[PATCH] Fixup gdb.python/py-value.exp for bare-metal aarch64-elf

Luis Machado lgustavo@codesourcery.com
Fri Oct 7 15:07:00 GMT 2016


On 10/07/2016 09:48 AM, Yao Qi wrote:
> On Fri, Oct 7, 2016 at 1:08 PM, Luis Machado <lgustavo@codesourcery.com> wrote:
>>
>> I went back and read the standard and we're dealing with a freestanding
>> environment for bare metal here.
>>
>> The descriptions above seem to make sense for a hosted environment, but not
>> for a freestanding one, correct?
>>
>
> IMO, bare metal != freestadning environment.  Since "main" function is used,
> it is a hosted environment.  See
> https://gcc.gnu.org/onlinedocs/gcc/Standards.html

I'm slightly confused. Are you implying that using main makes this a 
hosted environment even though there is no underlying OS facility at play?

The documentation states freestanding environments can use any type of 
startup routine they want, only having to honor a small subset of clauses.

This seems to imply that the use of "main" is allowed and the lack of 
underlying OS facilities make it freestanding.

Skipping the discussion of whether we have a hosted x freestanding 
environment, the testcase itself uses main/argc/argv. Does that mean it 
is supposed to be exercised only on hosted environments or only on 
targets providing sane argc/argv values?

>
> "a hosted environment, which is not required, in which all the library
> facilities are provided and startup is through a function int main
> (void) or int main (int, char *[])."
>
> On the other hand, in the C standard, function "main" is only mentioned in
> the section of "5.1.2.2 Hosted environment".
>

They are mentioned there because it is a requirement for a hosted 
environment but, as mentioned above, freestanding programs could have a 
function called main as the startup function. There is no restriction on 
that part, correct?

More importantly, does it really matter, from a gdb testsuite point of 
view, whether the freestanding implementation provides main/argc/argv or 
not? The tests get exercised anyway, with only a small adjustment being 
required to make it not crash due to an assumption that is not always 
true (presence of argc/argv).



More information about the Gdb-patches mailing list