77300 - ASSERTION: unable to use nl_langinfo(CODESET): '0'

Status: VERIFIED FIXED

(Core :: Internationalization, defect)

Description

[timeless]

SunOS 5.7 Generic_106541-15 sun4u sparc SUNW,Ultra-2
glib 1.2.9
libIDL 0.6.8
qt 2.2.0
mozilla CVS
~/.mozconfig
ac_add_options --without-gtk
ac_add_options --with-qt
ac_add_options --enable-toolkit=qt
ac_add_options --enable-ultrasparc
ac_add_options --disable-tests
ac_add_options --with-qtdir=/tmp/qt-2.2.0/
ac_add_options --with-libIDL-prefix=/tmp/idl/
mk_add_options MOZ_OBJDIR=@TOPSRCDIR@/../obj-qt-@CONFIG_GUESS@

./run-mozilla.sh ./mozilla-bin
MOZILLA_FIVE_HOME=.
  LD_LIBRARY_PATH=.:./plugins:/tmp/qt-2.2.0/lib/:/tmp/idl/lib:/tmp/glib/lib
     LIBRARY_PATH=.:./components
       SHLIB_PATH=.
          LIBPATH=.
       ADDON_PATH=.
      MOZ_PROGRAM=./mozilla-bin
      MOZ_TOOLKIT=
        moz_debug=0
     moz_debugger=
File descriptors set to 512
Type Manifest File: /tmp/obj-qt-sparc-sun-
solaris2.7/dist/bin/components/xpti.dat
nsNativeComponentLoader: autoregistering begins.
nsNativeComponentLoader: autoregistering succeeded
nNCL: registering deferred (0)
###!!! ASSERTION: failed to create encoder: '0', 
file /tmp/mozilla/intl/uconv/src/nsUNIXCharset.cpp,
 line 396
###!!! Break: at file /tmp/mozilla/intl/uconv/src/nsUNIXCharset.cpp, line 396
###!!! ASSERTION: unable to use nl_langinfo(CODESET): '0', 
file /tmp/mozilla/intl/uconv/src/nsUNIXCharset.cpp, line 320
###!!! Break: at file /tmp/mozilla/intl/uconv/src/nsUNIXCharset.cpp, line 320
Initialized app shell component {18c2f989-b09f-11d2-bcde-00805f0e1353}, 
rv=0x00000000
GFX: dpi=96 t2p=0.0666667 p2t=15 depth=32
WEBSHELL+ = 1
nsPluginHostImpl ctor
********** Got plugins path: /tmp/obj-qt-sparc-sun-solaris2.7/dist/bin/./plugins
WEBSHELL+ = 2
Disabling View Source StyleSheet
Enabling Quirk StyleSheet
Note: verifyreflow is disabled
Style Data Sharing is Enabled :)
Note: styleverifytree is disabled
Note: frameverifytree is disabled
Shut down app shell component {18c2f989-b09f-11d2-bcde-00805f0e1353}, 
rv=0x00000000
nsPluginHostImpl::Observe "xpcom-shutdown"
### beging nsCacheService::Shutdown()
###!!! ASSERTION: ### active entry = nsnull!: 'entry', 
file /tmp/mozilla/netwerk/cache/src/nsCacheService.cpp, line 1057
###!!! Break: at file /tmp/mozilla/netwerk/cache/src/nsCacheService.cpp, line 
1057
Segmentation Fault

Steps to reproduce: build moz Qt, run moz 
Expected Results: Mozilla Window.
Actual Results: No Window, +Seg Fault.

I'm filing two bugs, one for intl and one for cache

Comment 1

[kill this account]

timeless: 
1) can you run the 'locale' command so I can see what locale you are using?
2) can you find out what the charset name is when it fails?

thanks

Comment 2

[timeless]

y:~: locale
LANG=
LC_CTYPE="C"
LC_NUMERIC="C"
LC_TIME="C"
LC_COLLATE="C"
LC_MONETARY="C"
LC_MESSAGES="C"
LC_ALL=

The second request will take work, i can't run gdb because i've eaten my entire 
swap space by building.  I'll have the answer by friday.  However, if you 
provide a recommended debugging patch i can try to have it thursday noonish.

