* Do additional validation and is_array() check in LanguageConverter
* Make redirects be in content language again (remove from Title->getPageLanguage())
* Pass title object from ParserCache to ParserOptions->optionsHash
* Made variant tabs hidden on special pages (has no or sometimes wrong/inconsistent effect there; plus it's mainly in the user language)
* Made redirects be in content language again (remove from Title->getPageLanguage())
$wgContLang in CategoryPage was probably removed with the better directionality improvements.
$wgServer, $wgRequest in getFullUrl() unused since r94375
Other variables added in r95396 iwtransclusion merge.
* Introduce $wgCanonicalServer, which should typically be a fully qualified version of $wgServer but in practice can be anything that you'd like to be used in IRC/e-mail notifs
** Default value is $wgServer, expanded to http:// if protocol-relative
** This means you can easily set HTTPS as the 'default' protocol to use in IRC and e-mail notifs by setting $wgCanonicalServer to https://example.com
* Introduce Title::getCanonicalURL(). Similar to getInternalURL(), including a hook for WMF usage (which will be needed as long as secure.wikimedia.org is used)
** Also add escapeCanonicalURL(). Due to some ridiculous accident of history, the other escapeFooURL() functions don't have a $variant parameter; I decided not to follow that bad example
* Reinstate the spirit of r44406 and r44412: instead of calling getInternalURL() (or getCanonicalURL()) and regexing the title parameter out, obtain the path to index.php using $wgCanonicalServer . $wgScript and append params to that. Sadly, we need to add a hook here to support the secure server hack for WMF, but that's the price of saner code in this case
* Introduce the {{canonicalurl:}} and {{canonicalurle:}} parser functions, which work just like {{fullurl:}} and {{fullurle:}} except that they use getCanonicalURL() instead of getFullURL()
* Use {{canonicalurl:}} in the enotif_body message, fixing bug 29993 (protocol-relative URLs appear in e-mail notifications)
We should really, really break the connection between internal URLs used for Squid purges and URLs used for IRC, and use the latter in e-mails as well. This was briefly discussed in the CR comments on r44412 and I'll be working on that today.
Instead of having getFullURL and getLocalURL call each other, move the interwiki workload into getLocalURL, and make getFullURL simply call getLocalURL then expand it (use the wf method in case it's already a full url) and append the fragment to it.
- Allow extensions to hook into WebRequest::getPathInfo and add to or alter the way titles are extracted from paths
- Add a $variant argument to the GetLocalURL hook; It's always had $query, but never had $variant. As a result extensions using GetLocalURL never new if getLocalURL and have the possibility of trying to change the url in cases where they shouldn't and as a result breaking links on wiki with language variants.
- Add GetLocalURL::Internal hook for non-interwiki links. These kinds of links internally use a ugly hack for action=render and an extension using GetLocalURL can be buggy in render mode if they don't re-implement the same ugly hack that MW does. This ::Internal hook runs before the hack does so extension authors don't need to be exposed to our ugly hacky code.
- Add GetLocalURL::Article hook specifically for url tweaks to pretty urls (ie: Only when we would apply $wgArticlePath); This hook avoids the need for extensions that only want to tweak pretty url output. This hook avoids the need to make a bunch of tests for things like !$title->isExternal(), $query == '', and $variant === false which getLocalURL does and could potentially change in the future making wider GetLocalURL hooks change in function requiring extension updates.
Scanning all of core and all the extensions we have in svn, it doesn't look like anyone makes any calls to Interwiki::getURL using the argument so this should be safe enough to clean up the api.
Also update Title::getFullURL to make use of the first arg now.
