On 1/10/07, Ian Campbell <Ian.Campbell@xxxxxxxxxxxxx> wrote:

On Wed, 2007年01月10日 at 17:42 +0900, Magnus Damm wrote:
> Fix parse_size_and_unit() breakage in xen-3.0.4-testing (kdump broken)
>
> Kdump has been broken since xen-3.0.4-testing.hg-13109.
> Many thanks to Hiromichi Itou for tracking this down.
>
> Signed-Off-By: Magnus Damm <magnus@xxxxxxxxxxxxx>
I applied a similar fix to xen-unstable.hg in 13282:9865145e53eb. I'll
get it pulled into xen-3.0.4-testing.hg.

Great, thanks.

> --- 0001/xen/common/lib.c
> +++ work/xen/common/lib.c 2007年01月10日 11:18:38.000000000 +0900
> @@ -442,7 +442,7 @@ s64 __moddi3(s64 a, s64 b)
> unsigned long long parse_size_and_unit(const char *s, const char **ps)
> {
> unsigned long long ret;
> - const char *s1;
> + const char *s1 = NULL;
>
> ret = simple_strtoull(s, &s1, 0);
Are you sure this bit is necessary? simple_strtoull always initialises
&s1.

Probably no, but the reason why I put it there is that we would be
sure to crash if someone happened to change simple_strtoull(). If
simple_strtoull() would forget to write s1 then today we may get
silent errors where switch (*s1) is used on an uninitialized variable.
/ magnus
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel