KO Myung-Hun wrote:
> > Does the init.sh test failure go away with this patch?
> >
>
> 'shopt' related failure disappeared.
OK, I've pushed the patch.
> However, 'invalid path dir' is
> still. I looked into that. And ':' was a problem. Because ':' is not a
> path separator on OS/2. Instead it's used as a separator of a drive
> letter and directory parts.
>
> I tried to use $PATH_SEPARATOR. However, it has nothing.
>
> Any idea ?
PATH_SEPARATOR is the right approach, yes. In m4/progtest.m4 and m4/lib-ld.m4
you find a code snippet that determines PATH_SEPARATOR. You can copy this to
init.sh.
2017-10-07 Bruno Haible <br...@clisp.org>
test-framework-sh: Don't require bash on Windows and OS/2.
Reported by KO Myung-Hun.
* tests/test-init.sh: Use 'shopt' only when running in bash.
diff --git a/tests/init.sh b/tests/init.sh
index 470605c..0821c57 100644
--- a/tests/init.sh
+++ b/tests/init.sh
@@ -250,7 +250,7 @@ test -n "$BASH_VERSION" && unalias -a
# That is part of the shell-selection test above. Why use aliases rather
# than functions? Because support for hyphen-containing aliases is more
# widespread than that for hyphen-containing function names.
-test -n "$EXEEXT" && shopt -s expand_aliases
+test -n "$EXEEXT" && test -n "$BASH_VERSION" && shopt -s expand_aliases
# Enable glibc's malloc-perturbing option.
# This is useful for exposing code that depends on the fact that
