On Thu, 30 Oct 2025 at 13:35, Joel Sherrill <[email protected]> wrote:
>
>
>
> On Thu, Oct 30, 2025 at 2:56 AM Richard Biener <[email protected]>
> wrote:
>>
>>
>> I wonder if it would be generally useful to know when options were introduced
>> and whether we should document that in the manual even? OTOH that suggests
>> version checks in configure when functional testing is what should be done.
>
>
> I think this would be useful. I tripped across the warnings/version issue
> while adding
> more warning flags to RTEMS builds. We use GCC 15 primarily but some targets
> have to use slightly older versions (e.g. GCC 14 for nios2, microblaze is 13,
> etc.).
> I would add a warning and someone would notice quickly that it wasn't in GCC
> x.
> This is without considering release branches which are locked onto a specific
> less than current GCC version.
>
> Isn't functional testing for a new flag/feature added at the same time? Why
> would
> you have to test for supporting a feature?
I think what Richi meant is not about GCC's own configure script. I
think he meant that if we document which version added features, then
it encourages people to write configure scripts that test versions
("do we have at least GCC 14?") instead of testing "does $CC support
the -Wfoo warning flag?"
