221669 - [FIXr]xul-overlay processing instruction pointing to inexistent file make mozilla enter in some kind of loop

Status: RESOLVED FIXED

(Core :: XUL, defect, P1)

Description

[Cedric. S.]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4) Gecko/20030624
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.4) Gecko/20030624

referencing an inexistent file in a <?xul-overlay ...> Processing Intruction in
a xul file make mozilla unable to load this xul file : mozilla never stops to
load it in fact.

Reproducible: Always

Steps to Reproduce:
drag&drop the linked xul page into the browser. 



Actual Results:  
It looks like it never finishes to download it. (Sandglass indefinitely, hit
ESCAPE to stop).

Expected Results:  
retry without the 2nd line, it's ok.

<?xml version="1.0"?>
<?xul-overlay href="inexistent-file.xul"?>

<window
  xmlns="http://www.mozilla.org/keymaster/gatekeeper/there.is.only.xul"
  >
  
</window>

Comment 1

[Boris Zbarsky [:bzbarsky]]

The XUL code uses Open() to load overlays.  Is the original XUL page in chrome?
 a file:// url?  An http:// url?  That is, what's the base URI for the relative
URI resolution?

Also, the implementation of Open() got heavily changed recently, so testing with
anything but a very recent nightly (like today's) is not very useful at this
point...

Comment 2

[Boris Zbarsky [:bzbarsky]]

.

Comment 3

[Cedric. S.]

the bug also occurs in Mozilla 1.5 (release)

Comment 4

[Cedric. S.]

The bug still can be reproduced in Moz 1.6 (realase)

Comment 5

[Michiel van Leeuwen (email: mvl+moz@)]

I did this with a recent build. I tried to import a overlay from wallet into
calendar. Works fine, expect when wallet isn't there. And firefox/thunderbird
don't have it.

<?xul-overlay href="chrome://wallet/content/walletTasksOverlay.xul"?> is the
thing that fails. so, to answers bz's question: it's a chrome url, and a recent
build.

Comment 6

[Boris Zbarsky [:bzbarsky]]

Attachment: This should fix the problem

Comment 7

[Ben Goodger (use ben at mozilla dot org for email)]

Comment on attachment 153010 [details] [diff] [review]
This should fix the problem

r=[e-mail]

Comment 8

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 153010 [details] [diff] [review]
This should fix the problem

jst, would you sr?

Comment 9

[Boris Zbarsky [:bzbarsky]]

Taking.

Comment 10

[Johnny Stenback (:jst)]

Comment on attachment 153010 [details] [diff] [review]
This should fix the problem

- In nsXULDocument::ReportMissingOverlay():

+    nsresult rv;
+    nsCOMPtr<nsIConsoleService> consoleService =
+	 do_GetService(NS_CONSOLESERVICE_CONTRACTID, &rv);
+    if (NS_FAILED(rv))
+	 return;
+    nsCOMPtr<nsIScriptError> errorObject =
+	 do_CreateInstance(NS_SCRIPTERROR_CONTRACTID, &rv);
+    if (NS_FAILED(rv))
+	 return;
+    nsCOMPtr<nsIStringBundleService> stringBundleService =
+	 do_GetService(NS_STRINGBUNDLE_CONTRACTID, &rv);
+    if (NS_FAILED(rv))
+	 return;
+    nsCOMPtr<nsIStringBundle> bundle;
+    rv = stringBundleService->CreateBundle(
+	    "chrome://global/locale/xul.properties", getter_AddRefs(bundle));
+    if (NS_FAILED(rv))
+	 return;
+
+    nsCAutoString spec;
+    aURI->GetSpec(spec);
+
+    NS_ConvertUTF8toUTF16 utfSpec(spec);
+    const PRUnichar* params[] = { utfSpec.get() };
+
+    nsXPIDLString errorText;
+    rv =
bundle->FormatStringFromName(NS_LITERAL_STRING("MissingOverlay").get(),
+				       params, NS_ARRAY_LENGTH(params),
+				       getter_Copies(errorText));
+    if (NS_FAILED(rv))
+	 return;

Geez, I mean, could this get any more verbose? If there isn't one already,
wanna file a bug to make a non-pathetic API for our string bundle code (in
nsContentUtils for layout etc to use, if nowhere else), I mean, this is just so
unbelieveably sad.

sr=jst nonetheless tho :-)

Comment 11

[Boris Zbarsky [:bzbarsky]]

Dunno about string bundles, but bug 228205 should make the console service api
much easier to use (in particular, it should hide all the intl crap under the
console service).

I can still file a bug on a better stringbundle api if you want, though.

Comment 12

[Johnny Stenback (:jst)]

Yeah, I think it'd be worth overhauling how we use the stringbundle service etc
too to hide some of its bloatiness behind a less bloaty API, since we use that a
fair bit for non-console service purposes as well. So please file a followup bug
on that (and CC me).

Comment 13

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

I think that patch should have nsXULDocument.h changes somewhere (now that it
broke the tree). :-)

Comment 14

[Boris Zbarsky [:bzbarsky]]

Checked in for 1.8a3 or 1.8b or whatever this next milestone is.

Comment 15

[Boris Zbarsky [:bzbarsky]]

And yes, I did check in the nsXULDocument.h change to fix bustage (just declares
the function in the obvious way).