Re: [PATCH 1/2] string: _FORTIFY_SOURCE=3 using __builtin_dynamic_object_size

[Siddhesh Poyarekar <siddhesh@sourceware.org>]

On 12/11/20 12:40 AM, Paul Eggert wrote:
> On 12/10/20 10:13 AM, Siddhesh Poyarekar via Libc-alpha wrote:
> 
>> +checks are applied. If defined to @math{3}, @theglibc{} attempts to use
>> +compiler-specific checks that may be more computationally expensive.
> 
> Please change "attempts to use compiler-specific checks" to "may also 
> use checks". The documentation need not mention compilers here, and the 
> "also" documents that 3 subsumes 2.

Thanks, I'll fix this.

>>  # elif !__GNUC_PREREQ (4, 1)
>>  #  warning _FORTIFY_SOURCE requires GCC 4.1 or later
>> -# elif _FORTIFY_SOURCE > 1
>> -#  define __USE_FORTIFY_LEVEL 2
>> +# elif _FORTIFY_SOURCE > 2
>> +#  define __USE_FORTIFY_LEVEL 3
>>  # else
>> -#  define __USE_FORTIFY_LEVEL 1
>> +#  define __USE_FORTIFY_LEVEL _FORTIFY_SOURCE
>>  # endif
> 
> The existing code warns if you define _FORTIFY_SOURCE on a platform that 
> doesn't support _FORTIFY_SOURCE. Shouldn't the revised code warn if you 
> define _FORTIFY_SOURCE to 3 on a platform that doesn't support that? 
> Otherwise people may be lulled into the belief that -D_FORTIFY_SOURCE=3 
> means something even on platforms where it doesn't.

Thanks, that's a good idea, I'll add a check.  I suppose I ought to 
document in the NEWS too that _FORTIFY_SOURCE=3 at present only adds 
additional checks with llvm 9 and later and that gcc-10.2 does not have 
the needed support?

Thanks,
Siddhesh