101896 - Unresolved symbol in libgkcontent.so when starting mail/news

Status: RESOLVED INCOMPLETE

(SeaMonkey :: Build Config, defect)

Description

[Greg Onufer]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; SunOS sun4u; en-US; rv:0.9.4+) Gecko/20010924
BuildID:    2001092618

Starting Tuesday afternoon every build I've made (incremental,
clobber, from completely new source tree) has resulted in
the browser crashing when I start mail/news after starting
the browser (ie, view mail/news icon on component bar):

ld.so.1: ./mozilla-bin: fatal: relocation error: file 
/home/exodus/tmp/mozilla.bad/components/libgkcontent.so: symbol 
__1cIDistance6FrknRnsReadingIterator4CH__2_I_: referenced symbol not found
Killed

As far as I can tell, that's from this code (from inspection
of CC -S assembler output):

 nsCOMPtr<nsIAtom> atom = 
dont_AddRef(NS_NewAtom(Substring(first, iter)));

in

nsresult
nsXULOutlinerBuilder::TokenizeProperties(const nsAReadableString& aString,
                                          nsISupportsArray* aProperties)


(around line 1400 of nsXULOutlinerBuilder.cpp).

That particular file (nsXULOutlinerBuilder.cpp) hasn't changed
recently.

I'm using the latest Forte Compilers from Sun.  I've been using
those compiler for several weeks now with many successful mozilla
builds, this is new since Tuesday afternoon.  It may very well
be a compiler bug.  I pulled the nightly build from mozilla.org
earlier today and it did not have this problem but was probably
built with gcc.


Reproducible: Always
Steps to Reproduce:
1. Start mozilla normally
2. click on the mail icon in the component bar
3.

Actual Results:  Application died due to unexpected unresolved symbol.


Expected Results:  Mail/News window opening.

Comment 1

[Asa Dotzler [:asa]]

->outliner

Comment 2

[John Morrison]

do you have a stack trace for this?

Also, are you sure that (somehow) you haven't wound up with some cruft in 
your components directory? What happens if you delete 'component.reg' and
then restart?

at any rate, sounds more like platform/build system issues than xptoolkit code.

Comment 3

[Greg Onufer]

If I back this out then it looks like everything starts working
again (the incremental build isn't finished but what's in the dist
directory so far now works):

http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=SeaMonkeyAll&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Change+Size&hours=2&date=explicit&mindate=09%2F25%2F2001+02%3A34%3A00&maxdate=09%2F25%2F2001+02%3A35%3A59&cvsroot=%2Fcvsroot

It restores the missing symbol to libstring_s.a and libxpcom.so:

./lib/libstring_s.a
[52]    |       200|      76|FUNC |GLOB |0    |2     
|__1cIDistance6FrknRnsReadingIterator4CH__2_I_

./lib/libxpcom.so
[4357]  |    877932|      76|FUNC |GLOB |0    |10    
|__1cIDistance6FrknRnsReadingIterator4CH__2_I_

Comment 4

[jag (Peter Annema)]

Have you tried clobbering your build?

Comment 5

[John Morrison]

Noted above: 
> I'm using the latest Forte Compilers from Sun.

Possibly they don't like some of that template-fu. Pav, you've the Forte 
stuff?

Comment 6

[Greg Onufer]

As mentioned in the very first comment I found this on an incremental build,
then tried a clobber build, then pulled a completely new source tree.  This
is 100% reproducible no matter how I build.  As far as I can tell exactly
one reference to that function is being generated, no other object files
in the tree refer to it.  Until I back out that one fix and then references
to that symbol show up everywhere.

Comment 7

[jag (Peter Annema)]

Sorry, I missed your note that you tried a clobber build.

/me investigates

Comment 8

[Greg Onufer]

I pulled a completely new tree and built with the Sun Workshop 6 Update 1
compilers and reproduced this problem.  For the past few weeks I've been
(successfully until Tuesday) using a snapshot of the next upcoming Forte
release so I wanted to verify that older released compilers encountered
the same problem.  They do.

Comment 9

[hacker formerly known as [e-mail]]

Try upgrading to WS6U2. We're running it on nebiros for weeks and it doesn't
show this problem.  

Reassigning to jag since it was his checkin that triggered this.

Comment 10

[Greg Onufer]

The compiler I've been using for weeks is a current stable snapshot of
WS7 (Forte 7).  I only tried WS6U1 as a second opinion to see if the results
were identical (and they were).  I'll see if I can try WS6U2, I'm downloading
it now.

