Commit graph

11 commits

Author SHA1 Message Date
Tim Starling
4b04f1a2eb Fixed excessive memory usage in PCRE in wfMangleFlashPolicy(). Was causing OOM on various big pages. Possible upstream regression, deserves investigation. 2008-05-30 14:58:29 +00:00
Siebrand Mazeland
79d5225c0e * remove end of line whitespace
* remove empty lines at end of file
* remove "?>" where still present
2008-04-14 07:45:50 +00:00
Tim Starling
377561cd1e * Introduced OutputPage::addWikiMsg() and OutputPage::wrapWikiMsg(), to make it easier to avoid double-parsing. Double-parsing introduces differences between the old and new parsers.
* Used the new functions in the relevant places throughout the core. Where this was not possible, used wfMsgNoTrans() to avoid double-parsing.
* Calling wfEscapeWikiText() on a title before passing it to a message is unnecessary and causes bugs. The best idea is to let the message do its own escaping. Removed wfEscapeWikiText() from the caller of addedwatchtext, removedwatchtext, deletedtext, protect-text, nosuchusershort and movepage-moved. Updated all message files to use <nowiki>$1</nowiki> where appropriate in these messages. 
* Removed unnecessary empty message handling from semiprotectedpagewarning
* Removed nonsense warning about the output of wfMsg() not being safe for inclusion in HTML. 
* Changed wfShellExec() to use passthru/ob_get_contents instead of exec(). This avoids corruption of whitespace in the response. 
* Added $wgValidateAllHtml feature, to optionally validate HTML on output. 
* Fixed missing newline in cascadeprotected in MessagesEn.php
* Belatedly updated addedwatchtext to take account of the new "monobook" skin. No unwatch link in the sidebar anymore.
2008-02-18 07:25:35 +00:00
Tim Starling
55bd3a0279 Add an X-Vary-Options header, as per my proposed squid patch 2008-02-08 07:12:38 +00:00
Brion Vibber
455feda3ac * (bug 11795) Be more paranoid about confirming accept-encoding header is present 2007-10-29 17:41:55 +00:00
Brion Vibber
f4080c1838 Don't do gzip compression if the URL path ends in .gz or .tgz
This confuses Safari and triggers a download of the page,
even though it's pretty clearly labeled as viewable HTML.
Bad Safari! Bad!

Bug is still present in Safari 2.0.4/Mac and 3.0.2/Win.

We have had a live hack for this on Wikimedia sites for .gz files;
hadn't noticed the .tgz problem before though.

While I was in there, also made a tweak so wfGzipHandler() makes
sure the Vary: header is there whether the current agent sent us
an Accept-Encoding header or not.
2007-07-19 19:06:32 +00:00
Aryeh Gregor
a15c419b3d Remove ?>'s from files. They're pointless, and just asking for people to mess with the files and add trailing whitespace. (Yes, I looked over every one and reverted those that were bogus. Slash-enter a million times in less worked well enough, although it was a bit mind-numbing.) 2007-06-29 01:19:14 +00:00
Brion Vibber
0757f0a92e Fix regression: typo caused 'Vary: Accept-Encoding' to be sent overwriting more exact Vary lines such as the Vary: Accept-Encoding, Cookie used for squid mode 2007-02-20 05:04:36 +00:00
Brion Vibber
94e8f63eb7 Don't suppress gzip mode for zero-length output; failing to send the
Content-Encoding header with 304 responses confuses some browsers.
2007-02-20 04:58:54 +00:00
Brion Vibber
62ef8829fb * Further fix to Special:Export etc; instead of trying to suppress those
Content-Encoding and Content-Length headers, just don't send them 
  in the first place!
* Suppress Content-Length from wfOutputHandler if $wgDisableOutputCompression set
* Fix typo that caused $wgDisableOutputCompression to *enable* instead of *disable* compression ;)
2007-02-20 04:46:07 +00:00
Tim Starling
28dc3ec888 * Moved the main ob_start() from the default LocalSettings.php to WebStart.php.
The ob_start() section should preferably be removed from older 
  LocalSettings.php files.
* Give Content-Length header for HTTP/1.0 clients.
* Partial support for Flash cross-domain-policy filtering. Text entry points should be protected, but uploads are not.
2007-02-19 23:03:37 +00:00