159842 - parser has no mechanism to allow parsing of incomplete html fragments

Status: VERIFIED FIXED

(Core :: DOM: HTML Parser, defect)

Description

[Joe Francis]

There is no way for a client of the parser to communicate to the dtd, via the
parser, that it wants to parse a fragment of html that may not have all required
parents.  /for example, if you wish to parse a fragment that has tr's and td's
but no table, the dtd will drop the table substructre on the floor.  This causes
problems with copy/paste.  

Attaching a fix soon.  I'll mark ths as a blocker of some other bugs once I
track them down.

Comment 1

[Joe Francis]

Attachment: parser patch

Comment 2

[Joe Francis]

Attachment: take #2

first patch not quite intrusive enough.

Comment 3

[Akkana Peck]

What's the status of this patch and the one in bug 94176 (blocked by this one)?
Are you waiting for reviews, or do the patches need further work?

Comment 4

[Joe Francis]

Attachment: take #3

This includes some missing header file changes from prior patch, and also
updates content to have a CID/ContractID/Factory method for the
HTMLFragmentSink, which is the sink most folks would want to use with the
parser when they are parsing an html fragment.

Comment 5

[Joe Francis]

ooops, forgot to add content changes.  hold the phone....

Comment 6

[Joe Francis]

Attachment: content changes to export the HTML Fragment Sink

This does not obsolete the prior patch.  Use both.

Comment 7

[Joe Francis]

Attachment: parser diffs ignoring ws

This is the same diff as "take #3", but ignoring whitespace to aid reviewers. 
There is a lot of gobbledeegook in "take#3" that is really just diff getting
confused by a bunch of lines that were indented to be inside a new "if" block.

Comment 8

[Joe Francis]

marking bugs blocked by this

Comment 9

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Comment on attachment 96474 [details] [diff] [review]
parser diffs ignoring ws

r=heikki, with the suggestion that you make SetCommand(const char*) also
recognize the new command.

It would be great if you could run the parser regression tests before checkin
in (command line tool, see README in htmlparser/tests/html). Although I have to
say this patch alone looks pretty safe.

Also, if this is not terribly urgent I would advice you to wait till Monday to
get sr from bzbarsky (I think he had something to do with fragment suff).

Comment 10

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 96474 [details] [diff] [review]
parser diffs ignoring ws

sr=bzbarsky if you make NS_FRAGMENTSINK_CONTRACTID be
NS_HTMLFRAGMENTSINK_CONTRACTID (because we may have an XML fragment sink at
some point soonish)

Comment 11

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 96376 [details] [diff] [review]
content changes to export the HTML Fragment Sink

sr=bzbarsky

Comment 12

[Joe Francis]

fix (with suggested tweaks from reviewers) landed on trunk.  qa: i suggest
verifying this by testing 94176.  If that's fixed, this is fixed.

Comment 13

[Charles Rosendahl]

Please verify this bug based on comment #12

Comment 14

[rubydoo123]

per Joe's comment #12: I tested and verified bug 94176, so I am also marking
this verified.