Re: csh on arm failure?

On 5/15/07 6:56 AM, Ricardo Nabinger Sanchez wrote:
> On Tue, 15 May 2007 16:30:35 +0300
> Krassimir Slavchev <krassi_at_bulinfo.net> wrote:
> 
>> (gdb) bt
>> #0  0x20204bb0 in kill () from /lib/libc.so.7
>> #1  0x20200370 in abort () from /lib/libc.so.7
>> #2  0x00045fe4 in malloc (nbytes=16)
>>     at /work/arm/src/bin/csh/../../contrib/tcsh/tc.alloc.c:213
>> #3  0x00045fe4 in malloc (nbytes=16)
>>     at /work/arm/src/bin/csh/../../contrib/tcsh/tc.alloc.c:213
>> (gdb)
> 
> Any chances that this relates to Steven's PR 112408?  (Not the gdb part, just
> the tcsh regression.)

No, this is different. The tcsh malloc is showing it is out of memory and 
aborting. The thing I don't understand is the malloc is coming from tcsh 
instead of using the system malloc. Looking at this file:

src/bin/csh/config_p.h:

...
#if defined(__FreeBSD__)
#define NLS_BUGS
#define BSD_STYLE_COLORLS
/* we want to use the system malloc when we install as /bin/csh */
#define SYSMALLOC
/* Use LC_MESSAGES locale category to open the message catalog */
#define MCLoadBySet NL_CAT_LOCALE
#define BUFSIZE 8192
#endif

The SYSMALLOC define should cause it to use the system malloc and not the tcsh 
malloc. Could __FreeBSD__ not be defined?

Also, it is odd that it is out of memory in the first place. Has it worked 
before? Is there something that changed in the arm memory allocation schemes?

Mark