Breaks unit tests as below, not going to be able to fix them before I disappear for the evening, so might aswell leave trunk clean
ArticleTablesTest testbug14404
Error:
ArticleTablesTest::testbug14404
Undefined offset: 0
/home/ci/cruisecontrol-bin-2.8.3/projects/mw/source/tests/phpunit/includes/ArticleTablesTest.php:31
/home/ci/cruisecontrol-bin-2.8.3/projects/mw/source/tests/phpunit/MediaWikiTestCase.php:60
/home/ci/cruisecontrol-bin-2.8.3/projects/mw/source/tests/phpunit/MediaWikiPHPUnitCommand.php:20
/home/ci/cruisecontrol-bin-2.8.3/projects/mw/source/tests/phpunit/phpunit.php:60
ParserTests testParserTest #552 - testParserTest with data set #551
Failure:
ParserTests::testParserTest with data set #551 ('RAW magic word', '{{RAW:QUERTY}}', '<p><a href="/index.php?title=Template:QUERTY&action=edit&redlink=1" class="new" title="Template:QUERTY (page does not exist)">Template:QUERTY</a>
</p>', '', '')
RAW magic word
Failed asserting that two strings are equal.
--- Expected
+++ Actual
@@ @@
-<p><a href="/index.php?title=Template:QUERTY&action=edit&redlink=1" class="new" title="Template:QUERTY (page does not exist)">Template:QUERTY</a>
+<p><a href="/index.php?title=Template:RAW:QUERTY&action=edit&redlink=1" class="new" title="Template:RAW:QUERTY (page does not exist)">Template:RAW:QUERTY</a>
</p>
/home/ci/cruisecontrol-bin-2.8.3/projects/mw/source/tests/phpunit/includes/parser/NewParserTest.php:545
/home/ci/cruisecontrol-bin-2.8.3/projects/mw/source/tests/phpunit/MediaWikiTestCase.php:60
/home/ci/cruisecontrol-bin-2.8.3/projects/mw/source/tests/phpunit/MediaWikiPHPUnitCommand.php:20
/home/ci/cruisecontrol-bin-2.8.3/projects/mw/source/tests/phpunit/phpunit.php:60
* Marked Title::userCanEditCssSubpage() and Title::userCanEditJsSubpage() as deprecated since these were the lasts calls to that functions (core and extensions)
* Get the action parameter from Skin::userCanPreview() instead of requesting it from the callers
* User::getRights(), User::isAllowed() and User::getGroupPermissions now optionally accept a namespace parameter. If not set, it will check whether the user has the right for all namespaces.
* Anything that uses Title::getUserPermissionsErrorsInternal() automatically supports per-namespace permissions. This includes Title::getUserPermissionsErrors and Title::(quick)UserCan.
* Fix tests that set User::mRights
The next step would be to change all User::isAllowed() to Title::quickUserCan or pass the namespace to User::isAllowed().
I'm not 100% sure this is the right thing to do, see the bug for the details. But basically this doesn't include an edit in the count if its text is hidden and its hidden from admins. (Not sure if it should not be included only if everything is hidden). Its also weird to show people different things depending if they have suppress rights, without really indicating that.
Minor db note: This causes the query to no longer use a covering index. I don't think that matters but just thought i'd mention.
p.s. The upload page show deleted edits link is broken right now, (from before) I'll fix in a follow-up.
* Added Title::loadFromRow() function and made WikiPage::loadPageData() use it; avoids raw Title field accessing
* Added Revision::newFromPageId() function and changed WikiPage::loadLastEdit() to use it. This makes it try a slave first instead of always hitting the master. It also makes it more consistent with getLatest() for sanity.
* Made WikiPage::loadPageData() use accessor for Title::mRestrictionsExpiry
* Make TOC numberings be in the page content language instead of wiki content language.
* Update getPageLanguage() and add a hook (for bug 9360/28970).
* Show redirects (when viewing a page with &redirect=no) in the user language direction (not essential but nicer imo).
This is basicly a merge of r84386 & r84491, see their commit messages for
more details.
r84386 makes wikilinks nicer by updating the URL forge implemented by r2621
r84491 fix an issue with the (un)?watch links. getParamValue should not be
used to guess 'title' or 'action'
* Skip user block checks for Title::quickUserCan(). This restores the behavior from 1.16 rsp. before r65504.
* Remove unnecessary check for "$short && count($errors)", this is already handled by getUserPermissionsErrorsInternal().
Need a way to do a blank title without using the constructor (Title::newBlank() maybe?)
