* Use WikiPage::newFromID() directly instead of Title::newFromID() (for now it does not change anything but it will soon)
* Directly load the data from the master when creating a page from its name, it avoids both having out-of-date information
when doing the existence check and a second database query from LinkCache::addLinkObj() when calling (WikiPage|Title)::exists()
* Moved the page existence check to only be executed when creating a title from its name since if we have a WikiPage from its ID it must exist.
Change-Id: Ifaff725d955ce111c46d6b7f00610191820f3ced
Fix for SVN r113178; seems that I forgot to remove 'parseinline' on some messages when changing wfMsgExt() to $this->msg().
Change-Id: I7138de97717b0976d1ae9da31a5d46f1ce513c85
- Check http://jquery.com/:
* IE 6.0+, FF 3.6+, Safari 5.0+, Opera, Chrome
- It does not include an Opera number, but they are only
actively testing for Opera 11.x versions. Also Operate
is quite bold in their upgrade requirement in-browser,
so that shouldn't be much of an issue.
I've also updated
https://www.mediawiki.org/wiki/Compatibility#Browser
Note that this means that we no longer support enhanced functionality in
some browsers that are currently still above the 0.1% wmf-traffic threshold
we keep. I've therefor split [[mw:Compatibility#Browser]] in a "Grade A"
and "Grade B" section. One that we support for all functionality, and the
others we support because of traffic, but not "everything" will work.
Right now the only notable difference will be jQuery and depending modules.
In reality Grade B will still get the same data served as Grade A, since
startup.js only enforces IE6 right now (bug 35906)
Auto-updating Chrome will not need a check, and Chrome < 1.0 is very minor,
and will probably even work just fine.
Change-Id: If60de580fc892ced1a43a0bc4958988d98d087d9
Changeset 1 of https://gerrit.wikimedia.org/r/#change,3365:
763b57f9f2af131a2d8e65f520a23c00109be0e1
got mis-merged in changeset 2:
f752cf8042
And introduced into master wth
0a792a1dcb
Also, marking all new functions as @since 1.19
since we want to have them there later.
Patchset 2: whitespace fixes
Patchset 3: Fix <?php
Patchset 4: rebase to current master
Change-Id: Ic7d940dfec8890de20197128505962375fac4f06
Seems that I forgot to change one instance of $info to $svnInfo in I2f4839fb587903524fdaf041597174003ccfdfeb.
Change-Id: I7e47ca6278569290df4b32cac584c94020c13b64
The MockOutputPage can be needed in various tests suite beside
GlobalFunctions. Seems to make sense to get it in its own class so we
can easily use it.
Change-Id: Icbb34db81e8cc8636d95f6326dd228bc74a741d7
This reverts commit 5586e73fac.
This broke mergeMessageFileList.php by no longer running the extension
setup files in the global scope.
Change-Id: I8f22d0b3e938ba617f4297802edfc2012d78cc9b
If we want to sort the ops for a batch ID, we can just use a tiny to small filesort.
No need to make the index bigger.
Change-Id: Ibf7592930ec633a26e675e33d42aebe7763547b1
especially the surprising start-of-line behavior. This test will
be used to implement matching behavior in Parsoid.
Change-Id: Ibfa13dfd2be2d62c285449cb090e2b112fe30f9b
PHPUnit has a timeout protection system which let us put tests in three
categories having different timeout (default: 1s, 10s, 60s). The timeout
only happens when using strict mode and having PHPUnit Invoker
installed.
Recently, the continuous integration server has been upgraded and the
PHPUnit Invoker was installed to let us run a serie of tests related to
dumping the database to XML. They do need some timeout system. Suddenly,
some tests started failing.
By default all tests are in the 1 second timeout group. Since we use a
sqlite backend and the server hard disk can get busy, some test can take
more than one second to execute.
This patch raise the default (small group) timeout to 2seconds and
explicitly define the value for the two other groups (medium and large)
using PHPUnit default values.
Change-Id: I2f5613b0c0215023c413dec8e84804175d8c8d6e
- Raise from 5s to 10s. Should fix the false positives that sometimes
happens in TestSwarm if a VM is being slow (especially IE6).
Change-Id: I4bc00b34040f5659c8456ce6978627c32497656d
Namespace existence has to be checked before calling Title::makeName() because otherwise it will fallback to a page with the same name, but in the main namespace which is not what we want here.
Change-Id: I8c087390044c54dd0502c1b8bb9321dfaec6b5f7
Automatic rebasing is often unwanted when submitting a new patchset to
gerrit. That is a default behavior of git-review which is unwanted.
New release of git-review will have a new ini setting which let us
disable rebasing by default :-)
Change-Id: I634979b8fbc0ac23a5d13893fb5eaa60a5156764
Now that there is a link on the revision hash in Special:Version, do the same in the debug toolbar.
Change-Id: I953ac70e17e72be68251e3908e164dfa46c68bf5
The URL we are crafting are using a short sha1 (7 chars long) which can
have some good chance of pointing to two different sha1. I have seen git
log to use 8 chars from time to time because of sha1 collisions, 62
occurences outu of 45315 commits as of now.
So to play it safe, this patch makes the URL points to the full sha1.
Follow https://gerrit.wikimedia.org/r/43355734eac351 (commit)
9afc1eff1c (merge)
Change-Id: Ic65d0ca89916fe43999b9e3f257d9f322198f9f8
