https://bugzilla.wikimedia.org/show_bug.cgi?id=54541

--- Comment #1 from Kunal Mehta (Legoktm) legoktm.wikipedia@gmail.com --- Created attachment 13369 --> https://bugzilla.wikimedia.org/attachment.cgi?id=13369&action=edit Copy of attachment from http://sourceforge.net/p/pywikipediabot/patches/621/attachment/login-site-us...

https://bugzilla.wikimedia.org/show_bug.cgi?id=54541

Andre Klapper aklapper@wikimedia.org changed:

What |Removed |Added ---------------------------------------------------------------------------- Keywords| |patch, patch-need-review Status|PATCH_TO_REVIEW |NEW

--- Comment #2 from Andre Klapper aklapper@wikimedia.org --- [No attachment in Gerrit, hence resetting status and adding keyword. See https://bugzilla.wikimedia.org/describekeywords.cgi and https://www.mediawiki.org/wiki/Bug_management/Bug_report_life_cycle for more information.]

https://bugzilla.wikimedia.org/show_bug.cgi?id=54541

John Mark Vandenberg jayvdb@gmail.com changed:

What |Removed |Added ---------------------------------------------------------------------------- CC| |jayvdb@gmail.com

--- Comment #4 from John Mark Vandenberg jayvdb@gmail.com --- The is an interesting approach which I am guessing solves part of the 'SUL' problem.

Instead of having to add usernames['wikipedia']['*'] = '..'; usernames['wikisource']['*'] = '..'; to user-config.py etc etc for every family in the matrix of the global account.

If a account is logged into it.wp using an entry in user-config.py, and the bot visits en.wikipedia, the bot *is* logged in, as they have a unified account, and the cookie logs them in, and api.py automatically populates userinfo with the username.

That works wonderfully, and doesnt need this patch.

I am guessing this patch is about when there is a 'problem' with the existing login, which might be because of site.py tripping over itself (the login code is a mess), or maybe the server asking the account to login again.

If the password_file contains simple two-part matching entry for the username in siteinfo, the re-login will be automatic. i.e. (bot_username, bot_password) ; not (family, username, password) or other options.

I would like to hear more about how this patch helps, but it does appear useful in limited but crucial circumstances.