Can someone check the object files on your build machine for the
relevant symbols?  mozilla does start fine and may even pass generic
browser tests, it just dies when you try mailnews.

Comment 11

[Greg Onufer]

The build with WS6U2 is finished, it has the same exact
problem:

./lib/libgkconxultmpl_s.a
[344]   |         0|       0|FUNC |GLOB |0    |UNDEF 
|__1cIDistance6FrknRnsReadingIterator4CH__2_I_

./lib/components/libgkcontent.so
[25180] |         0|       0|FUNC |GLOB |0    |UNDEF 
|__1cIDistance6FrknRnsReadingIterator4CH__2_I_

And that symbol isn't defined in any other object.

% CC -V
CC: Sun WorkShop 6 update 2 C++ 5.3 2001/05/15
%

Here is my .mozconfig file:

# sh
# Build configuration script
#
# See http://www.mozilla.org/build/unix.html for build instructions.
#

# Options for client.mk.
mk_add_options MOZ_MAKE_FLAGS=-j3

# Options for 'configure' (same as command-line options).

ac_add_options --enable-optimize="-O -xO4 -xtarget=ultra3 -xspace -xildoff"
ac_add_options --enable-ultrasparc
ac_add_options --enable-crypto
ac_add_options --disable-tests
ac_add_options --disable-debug
ac_add_options --without-jpeg
ac_add_options --without-zlib
ac_add_options --without-png

Comment 12

[jag (Peter Annema)]

So when I build on linux using gcc (egcs-2.91.66), without -O, I get
Substring__FRCt17nsReadingIterator1ZUsT0, and no Distance symbol in
libgkconxultmpl_s.a.

When I build with -O or -O3, I get Distance__FRCt17nsReadingIterator1ZUsT0, and
no Substring symbol.

My libstring_s.a contains Distance__FRCt17nsReadingIterator1ZUsT0 for both -O3
and no -O, and doesn't contain Substring__FRCt17nsReadingIterator1ZUsT0 in
either case.

Comment 13

[hacker formerly known as [e-mail]]

*** Bug 103028 has been marked as a duplicate of this bug. ***

Comment 14

[hacker formerly known as [e-mail]]

Btw, I did verify the crash using Greg's config & WS6U2 (both with & without the
crazy optimization options).

Comment 15

[jag (Peter Annema)]

Does someone have the cycles to investigate what's going on here? Again, it
looks like it's a problem with the compiler.

Comment 16

[Roland Mainz]

