https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78935
--- Comment #3 from Dominique d'Humieres <dominiq at lps dot ens.fr> --- > Workaround (4) in the original submission is mistakenly redundant. What is "Workaround (4)"? > It's an interestingly fragile ICE that disappears with seemingly > minor changes to the code. For example, it also disappears if > the first allocation argument is a scalar. AFAICT the ICE disappears if the two parts are in the same file.
