Apply two regression fix patches to WebKitGTK

[Douglas R. Reno]

From upstream this morning:

TL;DR: We recommend applying the patches from [2] and [3] over 2.44.3,
and the issues are expected to be solved in the upcoming 2.44.4 releases.


On Tue, 13 Aug 2024 09:10:44 +0200 Carlos Garcia Campos <cgarcia@igalia.com> wrote:
> WebKitGTK 2.44.3 is available for download at:
>
> https://webkitgtk.org/releases/webkitgtk-2.44.3.tar.xz (35.5MB)
>    md5sum: 46cf81df314acbf62f811bcfd99f4769
>    sha1sum: c9bcb2097d8f774b2c64ca650a4f8a6365ff54f6
>    sha256sum: dc82d042ecaca981a4852357c06e5235743319cf10a94cd36ad41b97883a0b54
>
> This is a bug fix release in the stable 2.44 series.
>
> [...]

There are two known issues in this release, which affect both the GTK and WPE
WebKit ports:

  * LTO builds made with Clang may hit what we believe is a compiler issue,
    causing segfaults when loading many pages. The issue [1] is not new but
    it started showing up in stable releases in 2.44.3 -- a workaround is
    available [2], which may or not be merged as-is, but I would recommend
    that packagers apply it before the fix is available in the next stable
    release.

  * Sites using WebAssembly can easily hit a crash related to code generation.
    This was caused by a patch backport. For the moment we have decided to
    revert it in the release branch [3], and packagers may want to do the
    same. I am investigating in case may be possible to re-apply the patch
    without introducing regressions in the next stable release, but I am
    not sure at the moment.

Cheers,
—Adrián

---
[1] https://bugs.webkit.org/show_bug.cgi?id=274780
[2] https://github.com/WebKit/WebKit/pull/32240
[3] https://github.com/WebKit/WebKit/commit/9140ce712aa87091613874d802787ab476be0e39

We're not hit by the LTO bug, but we are by the WebAssembly fix. Upstream recommends packagers patch this set of issues ASAP, so I will do so today.

[Xi Ruoyao]

Hmm maybe it's related to the browser crash when I was tracking my package on usps.com. IIRC before updating WebKitGTK it had worked fine...

[Douglas R. Reno]

Replying to Xi Ruoyao:

Hmm maybe it's related to the browser crash when I was tracking my package on usps.com. IIRC before updating WebKitGTK it had worked fine...

I can reproduce it on usps.com as well... I'll check after I have the patches in. I keep encountering other issues with packages before I get there, so this might not be done today

[Xi Ruoyao]

Do we need the first patch? We don't use clang and we don't enable LTO.

[Xi Ruoyao]

The second patch seems just sed '/returnLocation.isStackArgument/,/returnLocation = canonicalLocation/d' -i Source/JavaScriptCore/wasm/WasmBBQJIT.cpp.

[Xi Ruoyao]

usps.com package tracking page works with the second patch applied.

[Douglas R. Reno]

Replying to Xi Ruoyao:

Do we need the first patch? We don't use clang and we don't enable LTO.

I'm thinking that we don't need it. I'll just test and then add the sed :)

I'm very glad that the USPS website works now

[Douglas R. Reno]

Fixed at ade0ad4ce59fe17f4842465fa04ebd095b604171