303522 - !!! Object's nsIClassInfo lies about it's interfaces!!! classname: HTMLDocument unimplemented interface name: nsIDOMXPathEvaluator

Status: RESOLVED FIXED

(Core :: XML, defect)

Description

[timeless]

!!! Object's nsIClassInfo lies about it's interfaces!!!
   classname: HTMLDocument
   contractid: <unknown>
   unimplemented interface name: nsIDOMXPathEvaluator

Comment 1

[Peter Van der Beken [:peterv]]

*** Bug 303521 has been marked as a duplicate of this bug. ***

Comment 2

[Peter Van der Beken [:peterv]]

Timeless: this is a useless bug report. We need way more details (you might want
to read https://bugzilla.mozilla.org/page.cgi?id=bug-writing.html).

Comment 3

[timeless]

background: various dlls randomly refused to load when i was performing this
round of testing. the issue is presumably whatever adds support for xpath does
it unconditionally for classinfo and conditionally for qi.

Comment 4

[timeless]

so here's a short explanation:

http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/content/base/src/nsDocument.cpp&rev=3.564&mark=715-739#692

dynamically /tries/ to add tearoff support for xpath.

http://bonsai.mozilla.org/cvsblame.cgi?file=mozilla/dom/src/base/nsDOMClassInfo.cpp&rev=1.291&mark=1565-1567,1613-1621,1673,1677,1795,1799,2231,2235,2319,2324,2383,2387#1790

statically claims to always support xpath, even though there's code that clearly
decides it sometimes doesn't.

Comment 5

[Peter Van der Beken [:peterv]]

Yes, well, I'm not going to add code that instantiates an object at startup.
Checking for the ContractID is good enough, if actually instantiating that
component fails then that's too bad.

Comment 6

[Peter Van der Beken [:peterv]]

And DEBUG_CheckClassInfoClaims should check for NS_ERROR_NO_INTERFACE imho, any
tearoff that fails to instantiate (for example due to out-of-memory) will
trigger that printf.

Comment 7

[timeless]

i figure that we should be able to use one of these two info bits:
 721                      gCheckedForXPathDOM = PR_TRUE;
 722                      gHaveXPathDOM = (evaluator != nsnull);

1615                       PRBool haveXPathDOM;
1616                       cr->IsContractIDRegistered(NS_XPATH_EVALUATOR_CONTRACTID,
1617                                                  &haveXPathDOM);
1618                       if (haveXPathDOM) {
1619                         xpathEvaluatorIID = &NS_GET_IID(nsIDOMXPathEvaluator);

Comment 8

[timeless]

i've posted a bug and patch for oom. the whole *point* of the function is to
complain about people who declare classinfo and then do something like throw
nointerface.

note that there's an ASSERT version of this code which means eventually jst will
crash with a message telling him to fix either classinfo or queryinterface for this.

Comment 9

[Peter Van der Beken [:peterv]]

Attachment: v1

Whatever.

Comment 10

[Johnny Stenback (:jst)]

Comment on attachment 191744 [details] [diff] [review]
v1

Sure, r+sr=jst