https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89639
--- Comment #6 from Steve Kargl <sgk at troutmask dot apl.washington.edu> --- On Sat, Mar 09, 2019 at 05:50:43PM +0000, dave.anglin at bell dot net wrote: > >> > > Thanks. When I wrote the test, I tried to skip the > > REAL128 case with the "if (real128 >) then" ... > > blocks, and completely overlooked that real128 is > > a named constant. This means the conversion functions > > real(..., real128) will try ti real128 = -1, which s/will try ti/will try to use > > of fails. > Now I understand why this just fails on hppa-linux.i > hppa-hpux has REAL128. I probably should change skip > reason to "No REAL128". > I added predefine macros for cpp. I'll someday move this testcase to ieee_9.F90, and use #if __GFC_REAL16__ to guard againt the issue. "Someday" may not be too soon.