Re: [Valgrind-developers] e500 SPE instructions and ppc_85xx for MontaVista - patch 2

[John M. <me...@ge...>]

Hi Bart,

On Thu, Jul 29, 2010 at 11:37 PM, Bart Van Assche <bva...@ac...> wrote:
> On Fri, Jul 30, 2010 at 12:52 AM, John Mehaffey <me...@ge...> wrote:
>> jmehaffe-lnx-1:/ws/jmehaffe/main_cge5/cles_moved_aside/vg/rpm/SOURCES
>>> cat valgrind-3.5.0_cross_compile.patch
>> --- valgrind-3.5.0/configure.in.orig    Wed Aug 19 06:37:48 2009
>> +++ valgrind-3.5.0/configure.in Thu Jul  1 15:01:20 2010
>> @@ -1472,9 +1472,10 @@ fi
>>  #----------------------------------------------------------------------------
>>  # Check for /proc filesystem
>>  #----------------------------------------------------------------------------
>> -AC_CHECK_FILES(/proc/self/fd /proc/self/exe /proc/self/maps,
>> -    [ AC_DEFINE([HAVE_PROC], 1, [can use /proc filesystem]) ],
>> -    [])
>> +# jfm AC_CHECK_FILES(/proc/self/fd /proc/self/exe /proc/self/maps,
>> +# jfm     [ AC_DEFINE([HAVE_PROC], 1, [can use /proc filesystem]) ],
>> +# jfm     [])
>> +AC_DEFINE([HAVE_PROC], 1, [can use /proc filesystem])
>
> It would be more elegant to keep the first test if [ $cross_compiling
> = "no" ] and to fall back to the second test only if [
> $cross_compiling = "yes" ]. The variable $cross_compiling is set by
> configure.

That sounds like a good idea.  I only know just enough about automake/
configure to be dangerous.

>> +/* Hack because glibc is not included,  but the math lib libgcc calls
>> it - jfm */
>> +
>> +void abort(void)
>> +{
>> +       vg_assert(0);
>> +}
>> +
>
> Against which Valgrind revision have you developed your patch ? An
> abort() function has been added to coregrind/m_main.c in r11172 on
> June 13, 2010.

I did my patches around the latest released version, 3.5.0.  It's good to
hear that it has been fixed in the "bleeding edge" version.

> Bart.

-mehaf