Page MenuHomePhabricator

allow protocol-relative URLs in $wgCaptchaWhitelist
Closed, ResolvedPublic

Description

A IP of de.wiki was telling, that he get a CAPTCHA for a edit, where only a comma is changed[1].

In my opinion is this a problem with protocol-relative URLs.

The page includes a fullurl to commons. Before the edit the database contains a http link to commons, after the reparse due to the edit the link to commons is changed to a protocol-relative URL and ConfirmEdit is seeing a different and trigger the CAPTCHA -> bad in this case.

We can fix this by editing all templates with fullurl to trigger a reparse with the Job Queue, but that sounds like many load for the server and also fixed that problem only on de.wiki.

[1] http://de.wikipedia.org/w/index.php?title=Simone_de_Beauvoir&diff=94101544&oldid=92622886


Version: unspecified
Severity: major

Details

Reference
bz31294
TitleReferenceAuthorSource BranchDest Branch
Write test exercising the case where eagerlyEvaluate hits a recursion limit.repos/abstract-wiki/wikifunctions/function-orchestrator!132apineapine-recursionmain
Write test exercising the case where eagerlyEvaluate hits a recursion limit.repos/abstract-wiki/wikifunctions/function-orchestrator!131apineapine-test-eagernessmain
Don't use RequestExceptionrepos/data-engineering/mediawiki-event-enrichment!52ottoexception0main
Fix bug in enrich_page_change_with_content where null record would be returnedrepos/data-engineering/mediawiki-event-enrichment!48ottofix0main
Use eventutilties_python stream_manager parameterizationrepos/data-engineering/mediawiki-event-enrichment!47ottoparamsmain
Issue async requests from EventProcessFunction.repos/data-engineering/eventutilities-python!40gmodenaT332948-minibatch-async-callmain
Window non keyed datastreamrepos/data-engineering/mediawiki-event-enrichment!40gmodenawindow-non-keyed-datastreamdev
Customize query in GitLab

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 21 2014, 11:51 PM
bzimport set Reference to bz31294.

Juergen wrote:

Shouldn't we allow external links to commons from IP Users without a captcha, even when they are really new and not caused by protocol-relative URLs as explained by Umherirrender ?

This mechanism seems to be working already in the case of external links to WP (for instance for links like the diff-link from the description).

Links to Commons already allowed, when there are containing http or https at the begin. See $wgCaptchaWhitelist in http://noc.wikimedia.org/conf/CommonSettings.php.txt

$wgCaptchaWhitelist needs changing for protocol-relative URLs
-> Using this as shell bug, changing summary and component.

Done:

  • $wgCaptchaWhitelist = '#^https?://([.a-z0-9-]+\\.)?((wikimedia|wikipedia|wiktionary|wikiquote|wikibooks|wikisource|wikispecies|mediawiki|wikimediafoundation|wikinews|wikiversity)\.org|dnsstuff\.com|completewhois\.com|wikimedia\.de|toolserver\.org)(/|$)#i';

+ $wgCaptchaWhitelist = '#^(https?:)?//([.a-z0-9-]+\\.)?((wikimedia|wikipedia|wiktionary|wikiquote|wikibooks|wikisource|wikispecies|mediawiki|wikimediafoundation|wikinews|wikiversity)\.org|dnsstuff\.com|completewhois\.com|wikimedia\.de|toolserver\.org)(/|$)#i';