Hi Eric,
> I just noticed this failure on NetBSD 5.0.2, i386-unknown-netbsdelf5.0.2:
>
> ../../gltests/test-nl_langinfo.c:115: assertion failed
> [1] Abort trap (core dumped) LC_ALL=${LOCALE_...
> FAIL: test-nl_langinfo.sh
>
> LC_ALL=fr_FR.ISO8859-1 ./test-nl_langinfo 1
>
> 114 const char *currency = nl_langinfo (CRNCYSTR);
> (gdb)
> 115 ASSERT (strlen (currency) >= (pass > 0 ? 1 : 0));
> (gdb) p currency
> $1 = 0xbbbc7761 ""
> (gdb) p pass
> $2 = 1
>
> $ grep NL_LANGINFO gllib/Makefile
> GNULIB_NL_LANGINFO = 1
> HAVE_NL_LANGINFO = 1
> REPLACE_NL_LANGINFO = 0
> -e 's|@''GNULIB_NL_LANGINFO''@|$(GNULIB_NL_LANGINFO)|g' \
> -e 's|@''HAVE_NL_LANGINFO''@|$(HAVE_NL_LANGINFO)|g' \
> -e 's|@''REPLACE_NL_LANGINFO''@|$(REPLACE_NL_LANGINFO)|g' \
OK, this means it is really the native NetBSD nl_langinfo() with the NetBSD
CRNCYSTR argument - since gnulib's langinfo.h replacement does not replace
CRNCYSTR when <langinfo.h> is present - which returns the empty string.
It is not formally a bug, since glibc also returns the empty string for
nl_langinfo (CRNCYSTR) in some locales (the C locale for example).
It's simply unexpected behaviour. Only the test needs to be fixed. I'm
committing this:
2010-10-21 Bruno Haible <br...@clisp.org>
nl_langinfo test: Avoid test failure on NetBSD 5.
* tests/test-nl_langinfo.c (main): Relax test of nl_langinfo(CRNCYSTR).
Reported by Eric Blake.
--- tests/test-nl_langinfo.c.orig Fri Oct 22 01:58:52 2010
+++ tests/test-nl_langinfo.c Fri Oct 22 01:57:21 2010
@@ -112,7 +112,11 @@
/* nl_langinfo items of the LC_MONETARY category */
{
const char *currency = nl_langinfo (CRNCYSTR);
- ASSERT (strlen (currency) >= (pass > 0 ? 1 : 0));
+ ASSERT (strlen (currency) >= 0);
+#if !defined __NetBSD__
+ if (pass > 0)
+ ASSERT (strlen (currency) >= 1);
+#endif
}
/* nl_langinfo items of the LC_MESSAGES category */
ASSERT (strlen (nl_langinfo (YESEXPR)) > 0);
