Re: GCC w/ C23 changes LDBL_EPSILON of IBM long double, causes test-float-h to fail

[C. Neidahl]

Am 06.07.25 um 21:37 schrieb Paul Eggert:

Thanks for reporting that problem. I installed the attached patch into
Gnulib on Savannah, which I hope fixes things for you. Please give it a
try on your platform.

Thanks, the patch does indeed fix that check!

The test is now failing on the equality check of LDBL_NORM_MAX and LDBL_MAX.

My GCC reports:

#define __LDBL_MAX__ 1.79769313486231580793728971405301e+308L
#define __LDBL_NORM_MAX__ 8.98846567431157953864652595394501e+307L





- Cosima

LDBL_MAX gets overridden in the C99 block of the generated float.h due to the platform & compiler, but still ends up in a similar 1.79...e+308 range. LDBL_NORM_MAX is taken as-is when defined, which it is in C23, so it doesn't get set to LDBL_MAX by your float.h.

publickey - opna2608@protonmail.com - 0xC802C863.asc
Description: application/pgp-keys

signature.asc
Description: OpenPGP digital signature