* (bug 28608) Fatal error: Call to protected Title::__construct()
Also make a few changes to the functions available. SpecialPageFactory::resolveAlias() now takes an optional subpage and returns array(<name>,<subpage>). Similarly merge getPage() and getPageByAlias(). There were many examples of (extensions particularly) making dubious assumptions about the presence or absence of subpages or canonical-ness.
I didn't deprecate SpecialPage::getTitleFor() as it's got over six hundred calls. I'm rather undecided on the best position of getPage()/executePath(). Although the latter needs cleanup anyway.
* Block.php: Set mExpiry to the value we store in the database, so that Block::equals() returns true. Fixes BlockTest::testInitializerFunctionsReturnCorrectBlock.
* In Title::checkUserBlock(): Convert message parameters to plain strings to make strict comparisons work. Fixes TitlePermissionTest::testUserBlock().
* ApiBlock.php: Follow-Up r85166, add 'Confirm' parameter to data array. Fixes ApiBlockTest::testMakeNormalBlock().
* ApiQueryTest.php: Use $wgMetaNamespace instead of $wgSitename to account for customizations.
* RandomImageGenerator.php: Use default value for constructor param to prevent fatals.
* ApiTestCaseUpload.php:
** Fix undefined variable $fileName.
** Remove stray ini_set statements, they don't belong there. Especially ini_set( 'error_reporting', 1 ) will effectively disable all reporting, which makes errors hard to track.
* ArticleTablesTest.php: Put testbug14404() in group Broken, as the fix for the bug has been reverted in r83868.
Added parser test. Added another parser test for what it seems to try to account for
but that in fact doesn't change when trying to achieve, since ids don't seem to be
affected when flipping $wgHtml5 and $wgExperimentalHtmlIds (so r68358 seems useless).
* Rewrote countRevisionsBetween/countAuthorsBetween to avoid assuming rev_id is in chronological order
* Made countAuthorsBetween use $limit+1 automatically for convenience; updated callers
* Work around HipHop issue 314 (volatile broken) and issue 308 (no compilation detection) by adding some large and ugly compilation detection code to WebStart.php and doMaintenance.php.
* Provide an MW_COMPILED constant which can be used to detect compiled mode throughout the codebase.
* Introduced wfIsHipHop(), which detects either compiled or interpreted mode. Used this to work around unusual eval() return value in eval.php.
* Work around lack of ini_get() in Maintenance.php, by duplicating wfIsHipHop().
* In Maintenance::shouldExecute(), accept "include" as an inclusion function name, since all kinds of inclusion give this string in HipHop.
* Introduced new class MWInit, which provides some static functions in the pre-autoloader environment.
* Introduced MWInit::compiledPath(), which provides a relative path for invoking a compiled file, and MWInit::interpretedPath(), which provides an absolute path for interpreting a PHP file. Used these new functions in the appropriate places.
* When we are running compiled code, don't include files which would generate duplicate class, function or constant definitions. Documented the new requirements on the contents of Defines.php and UtfNormalDefines.php.
* In HipHop compiled mode, it's not possible to have executable code in the same file as a class definition.
** Moved MimeMagic initialisation to the constructor.
** Moved Namespace.php global variable initialisation to Setup.php.
** Moved MemcachedSessions.php initialisation to the caller in GlobalFunctions.php.
** Moved Sanitizer.php constants and global variables to static class members. Introduced an accessor function for the attribs regex, as a new place to put code formerly at file level.
** Moved Language.php initialisation of $wgLanguageNames to Language::getLanguageNames(). Removed the global variable, marked "private" since forever.
* In two places: don't use error_log() with type=3 to append to a file, HipHop doesn't support it. Use file_put_contents() with FILE_APPEND instead.
* Work around the terrible breakage of class_exists() by using MWInit::classExists() instead in various places. In WebInstaller::getPageByName(), the class_exists() was marked with a fixme comment already, so I replaced it with an autoloader solution.
The length of this message varries by localization, so the previous solution of telling truncate to truncate 5 bytes
less than needed is not good since this will be too little or too much.
Updated places where its used. Some places I left as is, as it looked like the new behaviour would work fine for them to.
