reason for performing these upgrades was that when
> > +        visiting https://youtube.com, WPE was getting TLS certificate
> > +        errors. After upgrading glib-networking, these were fixed, but the
> > +        upgrade introduced dependencies on newer versions of the other
> > +        packages upgraded in this commit.
> > +
> > +        The upgrade to glib caused a linking error in gstreamer, the
> > +        following errors were being logged during linking,
> > +
> > +        //usr/lib/x86_64-linux-gnu/libpangoft2-1.0.so.0: undefined reference to `hb_glib_script_from_script'
> > +        //usr/lib/x86_64-linux-gnu/libpangoft2-1.0.so.0: undefined reference to `hb_glib_get_unicode_funcs'
> > +        //usr/lib/x86_64-linux-gnu/libpangoft2-1.0.so.0: undefined reference to `hb_glib_script_to_script'
> > +
> > +        This was fixed by removing the --with-glib=no from the harfbuzz
> > +        package.
> 
> ....

I don't know what your dots mean, but I rephrased this sentence.

> 
> > Tools/wpe/jhbuild.modules:122
> >    <autotools id="harfbuzz" autogen-sh="configure"
> > -        autogenargs="--with-cairo=no --with-glib=no --with-freetype=yes --with-fontconfig=yes">
> > +        autogenargs="--with-cairo=no --with-freetype=yes --with-fontconfig=yes">
> >      <dependencies>
> >        <dep package="freetype6"/>
> >        <dep package="fontconfig"/>
> 
> Why is better passing "--with-glib=no" to the configure instead of adding a
> '<dep package="glib">' entry in the module build dependencies ?

I don't have an answer to that question, I dropped to it rely on the auto package behavior and pick up the glib in the jhbuild. I switched to your suggestion.

[Michael Catanzaro]

Comment on attachment 332678 [details]
Patch

OK, but watch the bots when this lands because this is likely to both break and fix many tests. You'll need to update the expectations accordingly....

[Charlie Turner]

A(In reply to Michael Catanzaro from comment #8)
> Comment on attachment 332678 [details]
> Patch
> 
> OK, but watch the bots when this lands because this is likely to both break
> and fix many tests. You'll need to update the expectations accordingly....

Agreed, I didn't notice much running the test deltas locally, but I will be checking the results after it lands, which I'll do tomorrow when I can baby sit it.

[EWS Watchlist]

Comment on attachment 332678 [details]
Patch

Attachment 332678 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.webkit.org/results/6264721

New failing tests:
imported/w3c/web-platform-tests/service-workers/service-worker/navigation-redirect.https.html

[EWS Watchlist]

Created attachment 332697 [details]
Archive of layout-test-results from ews104 for mac-sierra-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: ews104  Port: mac-sierra-wk2  Platform: Mac OS X 10.12.6

[Charlie Turner]

Created attachment 332757 [details]
Patch

Rebase patch

[Charlie Turner]

Created attachment 332758 [details]
Patch for landing

[Charlie Turner]

Is this patch stuck in the commit queue because of comment 11 (which surely can't be caused by this patch, surely!)

https://trac.webkit.org/wiki/CommitQueue
Q: How long until a patch lands after I set commit-queue+?

10-15 minutes. Depends on if the ​tree is red or not. The commit-queue will not commit when buildbots for the ​core platforms are red.

[WebKit Commit Bot]

Comment on attachment 332758 [details]
Patch for landing

Clearing flags on attachment: 332758

Committed r227900: <https://trac.webkit.org/changeset/227900>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Michael Catanzaro]

The wiki page is wrong/obsolete. commit-queue normally takes an hour or thereabouts.