On Fri, Aug 20, 2021 at 03:47:54PM +0200, Tobias Burnus wrote:
> On 20.08.21 13:56, Jakub Jelinek wrote:
>
> > On Fri, Aug 20, 2021 at 01:50:00PM +0200, Tobias Burnus wrote:
> > > Comments? OK?
> > LGTM (except that the last hunk won't apply anymore).
>
> Now applied as r12-3044; I have now changed it to %wd ...
>
> ... but as discussed in another email in the thread, I think the
> gfc_error should be removed (as unused); possibly with some additional
> testcases running into the resolve errors for those (out of bound).
Yeah, it makes sense that if there are errors in user code, they are
diagnosed during resolve, no matter whether simplification is possible
or not and simplification either asserts the errors aren't there or just
punts on the simplification, but doesn't repeat those diagnostics.
Jakub