An attempt to build a StaticBuild Mozilla using the Workshop compiler fails with
the following error... I assume it's the same problem as this one:
-- snip --
/opt/SUNWspro/bin/CC -o mozilla-bin -I/usr/openwin/include -mt -O  -DDEBUG
-DDEBUG_mozilla -DTRACING -g -DWIDGET_DLL=\"libwidget_gtk.so\"
-DGFXWIN_DLL=\"libgfx_gtk.so\" -I/usr/local/lib/glib/include
-I/usr/local/include -I/usr/openwin/include  nsAppRunner.o nsSetupRegistry.o
nsWindowCreator.o nsSigHandlers.o showOSAlert.o nsStaticComponents.o -xildoff   
-L../../dist/bin -L../../dist/lib -L../../dist/lib/components -lxpconnect
-ljsloader -lxpctest -ltimer_gtk -lnkcache -lnecko -lnecko2 -luriloader -luconv
-lucvja -lucvcn -lucvlatin -lucvtw -lucvtw2 -lucvko -lucvibm -lucvmath
-lunicharutil -lnslocale -lstrres -llwbrk -lchardet -lpref -loji -lcaps -lchrome
-lrdf -lhtmlpars -lgfxps -lgfx_gtk -lgfxxprint -lgfx2 -limglib2 -limgppm
-limgpng -limggif -limgjpeg -lgkplugin -lnpsimple -ljsurl -ljsdom -lgkview
-lwidget_gtk -lxremote_client -lgkcontent -lgklayout -lmork -ldocshell
-lembedcomponents -lwebbrwsr -lcomposer -ltxmgr -ltxtsvc -lprofile -lnsprefm
-laccessibility -lnsappshell -lmozfind -lregviewer -lmozxfer -lshistory
-lappcomps -lxremoteservice -lmozbrwsr -lpipnss -lpippki -lmozldap -lcookie
-lwallet -lwalletviewers -lxmlextras -ltransformiix -lmailnews -lmsgdb -lmsgnews
-llocalmail -lmimeemitter -lvcard -lsmimestb -lmime -lmsgcompose -lmsgimap
-laddrbook -limport -limpText -labsyncsvc -lmozjs -lxptinfo -lxpcom -ljsj -lmozz
-lmozjpeg -lgtksuperwin -lgtkxtbin -lmozpng -lmozmng -lxlibrgb -lgkgfx
-lgtkembedmoz -lmsgbaseutil  -L../../dist/lib -lmozpng -L../../dist/lib -lmozmng
-L../../dist/lib -lmozjpeg -L../../dist/lib -lmozz  -lXIE -L/usr/openwin/lib
-R/usr/openwin/lib -lXp -lXext -lX11  ../../dist/lib/libcrmf.a
../../dist/lib/libssl.a ../../dist/lib/libnss.a ../../dist/lib/libssl.a
../../dist/lib/libpkcs12.a ../../dist/lib/libpkcs7.a ../../dist/lib/libcerthi.a
../../dist/lib/libpk11wrap.a ../../dist/lib/libcryptohi.a
../../dist/lib/libcerthi.a ../../dist/lib/libpk11wrap.a
../../dist/lib/libsoftoken.a ../../dist/lib/libcertdb.a
../../dist/lib/libfreebl.a ../../dist/lib/libsecutil.a ../../dist/lib/libdbm.a  
-lldap40 -llber40  -L/usr/openwin/lib -R/usr/openwin/lib -lXt -L../../dist/bin
-lxpcom -ljsj
-L/shared/bigtmp/mozilla/2001-10-06-08-trunk/objdir_ws6_gtk_static/dist/lib
-lplds4 -lplc4 -lnspr4 -lpthread -ldl   -L/usr/local/lib -L/usr/openwin/lib
-R/usr/openwin/lib -lgtk -lgdk -lgmodule -lglib -ldl -lXext -lX11 -lsocket -lnsl
-lm -lsocket -ldl -lm     
Undefined                       first referenced
 symbol                             in file
unsigned Distance(const nsReadingIterator<unsigned short>&,const
nsReadingIterator<unsigned short>&)
../../dist/lib/components/libgkcontent.a(nsXULOutlinerBuilder.o)
ld: fatal: Symbol referencing errors. No output written to mozilla-bin
gmake[1]: *** [mozilla-bin] Error 1
gmake[1]: Leaving directory
`/shared/bigtmp/mozilla/2001-10-06-08-trunk/objdir_ws6_gtk_static/xpfe/bootstrap'
gmake: *** [install] Error 2
-- snip --

Comment 17

[Roland Mainz]

Can anyone check whether this issue exists in the 0.9.5_branch, too ?
This looks like a milestone blocker for me...

Comment 18

[Roland Mainz]

Attachment: patch for 2001-10-06-08-trunk (more a workaround than a real fix)

Comment 19

[jag (Peter Annema)]

Comment on attachment 52530 [details] [diff] [review]
patch for 2001-10-06-08-trunk (more a workaround than a real fix)

r=jag

Comment 20

[Roland Mainz]

BTW: The real cause of this bug is more or less described in
http://lxr.mozilla.org/seamonkey/source/string/public/nsStringIteratorUtils.h#41
My patch is only a workaround - this issue can pop-up again anytime, anywhere
and is hard to track...
Can anyone please file a bug to fix that, please ?

Comment 21

[Roland Mainz]

StaticBuild compiles now without problems (but crashes at startup, I filed bug
103746 for that issue) ...

Comment 22

[jag (Peter Annema)]

Greg: could you see if this patch does the trick for you?

Comment 23

[Greg Onufer]

Yes, I un-backed-out the changes that introduced this and applied the
workaround.  All seems well, no trace of the unresolved symbol.  Thanks,
Roland.

Comment 24

[Roland Mainz]

*** Bug 103008 has been marked as a duplicate of this bug. ***

Comment 25

[jag (Peter Annema)]

Attachment: Wild guess

Comment 26

[jag (Peter Annema)]

Roland, this is just a wild guess. Could you see if this instead of the first
patch fixes your problem?

