Am 17.05.26 um 5:56 PM schrieb Richard Biener:
On Fri, May 15, 2026 at 11:17 PM Harald Anlauf <[email protected]> wrote:
Hi Richi,
Gesendet: Freitag, 15. Mai 2026 um 22:09
Von: "Richard Biener" <[email protected]>
An: "Harald Anlauf" <[email protected]>
CC: fortran <[email protected]>
Betreff: Re: PR125311: gfortran, d, or driver issue?
Conflicting options cannot exist, I guess if gfortran was first it’s a D bug.
OTOH the Alias is conditional on D, but maybe Alias processing isn’t properly
language specific or, the conflict isn’t reported by the gen script …
Well, fortran added -fcheck=... options with r0-92144, and d's lang.opt was
added with r9-3689. So gfortran was clearly first.
Does the Alias processing take into account the language? I could not figure
this out.
(I would naively expect that this happens for each language separately, as one
could mix e.g. fortran and c source files on the gfortran command line).
All .opt files are combined to a single one, the language
specifications apply (accumulate), but i have no idea whether
they are even relevant for 'Alias' ...
I can see the combinations in the generated options.cc,
and I even see an #error directive inserted when there is some
obvious conflict detected early enough.
In my previous mail I was pointing at the commit where D was
merged. However, the issue does not show up in my tests until
11.5.0, but in 12.2.0, so there was some later change that lead
to the issue observed now.
I've therefore changed the component to "driver".
Thanks,
Harald
Harald