(for example, the autosummary feature - it was cutting off at 200 bytes, which is no where near 250 limit, so I presume that
was for asethic reasons rather then to fit as much in before the db limit).
Will do another commit for extension callers in a moment.
* Internalise $mAddress/$mUser, $mBy/$mByName, $mEnableAutoblock, $mId as getTarget(), getBlockers(), isAutoblocking(), getId().
* This required editing AbuseFilter and CheckUser backwards-incompatibly, so push the rest of the changes out to those extensions.
* Attack the evil 14-parameter constructor and gratuitously-confusing newFromDB( $notVeryImportantParameter, $moreImportantParameter)
* Reimplement the hack for bug 13611 in a slightly less fragile fashion; could still do with further cleanup, but then again the login frontend is its own can of worms... :S
* Remove transitionary getTargetAndType() and newFromTargetAndType() methods
* Some optimisation in parseTarget()
* Fix the broken phpunit test mentioned in r84251
This patch mostly impact Title:getLocalUrl() with two changes:
1) Abstraction:
$wgActionPath related code is now in Title::resolveActionPath(). It now
interprets query parameters using an array which is easier to read than
playing with a regexp and the evil $matches[].
2) Tweaking:
Change the 7 years old (r2621) URL forge:
"{$wgScript}?title={$dbkey}&{$query}"
=> /w/index.php?title=FOO&q=2
To a nicer
"str_replace( '$1', $dbkey, $wgArticlePath )"
=> /wiki/FOO?q=2
Additional changes:
* Parser tests were modified to reflect the changes.
* PHPUnit TitleTest amended to test getLocalUrl() (add more!!)
* When using wgActionPaths, makes sure the 'view' action is a sane default
Except for some whitespace changes, this is Baowolff's patch from http://bug-attachment.wikimedia.org/attachment.cgi?id=8312 which I am applying in the spirit of “apologize later”.
Baowolff writes:
Don't show new page entries in Special:newpage for supressed redirects.
Patch to remove the new page entry from Special:RecentChange and
special:newpages, if the new page gets moved with redirect supressed.
I'm not a 100% sure this is a good idea. This will only suppress the initial
new page line in the RC (And newpages), and won't supress any other edits to
the page. Does this behaviour really make sense? (Thats not rhetorical). It
seems somewhat different from deleting a page, where none of the content is
relevent, but when you move a page, the edits don't go away, they're just in a
new title.
I suppose we could just change the new page entry in the RC to point to the
other page, but that seems rather wrong too.
So anyhow, not sure what the right approach is.
But the cl_sortkey_prefix are kept to the right value for the cl_sortkey's that are set to the wrong value, totally messing things up.
End result is: Have a page, with some categories having different sortkeys, move that page, now all the categories have the same
sortkey, but still have their original prefix, so linksupdate things the sortkey is correct. End result is null edits don't fix
the sortkey, nothing but removing and re-adding category will reset sortkey to correct value.
** automatically add redirect=no to all links on the redirect page in Article::viewRedirect.
** properly check if redirects are enabled if $wgMaxRedirects < 1 (moved check from Title::newFromRedirectArray to Title::newFromRedirectInternal).
Title::getNotificationTimestamp() is called for every line in HistoryPager, to compare against the revision's timestamp to determine if it's a new edit since the viewer's last visit to the page.
The function kept an internal cache within the Title object to avoid having to make the query over and over in this situation, but in the case that it returns null, the cached entry would be accidentally ignored on the next round due to use of isset() to check for the array entry. (Most of the time it's great, but if the value of the key actually *is* null, isset() returns false.) Switched to array_key_exists(), which doesn't have this false negative. Now only one query gets issued for this on a history page, even if the answer is null.
* MessageCache::singleton() calls wfGetMessageCacheStorage() directly instead of using $messageMemc, just in case this would be called before that variable is set
* Per TimStarling: also removed deprecated methods in MessageCache class: addMessages() and related, [get|set|enable|disable]Transform(), loadAllMessages(), loadMessageFile() and some others. Same for the legacyData stuff in LocalisationCache that was only used by MessageCache::addMessages() and related.
