Re: RedHat / Fedora build of ncl

Hi Orion,
I agree that something has gone awry before you get into the _Nclstr_match_ic function. For what it's worth, I ran the same code in gdb here at NCAR as follows:

ncl 0> class_out = NhlGetClassResources("contourPlotClass", "")
ncl 1> print(dimsizes(class_out))
(0) 610
ncl 2> cnConst = str_match_ic(class_out, "cnConst")

Breakpoint 1, _Nclstr_match_ic () at userAddFuncs.c:3760
3760 int has_missing = 0;
(gdb) n
3771 ng_size_t output_str_size = 0;
(gdb)
3780 input_strs = (string *) NclGetArgValue(
(gdb)
3790 if (input_strs == NULL)
(gdb)
3796 if(type != NCL_string)
(gdb)
3802 if(has_missing_input_strs)
(gdb)
3804 ret_missing.stringval = missing_input_strs.stringval;
(gdb)
3805 has_missing = 1;
(gdb)
3812 input_expr = (string *) NclGetArgValue(
(gdb)
3822 if (input_expr == NULL)
(gdb)
3828 if(type == NCL_string)
(gdb)
3830 char *reg_exp = NrmQuarkToString(input_expr[0]);
(gdb)
3835 if(strlen(reg_exp) > 0)
(gdb)
3837 if(regcomp(&expr,reg_exp,REG_ICASE|REG_EXTENDED) != 0)
(gdb)
3850 str_size = 1;
(gdb)
3851 for(i=0; i<ndim_input_strs; i++)
(gdb)
3852 str_size *= dimsz_input_strs[i];
(gdb) p dimsz_input_strs
$1 = {610, 8589934592, 4294967298, 4354594080, 0, 0, 4294971392, 4354594080, 0, 4353687552, 4353032194, 4353069440, 140734799799664, 2048, 140734799799920, 4296697968, 4294967298, 4354594080, 0, 0, 17592186048512, 4354158336, 261993005056, 0, 140734799799776, 4353687552, 0, 4310887584, 0, 0, 0, 0}
(gdb) n
3851 for(i=0; i<ndim_input_strs; i++)
(gdb) n
3854 output_strs = (string *) NclMalloc(str_size*sizeof(string));
(gdb) n
3855 if (! output_strs)
(gdb) p output_strs
$2 = (string *) 0x104106800
(gdb) p str_size*sizeof(string)
$3 = 4880

It would be hard to diagnose this without running the version that is crashing. Now I see that Mary has responded to you as well, but maybe this at least shows you what it should look like for this code fragment.
 -dave

On Jul 13, 2012, at 4:37 PM, Orion Poplawski wrote:

> Sending this to the ncl-talk list to see if we can get any more insight.
>
> Summary - seeing a segfault in Fedora/EPEL RPMs of ncl.
>
> This is where I've tracked it to now.
>
> In userAddFuncs.c:_Nclstr_match_ic()
>
> str_size = 1;
> for(i=0; i<ndim_input_strs; i++)
> str_size *= dimsz_input_strs[i];
>
> output_strs = (string *) NclMalloc(str_size*sizeof(string));
> if (! output_strs)
> {
> NHLPERROR((NhlFATAL,ENOMEM,NULL));
> return NhlFATAL;
> }
>
> It appears that ndim_input_strs is 1, and
>
> (gdb) print dimsz_input_strs
> $17 = {0 <repeats 13 times>, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,
> 0, 0, 0}
>
> so str_size=0, NclMalloc(0) returns null and we return NhlFATAL which somehow
> ends up segfaulting.
>
> I can't grok how ncl handles arguments, seems very complicated.
>
> I believe things have gone awry before this point, but that is the symptom I see.
>
> continuing:
>
> 3485 NHLPERROR((NhlFATAL,ENOMEM,NULL));
> (gdb)
> fatal:["userAddFuncs.c":3485]:[errno=12]
> 3486 return NhlFATAL;
> (gdb)
> 3532 }
> (gdb)
> 0x00000000 in ?? ()
>
> so perhaps the stack is already smashed.
>
> FWIW - get the same with 6.1.0-beta.
>
> Any more suggestions for debugging this?
>
> On 06/22/2012 01:57 AM, Paul Scorer wrote:
>> Greetings,
>>
>> I write as I have stamped on a bug in both the i686 and x86_64 versions
>> of ncl as distributed with F15 and F16. I have not tested other
>> versions.
>>
>> The specific packages tested were
>>
>> ncl-6.0.0-2.fc16.i686
>> and
>> ncl-6.0.0-2.fc15.x86_64
>>
>> A simple test is to start ncl and enter the script as below. This is an
>> example script from
>> http://www.ncl.ucar.edu/Document/Functions/Built-in/str_match.shtml
>>
>> $ ncl
>> Copyright (C) 1995-2011 - All Rights Reserved
>> University Corporation for Atmospheric Research
>> NCAR Command Language Version 6.0.0
>> The use of this software is governed by a License Agreement.
>> See http://www.ncl.ucar.edu/ for more details.
>> ncl 0> class_out = NhlGetClassResources("contourPlotClass", "")
>> ncl 1> cnConst = str_match_ic(class_out, "cnConst")
>> Segmentation fault
>> $
>>
>> It the same for both str_match() and str_match_ic()
>>
>> I have taken this up with Mary Haley <haley@ucar.edu> of
>> ncl-talk@ucar.edu
>>
>> Clearly, they cannot support a build done by others. However, she has
>> noted that there have been a number of other issues with the RedHat
>> build of ncl, and has suggested that this may relate to the optimiser
>> options used.
>>
>>> From my point of view the matter has been resolved by downloading a
>> version of the ncl binary from Earth System Grid, but I am sure you
>> would agree that this is something that should get fixed.
>>
>> HTH
>>
>> PaulS
>>
>
>
> --
> Orion Poplawski
> Technical Manager 303-415-9701 x222
> NWRA/CoRA Division FAX: 303-415-9702
> 3380 Mitchell Lane orion@cora.nwra.com
> Boulder, CO 80301 http://www.cora.nwra.com
>
>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Fri Jul 13 17:59:05 2012