Comment 27

[Roland Mainz]

Nope, the new patch does not fix it... ;-(

It still goes to hell when I try to open the DOM inspector:
-- snip --
###!!! ASSERTION: Popup set is already defined! Only 1 allowed.:
'!mPopupSetFrame', file
../../../../../../../src/2001-10-15-08-trunk/mozilla/layout/xul/base/src/nsRootBoxFrame.cpp,
line 302
###!!! Break: at file
../../../../../../../src/2001-10-15-08-trunk/mozilla/layout/xul/base/src/nsRootBoxFrame.cpp,
line 302
ld.so.1: ./mozilla-bin: fatal: relocation error: file
/home/mozilla/builds/2001-10-15-08-trunk/objdir_ws6_xlib/dist/bin/components/libgkcontent.so:
symbol __1cIDistance6FrknRnsReadingIterator4CH__2_I_: referenced symbol not
found
Killed
-- snip --

Comment 28

[jag (Peter Annema)]

Roland: could you e-mail me the pre-processed output of
content/xul/templates/src/nsXULOutlinerBuilder.cpp without and with your patch?

Comment 29

[Roland Mainz]

Attachment: content/xul/templates/src/nsXULOutlinerBuilder.cpp without patch 52530

Comment 30

[Roland Mainz]

Attachment: content/xul/templates/src/nsXULOutlinerBuilder.cpp with patch 52530

Comment 31

[Roland Mainz]

Attachment: cat without_patch.txt | grep -v "^#" >without_patch1.txt; cat with_patch.txt | grep -v "^#" >without_patch1.txt; gdiff -u -w without_patch1.txt with_patch1.txt

Comment 32

[Roland Mainz]

It would be nice to check-in my workaround for now. 
I have reports that this issue always "happens" with the current recommended
patch cluster for Solaris 7 and the newest Sun Workshop 6U2 patch set (AFAIK
"nebiros" wasn't updated yet...).

Comment 33

[jag (Peter Annema)]

Work-around checked in.

-> future, helpwanted for figuring out what the right fix for this is.

Comment 34

[Roland Mainz]

We hit this issue another time, see bug 135731 ("opening address book crashes
mozilla") ...

Comment 35

[hacker formerly known as [e-mail]]

*** Bug 167722 has been marked as a duplicate of this bug. ***

Comment 36

[Donnie Cranford]

I also am seeing this except now its in libnecko 

I am using Forte 7 with the latest patches for C and C++ from sunsolve
also using Solaris 9 with what as of 9/18/02 is the latest patch cluster

Here is my environment

# env
CC=/opt/SUNWspro/bin/cc
CXX=/opt/SUNWspro/bin/CC
PATH=/usr/sbin:/usr/bin:/usr/dt/bin:/usr/openwin/bin:/bin:/usr/ccs/bin:/opt/SUNW
spro/bin:/usr/local/bin:/usr/ucb
LD_LIBRARY_PATH=/usr/lib:/usr/sfw/lib:/usr/local/lib

# cat .mozconfig

mk_add_options MOZ_MAKE_FLAGS=-j4
ac_add_options --disable-tests
ac_add_options --disable-debug
ac_add_options --enable-optimize=-xO2
ac_add_options --without-system-nspr
ac_add_options --without-system-zlib
ac_add_options --without-system-jpeg
ac_add_options --without-system-png
ac_add_options --without-system-mng
ac_add_options --enable-crypto

Comment 37

[Roland Mainz]

Donnie Cranford wrote:
> I also am seeing this except now its in libnecko 
>
> I am using Forte 7 with the latest patches for C and C++ from sunsolve
> also using Solaris 9 with what as of 9/18/02 is the latest patch cluster

I don't see this in Solaris 2.7/SPARC with Sun Workshop 7+latest WS patches...
what's your buildid ?

Comment 38

[Donnie Cranford]

Roland,

I have given more info on this in bug 167722

Comment 39

[Robert Kaiser]

This bug report is registered in the SeaMonkey product, but has been without a comment since the inception of the SeaMonkey project. This means that it was logged against the old Mozilla suite and we cannot determine that it's still valid for the current SeaMonkey suite. Because of this, we are setting it to an UNCONFIRMED state.

If you can confirm that this report still applies to current SeaMonkey 2.x nightly builds, please set it back to the NEW state along with a comment on how you reproduced it on what Build ID, or if it's an enhancement request, why it's still worth implementing and in what way.
If you can confirm that the report doesn't apply to current SeaMonkey 2.x nightly builds, please set it to the appropriate RESOLVED state (WORKSFORME, INVALID, WONTFIX, or similar).
If no action happens within the next few months, we move this bug report to an EXPIRED state.

Query tag for this change: mass-UNCONFIRM-20090614

Comment 40

[Robert Kaiser]

Our code and compilers have evolved, I don't think there's any valid information for current issues left here.