On 2023-10-07 02:42, Bruno Haible wrote:
If we start to use 'test "$arg"' here and there, at some point someone
will assume it's safe (because he doesn't remember the long rule), and
then we're too close to having an actual bug.
I guess it's OK for now, but when Solaris 10 stops being supporte
Paul Eggert wrote:
> >* If $ISNAND_LIBM is empty, it executes
> >test ''
> > which is invalid in at least some version of 'test'.
>
> Which version would that be? If there is one, it doesn't conform to
> POSIX and we should document it in the Autoconf manual.
>
>
> >* If $I
On 2023-10-05 13:06, Bruno Haible wrote:
Hi Paul,
+ if test $gl_cv_func_isnand_no_libm = yes || test "$ISNAND_LIBM"; then
This doesn't look right:
* If $ISNAND_LIBM is empty, it executes
test ''
which is invalid in at least some version of 'test'.
Which version would that be
Hi Paul,
> + if test $gl_cv_func_isnand_no_libm = yes || test "$ISNAND_LIBM"; then
This doesn't look right:
* If $ISNAND_LIBM is empty, it executes
test ''
which is invalid in at least some version of 'test'.
* If $ISNAND_LIBM is -lm, it executes
test -lm
which is always
insertions(+), 9 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index 3ace239450..f09de5233e 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,11 @@
2023-10-05 Paul Eggert
+ isnan: slightly simplify configuration
+ * m4/isnand.m4 (gl_FUNC_ISNAND):
+ * m4/isnanf.m4 (gl_FUNC_ISNANF
