298298 - Google Maps fails completely on Firefox 1.0/Solaris 10

Status: RESOLVED DUPLICATE of bug 297750

(Firefox :: General, defect)

Description

[Blake Jones]

User-Agent:       Mozilla/5.0 (X11; U; SunOS i86pc; en-US; rv:1.7.5) Gecko/20041109 Firefox/1.0
Build Identifier: Mozilla/5.0 (X11; U; SunOS i86pc; en-US; rv:1.7.5) Gecko/20041109 Firefox/1.0

Maps.google.com completely fails to work for me using Firefox 1.0 on Solaris 10 
on x86.

Going to http://maps.google.com/ directly gives me a gray map window (see http:/
/foo.net/~blakej/tmp/bugs/ff-1/1.jpg).

Entering in a search term (such as [san francisco, ca]) gives the world map tile
d four times vertically (see http://foo.net/~blakej/tmp/bugs/ff-1/2.jpg).  I can
 pan this map left and right, with click-and-drag or the arrow buttons or the ar
row keys, but I can't pan it up and down (with any of those three methods).  I c
an zoom in and out, but it stays centered around northern Canada/Greenland, and 
I still can't pan up or down with it zoomed in.  The zoomed-in map still has the
 map tiled four times vertically (see http://foo.net/~blakej/tmp/bugs/ff-1/3.jpg
).

Clicking on the "Satellite" link causes the browser to become unresponsive.  Aft
er ten or so seconds, a dialog box pops up pointing out that "A script on this p
age is causing mozilla to run slowly. [...]"  As long as I click "Cancel" (i.e. 
don't abort the script), the browser continues to be unresponsive.

I grabbed some stack traces of the browser (using pstack) while it was being unr
esponsive.  Unsurprisingly, it looks like it's working on JavaScript processing.
  I've saved the stack traces at http://foo.net/~blakej/tmp/bugs/ff-1/stacks.txt
 .  I also ran these through "c++filt", to make the C++ symbols easier to read; 
that's saved at http://foo.net/~blakej/tmp/bugs/ff-1/filtered.txt .

The JavaScript console does show one error on the page:

    Error: [Exception... 
      "'Permission denied to get property XULElement.selectedIndex' 
       when calling method: [nsIAutoCompletePopup::selectedIndex]"
      nsresult: "0x8057001e (NS_ERROR_XPC_JS_THREW_STRING)"
      location: "JS frame :: http://maps.google.com/mapfiles/maps.8.js :: anonym
ous :: line 490"
      data: no]
    Source File: http://maps.google.com/mapfiles/maps.8.js
    Line: 490

This error comes up as soon as I click in the map window, regardless of whether 
I have entered a search query.  It doesn't give any more errors when I click on 
"Satellite", i.e. when the browser is hung.

Again, this is all using Firefox 1.0 on Solaris 10/x86, using a fresh profile an
d no proxies (that I know of).

If there is additional data I could gather to help solve the problem(s), I will
gladly do so.

Reproducible: Always

Steps to Reproduce:
See Details.

Actual Results:  
See Details.

Expected Results:  
See Details.

As described in Details, more information is available at

    http://foo.net/~blakej/tmp/bugs/ff-1/1.jpg
    http://foo.net/~blakej/tmp/bugs/ff-1/2.jpg
    http://foo.net/~blakej/tmp/bugs/ff-1/3.jpg
    http://foo.net/~blakej/tmp/bugs/ff-1/stacks.txt
    http://foo.net/~blakej/tmp/bugs/ff-1/filtered.txt

Comment 1

[Kevin Brosnan]

Firefox 1.0 is nearly a year old, it would be helpful if you checked if it
worked with the deer park alpha 1 availble at
http://ftp.mozilla.org/pub/mozilla.org/firefox/releases/deerpark/alpha1/contrib/

Comment 2

[Blake Jones]

Deerpark Alpha 1 shows almost identical symptoms:

- searching for [san francisco, ca] gives a multiply-tiled world map
- can't pan north/south on the world map
- clicking "satellite" hangs the browser for a few seconds

The only difference is that the "satellite" link functions a tiny bit - it
doesn't always cause the browser to hang completely.  Sometimes the browser will
become responsive in as few as three seconds, in contrast to the unending hang
that Firefox 1.0 showed.  It didn't successfully display any actual maps,
though, and after a couple zooms in and out it did in fact go into "this script
is wedging your browser, should I kill it?" mode.

Comment 3

[Blake Jones]

Deerpark Alpha 1 showed different complaints in the Javascript Console window. 
This one showed up whenever I clicked or panned on Google Maps:

    Error: Error in parsing value for property 'top'.  Declaration dropped.
    Source File: http://maps.google.com/
    Line: 0

Clicking on the "Satellite" link gave a flood of these errors, along with a
bunch that were the same except were about "property 'left'".

Comment 4

[John Weeks]

When compiling on Solaris x86, __LITTLE_ENDIAN is not defined:

 fdlibm.h

 ...
 #if defined(linux) && (defined(__i386__) ||
 defined(__x86_64__))
 #define __LITTLE_ENDIAN
 #endif

 /* End here. The rest is the standard file. */

 #ifdef __NEWVALID       /* special setup for Sun test regime */
 #if defined(i386) || defined(i486) || \
        defined(intel) || defined(x86) ||
        defined(i86pc)
 #define __LITTLE_ENDIAN 
 ...

I don't know what the purpose of __NEWVALID is, but it's not defined either.
Thus, we are building with the wrong byte swapping.

Comment 5

[Leon Sha]

Attachment: Patch

Comment 6

[Brendan Eich [:brendan]]

This is a dup of bug 297750, which has a smaller spot-fix patch that preserves
the mess that is fdlibm's preprocessor portability -- why not help get that
patch in?

/be

*** This bug has been marked as a duplicate of 297750 ***