* Converted remaining extensions
* Added a maintenance script which generates a list of first letters. Unified Han are omitted for performance, and because they shouldn't be used as headings anyway. A future collation specific to Chinese would provide the KangXi radicals as "first letters".
* Provided a precomputed list of first letters. Used Unicode 6.0.0 data and ICU 4.2.
* Moved collation functionality from Language to a Collation class hierarchy with factory function. Removed the recently-added methods from Language and updated all callers.
* Changed Title::getCategorySortkey() to separate its parts with a line break instead of a null character. All collations supported by the intl extension ignore the null character, i.e. "ab" == "a\0b". It would have required a lot of hacking to make it work.
* Fixed the uppercase collation to handle non-ASCII characters, redundantly with r80436. I don't think it's necessary to change the collation name as was done there, so I reverted that in the course of my conflict merge. A --force option to updateCollation.php might be nice though.
Additionally, *cough* *cough*:
* Add a Title::isMainPage helper for the fairly common $title->equals( Title::newMainPage() ); test.
* Update wfMessageFallback to also accept an array of message keys instead of requiring them listed as arguments to the function.
* Move the bulk of wfMessageFallback code into Message.php instead of leaving it in GlobalFunctions.php
* Change the wfMessageFallback implementation so that the Message class handles the fallbacks themselves eliminating any side effects caused by the fact that wfEmptyMsg always used usedb=false, language=userlang when one might actually use a different language or usedb setting in the message object that actually returned the text (this may be considered a wfEmptyMsg regression in 1.18).
* Make blank "" message contents fallback like nonexistant messages do.
* Re use the new tabAction array handling used to support mainpage-nstab in the talk and view tabs instead of making wfEmptyMsg calls directly in SkinTemplate.
Remove invalidateTitleProtectionCache().
Set $mTitleProtection to the proper value when we change it instead of forcing a reload via the protected_titles table.
Make Title::loadRestrictionsFromRows() private since it is only called by the private loadRestrictionsFromResultWrapper() and nobody should be fiddling with such title internals anyway (follow up to r54527).
Removed the invalidateTitleProtectionCache() from getCascadeProtectionSources() since that function only deals with titles which do exist. There's nothing to invalidate there.
has a % sign in it, but interperted as a + if no % sign is present.
This changes how the parser interperts links slightly. However:
*I can't imagine anyone is relying on this behaviour
*Things should be consistent. a + sign shouldn't magically change meaning
if there is a % sign somewhere else in the link.
*Pages are allowed to contain % signs in their title, and + signs,
you should be able to link to such pages just by typing there name without
resorting to %2B.
*If you have a page named foo%+ having [[{{PAGENAME}}]] link to a different
page seems inherently wrong.
*The previous behaviour seemed accidental.
* Don't set $IP. The entry point already sets it, so there's no point. Users can still set it if they can figure out a way to do it without breaking things.
* Don't include DefaultSettings.php. Doing that is code, not configuration, and it can easily be done in WebStart/Maintenance. Some non-standard entry points in extensions may be broken by this. That's their fault for being non-standard. Backwards compatibility is preserved thanks to require_once().
* Introduce $wgInvalidateCacheOnLocalSettingsChange, which when set, causes $wgCacheEpoch to be updated in the way that it previously was in the default LocalSettings.php.
* Don't set $wgLocalInterwiki to some nonsense value (possibly including spaces and punctuation in the new installer). It should be only for actual interwiki prefixes. Since most wikis don't have them, I set this to false by default and adjusted the referring code to accept this.
* Removed the guard for web invocation with $wgCommandLineMode set. This ancient code is redundant with modern protections in doMaintenance.php.
* In DefaultSettings.php, fixed fold terminator placement near $wgLoadScript
* Add a web entry point guard. That's one piece of code which really is necessary.
* Added static DifferenceEngine::intermediateEditsMsg() function
* diff-multi msg use by FlaggedRevs wasn't updated after $2 param was added. Fixed this.
* Minor cleanups to getMultiNotice()
2. Add more Accept-Language XVO cache name for IE;
3. Use hreflang to specify canonical and alternate links, it's search engine friendly
when a wiki has multiple variant languages.
