Summary: | [GTK] Remove unneeded modules from the jhbuild | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Carlos Garcia Campos <cgarcia> | ||||
Component: | WebKitGTK | Assignee: | Nobody <webkit-unassigned> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | bugs-noreply, mcatanzaro | ||||
Priority: | P2 | Keywords: | Gtk | ||||
Version: | WebKit Nightly Build | ||||||
Hardware: | Unspecified | ||||||
OS: | Unspecified | ||||||
See Also: | https://bugzilla.gnome.org/show_bug.cgi?id=760367 | ||||||
Bug Depends on: | |||||||
Bug Blocks: | 170500 | ||||||
Attachments: |
|
Description
Carlos Garcia Campos
2017-04-06 00:52:07 PDT
Created attachment 306369 [details]
Patch
Let's start by removing what we don't ned to build. This shouldn't affect the tests.
It turns out it does affect tests, there are only a few failures. The libxml version in the bot is exactly the same, so it must be libxslt. However, I can reproduce some of the failures in my laptop with even newer versions. Ok, the problem is not libxslt but libxml. It's weird because we used 2.9.1 which is what the bot has installed ii libxml2:amd64 2.9.1+dfsg1-5+deb8u4 I can also reproduce the differences with my libxml from debian testing ii libxml2:amd64 2.9.4+dfsg1-2.2 I'd rather introduce new failures and update test expectations accordingly than continue using the older version. So, this regressed in libxml2 and we never noticed it because our tests ran fine with the old version of libxml2. However, it has been broken for our users all this time, since most of the distros are shipping a libxml2 version with the regression. This is a good example of why it's not a good idea to use old versions of our deps to run the tests. Committed r218550: <http://trac.webkit.org/changeset/218550> You need to add libffi to install-dependencies: libffi-dev (Debian) libffi-devel (Fedora) libffi (Arch) |