356362 - Feed date parser breaks on news.google.com's atom feed's out-of-spec dates

Status: RESOLVED WONTFIX

(Firefox Graveyard :: RSS Discovery and Preview, defect)

Description

[Manish Singh]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.0.7) Gecko/20060830 Firefox/1.5.0.7 (Debian-1.5.dfsg+1.5.0.7-2)
Build Identifier: 

news.google.com's atom feed has dates like:

2006-010-12T00:16:45+00:00

Note the 3 digit month: "010"

While this is clearly a spec violation, I think the feed code should be as forgiving as reasonably possible, especially for a popular site like google news.

Reproducible: Always

Comment 1

[Manish Singh]

Attachment: Patch to change the regexp so 3-digit months match

This new regexp handles this. I changed the rest of the regexp to use curly brace numerical matching just for consistency.

Comment 2

[Robert Sayre]

(In reply to comment #0)
> While this is clearly a spec violation, I think the feed code should be as
> forgiving as reasonably possible, especially for a popular site like google
> news.

Things to consider:

How IE 7 behaves.
How Safari behaves.
How other popular aggregators behave.

Willing to take a fix here, but not if we end up creating an Atom profile.

Comment 3

[Robert Sayre]

Attachment: atom feed from Google News to test

Comment 4

[Robert Sayre]

OK, Safari deals with that.

Comment 5

[Robert Sayre]

NetNewsWire shows the wrong date.

"01/01/06 12:00am". 

Comment 6

[Manish Singh]

IE7 doesn't handle this.

Comment 7

[Manish Singh]

Safari looks like it doesn't handle it either, and substitutes the current date instead. (OS X 10.4.8)

Comment 8

[Robert Sayre]

(In reply to comment #7)
> Safari looks like it doesn't handle it either, and substitutes the current date
> instead. (OS X 10.4.8)
> 

Oh yeah, you're right. Thunderbird will probably do that too.

Comment 9

[Robert Sayre]

Hmm, universal feedparser add support for this yesterday.

http://feedparser.cvs.sourceforge.net/feedparser/feedparser/feedparser.py?r1=1.141&r2=1.142

"support for google dates". Hehe.

Thinking this over, we should take this. We're not particularly strict with date formatting anyway, since we accept RFC822, IS08601, and Unix timestamps in all date fields in all formats. 

r=sayrer if you add unit tests for RSS1, RSS2, and RFC4287.

Comment 10

[Manish Singh]

Do you mean specific unit tests for "google dates" or ISO8601 dates in general?

Comment 11

[Robert Sayre]

Google News doesn't serve this junk anymore. I don't feel strongly about not taking this patch. Reopen if you still want to land this w/ test case.

Comment 12

[Robert Sayre]

(In reply to comment #10)
> Do you mean specific unit tests for "google dates" or ISO8601 dates in general?
> 

I meant "google dates". Again, feel free to reopen.