[Scilab-Dev] scilab-5.0.3 segfault on solaris: strlen()

[Gerard Henry]

i tried, by adding:
--enable-build-localization

and now, the make hangs here:
./bin/scilab -ns -nwni -f 
modules/functions/scripts/buildmacros/buildmacros.sce
Warning: Localization issue: Error while binding the domain from 
/local/apps/src/scilab-5.0.3/../locale/ or 
/local/apps/src/scilab-5.0.3/locale/: Switch to the default language 
(English).


the debugger sais that there is a thread waiting:
Attached to process 12493 with 3 LWPs
t at 1 (l at 1) stopped in __lwp_park at 0xfa9a47fb
0xfa9a47fb: __lwp_park+0x000b:  jae      __lwp_park+0x19        [ 
0xfa9a4809, .+0xe ]
Current function is basin
    34                  call 
zzledt(string,len(string),lline,status,menusflag,0)

(dbx) lwp l at 2
t at 2 (l at 2) stopped in _read at 0xfa9a52a7
0xfa9a52a7: _read+0x0007:       jae      _read+0x13     [ 0xfa9a52b3, 
.+0xc ]
(dbx) list
dbx: warning: No source file
(dbx) where
=>[1] _read(0x0, 0xfa9d680c, 0x1), at 0xfa9a52a7
   [2] __filbuf(0xfa9d1c10, 0xfd287c44, 0xfa57efb8, 0xfd273ed8), at 
0xfa9881a4
   [3] gchar_no_echo(0x0, 0x0, 0x0, 0x0, 0x0, 0x0), at 0xfd273f00
   [4] TermReadAndProcess(0xfa470200, 0xfd287c44, 0xfa57efd8, 
0xfd27247a, 0xfa470200, 0xfa9ce000), at 0xfd272c6c
   [5] getCommandLine(0xfa470200, 0xfa9ce000, 0xfa57efec), at 0xfd272398
   [6] watchGetCommandLine(0x0), at 0xfd27247a
   [7] _thr_setup(0xfa470200), at 0xfa9a44b7
   [8] _lwp_start(0x0, 0xfa9d680c, 0x1, 0xfd288174, 0x0, 0xfd287c44), at 
0xfa9a47a0
(dbx) kill

Why does it hang on a read? what's the meaning of the read in this context?
When i kill the lwp 2, the make finishes, and i can launch scilab, but 
it is unusable:
 > ./bin/scilab -nogui
                                 !--error 42
Incompatible input argument.



Jonathan Blanchard wrote:
> Ah the joys of Scilab on Solaris.
> 
> I remember having a similar issue quite a while ago. If I remember
> correctly it was something about localization. Either enabling or
> disabling it kind of fixed the problem.
> 
> Jonathan Blanchard
> 
> 
> 
> On Sat, Jan 10, 2009 at 4:12 PM, Gerard Henry
> <Gerard.Henry at cmi.univ-mrs.fr> wrote:
>> Gerard Henry wrote:
>>> hello all,
>>> after successfully compiling scilab on opensolaris, i did the same job
>>> on solaris 10, make went fine, but when i execute, it segfaults. Looking
>>> with the debugger, i got:
>>> Program received signal SIGSEGV, Segmentation fault.
>>> 0xfa855891 in strlen () from /usr/lib/libc.so.1
>>> (gdb) where
>>> #0  0xfa855891 in strlen () from /usr/lib/libc.so.1
>>> #1  0xfac62160 in callFunctionFromGateway ()
>>>   from /local/apps/src/scilab-5.0.3/modules/core/.libs/libscicore.so.5
>>> #2  0xfd297571 in gw_io ()
>>>   from /local/apps/src/scilab-5.0.3/modules/io/.libs/libsciio.so.5
>>> #3  0xfac51df3 in callinterf_ ()
>>>   from /local/apps/src/scilab-5.0.3/modules/core/.libs/libscicore.so.5
>>> #4  0xfac5ef42 in scirun_ ()
>>>   from /local/apps/src/scilab-5.0.3/modules/core/.libs/libscicore.so.5
>>> #5  0xfac4d54b in realmain ()
>>>   from /local/apps/src/scilab-5.0.3/modules/core/.libs/libscicore.so.5
>>> #6  0xfd242ad0 in mainscic ()
>>>   from /local/apps/src/scilab-5.0.3/modules/shell/.libs/libscishell.so.5
>>> #7  0x08050d7d in main ()
>>>
>>> even a "make clean" becomes impossible, because:
>>> devel1 at nemo:~/src/scilab-5.0.3$ make clean
>>> ./bin/scilab: line 453: 13413 Segmentation Fault      (core dumped)
>>> "$SCILABBIN" "$@"
>>> ./bin/scilab: line 453: 13439 Segmentation Fault      (core dumped)
>>> "$SCILABBIN" "$@"
>>> make: *** [clean-doc] Error 139
>>>
>>> Any idea?
>>>
>>> thanks in advance for help,
>>>
>>> gerard
>>>
>>> PS: some notes in french:
>>> http://www.latp.univ-mrs.fr/support/doku.php?id=scilab
>>>
>> after recompiling with -g, i have more information:
>> t at 1 (l at 1) program terminated by signal SEGV (no mapping at the fault
>> address)
>> 0xfa705891: strlen+0x0031:      cmpb     $0x00000000,(%eax)
>> Current function is callFunctionFromGateway
>>   36           if (*(Tab[Fin-1].f) != NULL) (*(Tab[Fin-1].f))
>> (Tab[Fin-1].name,(unsigned long)strlen(Tab[Fin-1].name));
>>
>> (dbx) where
>> current thread: t at 1
>>  [1] strlen(0xe, 0x0), at 0xfa705891
>> =>[2] callFunctionFromGateway(Tab = 0xfd2be908), line 36 in
>> "callFunctionFromGateway.c"
>>  [3] gw_io(), line 102 in "gw_io.c"
>>  [4] callinterf_(k = 0xfab07b04), line 155 in "callinterf.c"
>>  [5] scirun_(startupCode = 0x8081358
>> "exec('modules/functions/scripts/buildmacros/buildmacros.sce',-1)",
>> startupCode_len = 64), line 147 in "scirun.c"
>>  [6] realmain(no_startup_flag_l = 1, initial_script = 0x8046f3f
>> "modules/functions/scripts/buildmacros/buildmacros.sce", initial_script_type
>> = SCILAB_SCRIPT, memory = 180000), line 145 in "realmain.c"
>>  [7] mainscic(argc = 5, argv = 0x8046e00), line 132 in "mainscic.c"
>>  [8] main(argc = 5, argv = 0x8046e00), line 26 in "main_linux.c"
>>
>>
>> is it possible that the value of Fin is out of range? Who set the value of
>> Fin?
>>
>> thanks for help,
>>
>> gerard
>>
>> --
>> Gerard HENRY
>> LATP UMR6632
>> 39, rue F. Joliot Curie
>> 13453 MARSEILLE Cedex 13
>> Tel.: 04 91 11 35 53 Fax.: 04 91 11 35 52
>>


-- 
Gerard HENRY
LATP UMR6632
39, rue F. Joliot Curie
13453 MARSEILLE Cedex 13
Tel.: 04 91 11 35 53 Fax.: 04 91 11 35 52