140323 - XML page crashes prior to any display: Solaris

Status: VERIFIED FIXED

(Core :: XSLT, defect)

Description

[Jerry]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.0rc1)
Gecko/20020417
BuildID:    Mozilla 1.0 RC1 (2002041711)

This same code works when accessed via IE 6.0.0.2600 on XP and works with the
same Mozilla RC1 build when run locally on Windows 2000 box.
Talkbacks were sent..
Running on iPlanet Web Server 4.1 SP9 and also on 6.0 SP2 (special MIME Types
not set on either Web Server) fails. Runs ok locally.
 

Reproducible: Always
Steps to Reproduce:
1. Connect to URL: http://www.uc.usbr.gov:2525/xml/pollutionPages.xml crashes
Mozilla 1.0 RC1
2. There is no second step...
3.

Actual Results:  Mozilla crashes...

Expected Results:  display of xml document as per IE 6.0.0.2600

Running locally on the Windows box I had to set change
'href="pollutionPages.xslt"' to 'href="pollutionPages.xsl"' within
pollutionPages.xslt before Mozilla would display page correctly (at all). of
course, i had to change file extension from xslt to xsl also. Left "text/xsl" as
is...

<?xml-stylesheet type="text/xsl" href="pollutionPages.xsl" ?>

Comment 1

[jlarsen]

WFM in windows  (Just saying this so others won't waist there time testing it in
windows too.)
Any Solaris users out there who can test?

Comment 2

[Rakesh Mishra]

Changing Q/A contact

Comment 3

[Kai Lahmann (is there, where MNG is)]

there was a XSLT-crash in 1.0RC1, which is fixed in the latest nightlies as far
as I know...

Comment 4

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

Probably XSLT bug, sending there for first look.

Comment 5

[Axel Hecht]

wrong mime type for the stylesheet. it's text/plain, it should be text/xml or 
application/xml. Probably the same thing as bug 139385, if only someone
could reproduce it.

Comment 6

[Shanmugavelu Shanmuganathan (gone)]

I am seeing this on  a Compaq's Tru64 UNIX too.
Looks like its a 64 bit problem. 
The data is not initialized (or wiped out).

ladebug 
>0  0x3ffbff375ac in StrAppend__12nsStrPrivateXR5nsStrRC5nsStrUii(aDest=& struct
nsStr { ... }, aSource=& struct nsStr { ... }, anOffset=0, aCount=3)
"nsStr.cpp":193
#1  0x3ffbff40a68 in
((nsString*)0x0)->nsString::AppendWithConversion(aCString=0x3ffbd6c8f48="1.0",
aCount=3) "nsString2.cpp":725

Comment 7

[Axel Hecht]

shanmu, could you attach a stack trace?

Comment 8

[Jim Crumley]

Works for me on Sparc Solaris with a local version of RC1, a local build from
20020430, and nightly build 2002042822. 

Comment 9

[Donnie Cranford]

This bug WORKSFORME on Solaris distribution of RC1 and also latest 1.0 branch
build from 04/30/2002

Comment 10

[Axel Hecht]

changing the URL. the original one is fixed on the server side.
Please give this a bit to stage.

Comment 11

[Shanmugavelu Shanmuganathan (gone)]

Attachment: Stack trace from a Tru64 UNIX machine.

I am still seeing this crash on a Tru64 UNIX.
I have attached the stack trace here.

Comment 12

[Shanmugavelu Shanmuganathan (gone)]

This bug looks like a machine dependant.
Mozilla is confused about the constructor for "String" class.
On a Tru64 UNIX mozilla was trying to use the constructor from
the libcxx.so. Instaed of looking at the local symbols first it was using
the one from libcxx.so and caused grief. I made it look at the libtransformix.so
first by  using the loader option "-B symbolic". It seems to work fine
when I make this one change in the Makefile.  
	Any idea which onfiguration file I should put this option to make 
it reflect in the /extensions/transformiix/build/makefile?

Comment 13

[Shanmugavelu Shanmuganathan (gone)]

Attachment: This fix adds "-Bsymbolic" to the loader option.

This is a tru64 UNIX only fix.

Comment 14

[Shanmugavelu Shanmuganathan (gone)]

Huh. I just realized that I have to file a bug against build config. 
So never mind this fix. I will move this to another bug.

Comment 15

[Chul Su Park]

Shanmu, your fix for rules.mk didn't work, it may need to be

< EXTRA_DSO_LDOPTS += -Wl,-Bsymbolic
> EXTRA_DSO_LDOPTS += -Wl,-B,symbolic

at least for dec osf 4.0f cxx/ld.

Comment 16

[Axel Hecht]

marking fixed, looks like rules.mk knows how to handle this now

Comment 17

[Axel Hecht]

mass verifying