115642 - [BUGZILLA] oopsbot should use XML to get bug information

Status: VERIFIED FIXED

(Webtools Graveyard :: Mozbot, defect)

Description

[Jacob Steenhagen]

Bugzilla.bm should use XML instead of trying to parse the HTML.  Right now,
xml.cgi exists.  However, due to bug 109530 the results of using it aren't
pretty.  This bug has been fixed in CVS and as soon as b.m.o updates, it'll make
those special characters appear correctly in ssdbot's output.

However, the way it's done right now is only half-way.  Because there is no way
to get query results in XML it still has to parse HTML for the query interface:

ssdbot: bugs &foo=bar&baz=snafu

Once bug 103778 is fixed, the fix for bug 48707 should be trivial (Myk's words)
which will allow Bugzilla.bm to go completely XML.

Comment 1

[Jacob Steenhagen]

BTW, my current implimentation is to use the XML::LibXML package from CPAN which
is much easier to work with the the expat based XML::Parser.

Comment 2

[Hixie (not reading bugmail)]

Pushing out to 2.6 because I'd like the relevant Bugzilla version to have been
released well in advance of the relevant mozbot release.

Comment 3

[Hixie (not reading bugmail)]

*** Bug 133202 has been marked as a duplicate of this bug. ***

Comment 4

[Bradley Baetz (:bbaetz)]

Taking. The patch I'll attach is based on Jake's.

Comment 5

[Bradley Baetz (:bbaetz)]

Attachment: patch

Jake's version which zach emailed me was not basde against current cvs. I don't
think I mismerged anything...

Comment 6

[Bradley Baetz (:bbaetz)]

That patch makes the "n bugs found" announcment even when it shouldn't - it
should only do so if there was more than one bug and we weren't using the single
query xml.cgi stuff. Is that correct?

Comment 7

[Bradley Baetz (:bbaetz)]

Attachment: v2

Fixed that bug. cvs mozbot gets this wrong, btw:

bug 1 bug 2

makes it say '2 bugs found'

Comment 8

[Bradley Baetz (:bbaetz)]

Hixie, review?

Comment 9

[Hixie (not reading bugmail)]

That's a big patch, it will take me a while to review, and I'm rather busy with
university at the moment. I'll definitely take it for the next release, don't
worry. (I also need to make sure it plays well with any fixes for the other
Bugzilla module bugs.)

Comment 10

[Bradley Baetz (:bbaetz)]

*** Bug 145384 has been marked as a duplicate of this bug. ***

Comment 11

[Robert Kaiser]

thebot on irc.mozilla.org (#mozilla.de, #mozbot, #ui, #mozdev, #mozl10n) now
uses the Bugzilla.pm module from cwbot that has bbaetz's patch.

With the Bugzilla module from current mozbot CVS, I didn't get Bugzilla to work,
but with bbaetz's module it works again...

Comment 12

[Christian Reis]

Robert, you can also check out Bug 145384 which has an alterate approach to the
problem (making the HTML parsing stronger) thus allowing it to work on any
version of Bugzilla I've seen so far.

Comment 13

[Robert Kaiser]

kiko: I've seen that bug. Anyway, I like the XML approach better because XML
should be the format to exchange data between different formatting interfaces
(and I would consider mozbot to be such an interface), HTML should only be the
format to present data to a web browser (IMO).

Comment 14

[Bradley Baetz (:bbaetz)]

Attachment: v2.1

Truncation of large responses was broken in a channel - this fixes that thinko.

Comment 15

[Jacob Steenhagen]

Comment on attachment 93281 [details] [diff] [review]
v2.1

Had someone in #mozwebtools asking for this patch today... I looked it over and
it all looks sensible to me (of course, some of the code is mine so I can't
really r= it :).

Making this an "official" review request.

Comment 16

[Christian Reis]

I still insist we look at the alternate solution I implemented in Bug 145384
which is simpler, less intrusive, and makes mozbot work with *many*  versions of
Bugzilla (since I tested it myself in a number of them). What good is releasing
a version of mozbot that requires you to upgrade *Bugzilla* to get it working?

Comment 17

[Bradley Baetz (:bbaetz)]

kiko - this version works with old bugzillas, using html as a fallback. That
triggers on teh contenttype for buglist.cgi, and rdf output was added at the
same time as the html changed, so it should always work

I do think that we should have both patches, though, just for robustness

/me pokes hixie for review ;)

Comment 18

[Hixie (not reading bugmail)]

I will not be reviewing this until I have a chance to test it first-hand, which
won't happen before I have a Linux installation, which for various technical
reasons is not likely to happen before the end of the year. Sorry.

Comment 19

[Dave Miller [:justdave]]

Hixie, it's past the end of the year.  What's the status? :)  Had people on IRC
asking for this again.

Comment 20

[Dave Miller [:justdave]]

Attachment: v2.2

This is what ssdbot is currently using.

Comment 21

[Dave Miller [:justdave]]

Hixie told us the Bugzilla folks now had complete control over the Bugzilla.bm
module.  This has been in use on ssdbot (and other mozbots) for several months
now and works just dandy.  Calling somewhere between 6 and 15 months worth of
production use good enough for a review and checking it in. :)

Checking in BotModules/Bugzilla.bm;
/cvsroot/mozilla/webtools/mozbot/BotModules/Bugzilla.bm,v  <--  Bugzilla.bm
new revision: 2.7; previous revision: 2.6
done

Comment 22

[Bradley Baetz (:bbaetz)]

Yeah, I'd been meaning to check this in for ages. There are some bugs it doens't
parse, though - I'll have to look through irc logs for details.

Thanks for doing this for me. (I'll doubel check that that is what I actually
have locally, mind you)

Comment 23

[Tanner M. Young [:tmyoung]]

Verified Fixed.