Re: [PATCH] tests: use xmalloc to allocate implementation array

[Siddhesh Poyarekar via Libc-alpha <libc-alpha@sourceware.org>]

On 7/28/21 12:55 PM, Florian Weimer wrote:
> * Siddhesh Poyarekar via Libc-alpha:
> 
>> diff --git a/benchtests/bench-string.h b/benchtests/bench-string.h
>> index fd25264417..03de372cff 100644
>> --- a/benchtests/bench-string.h
>> +++ b/benchtests/bench-string.h
>> @@ -18,6 +18,7 @@
>>   
>>   #include <getopt.h>
>>   #include <sys/cdefs.h>
>> +#include <support/support.h>
>>   
>>   /* We are compiled under _ISOMAC, so libc-symbols.h does not do this
>>      for us.  */
>> @@ -200,8 +201,8 @@ static impl_t *impl_array;
>>   		 skip = impl;						      \
>>   	       else							      \
>>   		 impl_count++;						      \
>> -	     a = impl_array = malloc ((impl_count + func_count) *	      \
>> -				   sizeof (impl_t));			      \
>> +	     a = impl_array = xmalloc ((impl_count + func_count) *	      \
>> +				       sizeof (impl_t));		      \
>>   	     for (impl = __start_impls; impl < __stop_impls; ++impl)	      \
>>   	       if (impl != skip)					      \
>>   		 *a++ = *impl;						      \
> 
> I'm surprised the benchtests link against libsupport.  There's some
> non-trivial setup code involved in libsupport for failure handling, and
> that gets linked into xmalloc.  It might affect the benchmarks.

Yeah, it may not hurt to wean them away; it's really just the string 
benchmarks that were copied over from tests and adapted.  I'll use 
programs/xmalloc.h here then.

Siddhesh