10712 - [PP]RDFRemoteDataSource chokes on resource names with Windows file: URLs

Status: VERIFIED FIXED

(Core Graveyard :: RDF, defect, P3)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Test cases to be attached:
  * rdftest.html
  * rdftest.rdf (when you download it, keep this filename, and put the 2 in the
same directory)

Load rdftest.html, and you will see that RDFRemoteDataSource gives error
messages like this on Windows, but not Linux:

Browser Window Alert: Caught exception [[Component returned failure code: 0x8000
ffff [nsISimpleEnumerator.HasMoreElements, {file: file:///e|/rdftest.html, line:
 87}]]]

JavaScript error: uncaught exception: Component returned failure code: 0x8000fff
f [nsIRDFContainer.AppendElement, {file: file:///e|/rdftest.html, line: 108}]

The problem comes from changing the test case for bug 9236 to *correctly* use
RDF:Bag id="" rather than about="" so that the resource name (used for
GetResource()) includes a file: URL.  I think this could have something to do
with ":" or "|" in file: URLs or something like that, since it happens on
Windows but not Linux.

If this is confusing, I'm tired.  And a bit confused and disoriented from
working under a now-unfamiliar operating system.  Ask me tomorrow :-).

Fails on the following windows builds:
 * July 25 apprunner
 * July 27 apprunner
 * July 27 viewer.

I've yet to verify that the test case itself works on Linux (I will when I
reboot), but I've seen the more complex case work/fail for the same reasons
depending on platform.  (Bad sentence, but you know what I mean.)

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: rdftest.html

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: rdftest.rdf

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Guess what?  That test case gives the exact same errors on Linux.

So now I've got to try and find something that works on Linux and fails on
windows.  But maybe the failing on Linux is a bug too...???

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: bugtest.html

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: bugtest.rdf

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

The above tests (bugtest.*) work under Linux and fail in Windows.

Comment 7

[Chris Waterson]

Move to M10.

Comment 8

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

This seems fixed in 1999-08-15-09-M9.

Comment 9

[johng]

Since this seems fixed in 8/15 M9 .... Phillip, can you verify so we can
take this off of Waterson's list for M11?

Reassigning to Phillip.

Comment 10

[phillip]

sorry, that's not how things are done around here. reassigning to waterson...

Comment 11

[phillip]

marking resolved fixed, since it is.

i'll verify when i get a stable linux build.

Comment 12

[phillip]

Attachment: stacktrace on linux 2.2

Comment 13

[phillip]

looks like we spoke too soon. this crashes all platforms:

1. choose either of the testcases above
2. in each case, the described alert appears, as it should not.
3. in each case, dismissing the alert crashes apprunner, as it should not.

this crash occurs on
     1999-09-14-09-M10 RedHat Linux 6.0 (GNOME/enlightenment)
     1999-09-14-09-M10 WinNT 4.0 sp5
     1999-09-14-09-M10 MacOS 8.51

reopening bug.

Comment 14

[Chris Waterson]

Okay, some comments:

1. rdftest.html and rdftest.rdf are _not_ compatible. The HTML file expects the
root resource to be "WhenCOM:Something"; the RDF file incorrectly uses a
lower-case "id=" to specify a document-relative resource ID. So that test case
is INVALID.

2. If you SAVE THE FILES LOCALLY, this does _not_ crash; it asserts.

Using the second set of test cases (bugtest.html and bugtest.rdf) seems to work
JUST FINE.

If you attempt to load the files thru bugsplat THIS WILL NOT WORK.

Comment 15

[phillip]

ok, my fault. i saved the files locally, but viewed them using http:// instead
of file://

sorry for all the noise, marking verified.