Comment 3

[kill this account]

Attachment: debug patch to printf the charset

Comment 4

[kill this account]

add Ulrich Drepper to cc:

Comment 5

[kill this account]

If I set LC_CTYPE to "C" I get the assert but the app runs.

timeless: do you find that the assert causes the app to stop?

Comment 6

[Ulrich Drepper]

Strange.  Not that I would ever have suggested to use nl_langinfo() on Solaris
but it seems to work fine for me.  Well, "fine" is the way that the return value
makes sense even if they are very non-standard.

drepper@blues drepper$ LC_ALL=en_US.UTF-8 ./u
codeset = UTF-8
drepper@blues drepper$ LC_ALL=en_US ./u
codeset = ISO8859-1
drepper@blues drepper$ LC_ALL=fr.ISO8859-15 ./u
codeset = ISO8859-15
drepper@blues drepper$ LC_ALL=C ./u
codeset = 646

This is a Solaris 7 machine.  On Solaris 8 the results are the same.  The only
bad things about this are the names of the charsets:

  646        should be  ISO_646.IRV:1991 or ISO646-US
  ISO8859-1  should be  ISO-8859-1

Instead of running locale, run this program with different locales as I did above:

#include <langinfo.h>
#include <locale.h>
int
main()
{
  setlocale (LC_ALL, "");
  printf ("codeset = %s\n", nl_langinfo (CODESET) ?: "**NULL**");
  return 0;
}


(OK, some gcc-ism in there but I'm lazy.)

Comment 7

[timeless]

You omited #include <stdio.h>

y:/tmp: g++ a.cpp
y:/tmp: ./a.out
codeset = 646
y:/tmp: setenv LC_ALL en_US.UTF-8
y:/tmp: ./a.out
codeset = UTF-8
y:/tmp: setenv LC_ALL en_US
y:/tmp: ./a.out
codeset = ISO8859-1
y:/tmp: setenv LC_ALL fr.ISO8859-15
y:/tmp: ./a.out
codeset = ISO8859-15
y:/tmp: setenv LC_ALL C
y:/tmp: ./a.out
codeset = 646

remember, this is solaris 7

I'm not sure if this assert is killing mozilla, actually i suspect it's the 
next assert that is, but I can't tell if this one is hurting, so I filed this 
bug [and its sibling for the other assert].

Can you fix it so we don't assert?

Comment 8

[kill this account]

I want to fix the assert.

Do we think this should be part of 0.9 or 0.9.1 ?

Comment 9

[kill this account]

  On Linux: set LC_CTYPE=C => nl_langinfo(CODESET) returns "ANSI_X3.4-1968"
On Solaris: set LC_CTYPE=C => nl_langinfo(CODESET) returns "646"

Frank: should we add mappings to charsetalias.properties:

  ansi_x3.4-1968=us-ascii
  646=us-ascii

Comment 10

[Erik van der Poel]

I believe Mozilla's aliases are for the Internet (e.g. HTTP, HTML, email).
nl_langinfo's name space is quite distinct from that. I believe the vendors
just came up with whatever names they felt like using. That doesn't mean, of
course, that those names are used on the Internet. Yeah, yeah, be liberal in
what you accept and all that, but being too liberal isn't good either. (Look at
the amount of "broken" HTML on the Web today.)

Comment 11

[Andreas Becker]

Changing QA contact to [e-mail]. Please re-assign further as 
appropriate.

Comment 12

[Frank Tang]

ANSI_X3.4-1968 is in http://www.iana.org/assignments/character-sets so we should 
add it to charsetalias.properties anyway

I think it is fine to also add 646 to charsetalias.properties

bstell- go ahead and make the patch. I will review it after you put into the 
attachment

Comment 13

[kill this account]

Attachment: patch defining the 2 charsets

Comment 14

[Frank Tang]

r=ftang

Comment 15

[Brendan Eich [:brendan]]

I hereby give rs= to bstell and others for all such property-file-only changes.

Any property-file changes that have code effects would have to come with code
changes, too, and all the changes would need full review.

/be

Comment 16

[kill this account]

Brendan emailed me a clarification:

> I hereby give rs= for all such changes -- so long as you're sure of the
> "only" part.  Any whiff of code changes, or effects on code that knows
> something about a property, or property naming conventions, etc., and we
> need to have full review.
> 
> /be

Comment 17

[kill this account]

fix checked in

Comment 18

[Masaki Katakai]

Verifiled on Solaris.