Bruno Haible <[EMAIL PROTECTED]> wrote: >> The *printf-posix modules should not work fine > The *printf-posix modules should _now_ work fine. (Jim noted the fatal typo.) > > Jim, it would be useful to use the printf-posix module in coreutils: > 1) The 'printf' program uses all floating-point directives, even with 'L'. > 2) The 'seq' program started to print garbage on BeOS when it was changed to > use 'long double' values, where it used 'double' values before. You can see > from gnulib/doc/functions/printf.texi that this affects also other > platforms > than BeOS.
I've been thinking about that. Considering the size and age of the code involved, it sounds a little risky, but that is counterbalanced by all of the new tests. Do you feel like adding a few coreutils tests that currently fail on common platforms but that would succeed if coreutils switched? E.g., here: coreutils/tests/seq/basic coreutils/tests/misc/printf Have you tried and compared code size on a modern Linux/libc system?
