[wp-trac] [WordPress Trac] #11219: could not be converted to UTF-8 / WordPress should cache failed feed fetches so as to avoid overloading feed sources

[WordPress Trac]

#11219: could not be converted to UTF-8 / WordPress should cache failed feed
fetches so as to avoid overloading feed sources
-------------------------------+--------------------------------------------
 Reporter:  Denis-de-Bernardy  |       Owner:                          
     Type:  defect (bug)       |      Status:  new                     
 Priority:  high               |   Milestone:  2.9.1                   
Component:  Optimization       |     Version:  2.9                     
 Severity:  major              |    Keywords:  has-patch tested blocker
-------------------------------+--------------------------------------------
Changes (by Denis-de-Bernardy):

  * keywords:  has-patch tested reporter-feedback => has-patch tested
               blocker


Comment:

 Replying to [comment:19 westi]:
 > Replying to [comment:18 Denis-de-Bernardy]:
 > > I hate to be punting this back to 2.9.1, but this ticket is the main
 reason that RSS widgets have caused so much trouble in WP 2.9. It *should*
 get fixed before we even think about rushing WP 2.9.1 in the wild; or at
 the very least, we should have an acceptable workaround.
 >
 > In order to understand this in more detail are you able to provide
 answers to the following questions:
 >
 >  * Is this a behavioural change between SimplePie 1.1.3 and SimplePie
 1.2 that means 2.9 is worse than 2.8?

 yes

 >  * Does a 2.8 install choke in the same way?

 no

 >  * Is there an easy step by step reproduction method for the issue?

 yes. I've been screaming how to reproduce this over and over further up.
 disable multibyte functions, and disable iconv. create a new site, using
 the UTF8 charset, and watch -- with amusement -- how the dashboard chokes
 in WP 2.9.

 >  * Have you tried SimplePie 1.1.3 + WP 2.9 to see if it helps?

 no, I haven't. the only thing I do know is where it comes from in WP 2.9:
 when SimplePie seeks to "convert" the UTF8 feed into UTF8, and finds
 neither of multibyte functions nor iconv, it fails miserably -- in spite
 of the fact that it technically has nothing to convert.

 so again, what my patch does is only the following: it adds a use case
 *after* it tries MB functions and iconv, which basically goes: if input
 charset = output charset, then don't convert a thing and just return the
 original data.

 I'll happily buy that there might be security implications if we accept a
 UTF8 string *as is*, so maybe we could try to at least validate the mess
 if we've the available tools and functions. the main difference between my
 first and second patch, in fact, is that I've simply placed the code
 further down in the convert function.

 Still, the point stands: SimplePie and WP should not fail, as it currently
 does, when trying to convert data from a charset to that same charset when
 there are neither of MB function and iconv.

-- 
Ticket URL: <http://core.trac.wordpress.org/ticket/11219#comment:20>
WordPress Trac <http://core.trac.wordpress.org/>
WordPress blogging software