[GRASS5] Re: [bug #2767] (grass) r.stats bug (due to recent

[tlaronde at polynum.com]

On Thu, Dec 09, 2004 at 10:35:16AM +0100, Markus Neteler wrote:
> 
> ... also crashing:
> 
> (gdb) r -anC fields,elevation.10m
> Starting program: /hardmnt/thuille0/ssi/software/cvsgrass57/dist.i686-pc-linux-gnu/bin/r.stats -anC fields,elevation.10m
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x00306657 in G_set_raster_cats_title (title=0xbfffb3c0 "", pcats=0x7a2d4fd0) at cats.c:1536
> 1536        pcats->title = G_store (title);

This is the call to G_store (see the previous answer to Paul) since with
the code:

char *buf;

if (s == NULL)
	*buf = '\0';

one is still writing the char '\0' to an unknown address since buf has
a random value (probably 0 in this case, leading to the SEGFAULT; and
no SEGFAULT on a system where buf, not initialized, happens to point
to some valid address---thus trashing some place it doesn't own).

So correct sequence here too: initialize (reserve address) before
writing to it:

if (s == NULL)
	buf = (char *) calloc(1,1); /* 1 byte allocated and destination zeroed */

( same as doing:
buf = (char *) malloc(1);
*buf = '\0';
)

But I still have the intuition that you should test in r.stats, and not
change G_store().

If the program is still segfaulting with such a change, there is another
bug somewhere else...

Cheers,
-- 
Thierry Laronde (Alceste) <tlaronde +AT+ polynum +dot+ com>
http://www.kergis.org/  |  http://www.kergis.com/
Key fingerprint = 0FF7 E906 FBAF FE95 FD89  250D 52B1 AE95 6006 F40C