https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69658
--- Comment #3 from Paolo Carlini <paolo.carlini at oracle dot com> --- Hi Jakub and thanks. I'm looking at your patchlet and putting it together my notes at the time, and IMO it makes a lot of sense: I even remember briefly wondering if calling reshape_init redundantly could be an issue :( Anyway, your patch passes all my regression tests so far. Are you going to send it to the mailing list? Ay this time I would only recommend adding a short comment before the new conditional explaining exactly what you explained here, that is that in the other cases reshape_init has been called already and where.
