On Thu, Jul 24, 2025 at 08:48:05PM +0530, Kishan Parmar wrote: > Hello, > > Changes from V1: > Corrected Commit Message. > > Configure script previously accepted unsupported --with-long-double-64 > option without any warning, leading to confusion. This option was > mistakenly documented in install.texi but never handled by configure > script, resulting in it being silently ignored. > > Patch updates configure script to error out when --with-long-double-64 > or --without-long-double-64 is used, and instructs to use the correct > --without-long-double-128 or --with-long-double-128 option instead. > > Additionally, the mistaken documentation reference in install.texi has > been corrected to reflect the actual supported option. > > 2025-07-23 Peter Bergner <[email protected]> > Kishan Parmar <[email protected]> > > ChangeLog: > * configure.ac: Error out on --with-long-double-64. > * configure: Regenerate. > > gcc/ChangeLog: > * doc/install.texi: Remove incorrect mention of --with-long-double-64.
I thought that I had used --with-long-double=64 several years ago when I was building 3 sets of GCC and comparing the results. 1: I built a compiler default to 128-bit IEEE long doubles; 2: I built a compiler default to 128-bit IBM long doubles; (and) 3: I built a compiler default to 64-bit long doubles. However, I see in the log for my build scripts, that on June 6th, 202 I changed from using --with-long-double-64 to using --without-long-double-128. So perhaps it stopped working in the 2022 through 2023 time frame. Your patch looks fine to me (though at this point, Segher still has to approve it). I like that you added code to actually check if the user used it by accident. -- Michael Meissner, IBM PO Box 98, Ayer, Massachusetts, USA, 01432 email: [email protected]
