Bug 227526 – [GTK] Pack header and .pc files in the built-product archive

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Philippe Normand]

The motivation for this is to enable Epiphany Canary builds from the GNOME CI.

[Philippe Normand]

Created attachment 432592 [details]
Patch

[Michael Catanzaro]

Comment on attachment 432592 [details]
Patch

I wonder, can you install WebKit into an empty install prefix, and then just copy everything it installed? That seems much less fragile.

[Philippe Normand]

Created attachment 432775 [details]
Patch

[Philippe Normand]

Created attachment 432776 [details]
Patch

[Philippe Normand]

Created attachment 432777 [details]
Patch

[Philippe Normand]

Created attachment 432778 [details]
Patch

[Philippe Normand]

Created attachment 432780 [details]
Patch

[Philippe Normand]

Aakash, the EWS fails because the new bot steps from the patch are not used (I think). How do you usually handle patches that change the EWS behavior?

[Philippe Normand]

Comment on attachment 432780 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=432780&action=review

> Tools/CISupport/built-product-archive:258
> +            contents = [os.path.join('install', directory) for directory in ['bin', 'include', 'lib', 'libexec']]

Ah, the extractBuiltProduct() will need to take this into account (moving install/* to $PWD)

[Philippe Normand]

Created attachment 432781 [details]
Patch

[Philippe Normand]

Created attachment 432782 [details]
Patch

[Michael Catanzaro]

Comment on attachment 432782 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=432782&action=review

Nice, this looks a lot better than the original version of this patch.

> Source/WebKit/gtk/webkit2gtk-web-extension.pc.in:5
> +revision=@SVN_REVISION@

I think I would land these revision=@SVN_REVISION@ changes in a separate bug/patch since they're largely unrelated to the bulk of this patch. You can set r=me.

P.S. I guess you know this will break as soon as we switch to GitHub. ;)

[Philippe Normand]

EWS looks green, but I suspect that if/when this lands it might require clean builds on the GTK fleet, because build-webkit might not regenerate properly the ninja defs based on the added --prefix option.

[Philippe Normand]

(In reply to Michael Catanzaro from comment #12)
> Comment on attachment 432782 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=432782&action=review
> 
> Nice, this looks a lot better than the original version of this patch.
> 
> > Source/WebKit/gtk/webkit2gtk-web-extension.pc.in:5
> > +revision=@SVN_REVISION@
> 
> I think I would land these revision=@SVN_REVISION@ changes in a separate
> bug/patch since they're largely unrelated to the bulk of this patch. You can
> set r=me.
> 
> P.S. I guess you know this will break as soon as we switch to GitHub. ;)

SVN_REVISION is used in various places across the tree. I expect that the migration to GitHub will take this into account :)

[Philippe Normand]

Created attachment 432798 [details]
Patch

[Philippe Normand]

(In reply to Philippe Normand from comment #13)
> EWS looks green, but I suspect that if/when this lands it might require
> clean builds on the GTK fleet, because build-webkit might not regenerate
> properly the ninja defs based on the added --prefix option.

I'd like to land this on Monday and have our QA folks around to help keep an eye on the bots for clean builds. :)

[Philippe Normand]

Comment on attachment 432798 [details]
Patch

I've check the build should be OK, build-webkit detects the new argument and re-runs cmake.

[EWS]

Committed r279563 (239394@main): <https://commits.webkit.org/239394@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 432798 [details].

[Radar WebKit Bug Importer]

<rdar://problem/80167129>

[Aakash Jain]

Restarted buildbot on build.webkit.org and EWS to pick up this change.