Eric Blake wrote:
> Is readlink already being replaced due to other bugs? (I guess I need
> to look more closely at your logs.)
It is not being replaced:
HAVE_READLINK='1'
REPLACE_READLINK='0'
> If so, then I think the replacement
> should work around this. If not, I'm perfectly fine relaxing the test
> (there's already a number of places where we can't guarantee the same
> errno as required by POSIX without lots of bloat; as long as a
> reasonable error happens, it's generally okay for a looser test).
OK, I'm applying this:
2010-07-31 Bruno Haible <br...@clisp.org>
readlink: Relax test a bit.
* tests/test-readlink.h (test_readlink): Allow different errno value
when readlink is called with a file name that ends in / and refers to
a file.
Suggested by Eric Blake.
Reported by Rainer Tammer.
--- tests/test-readlink.h.orig Sat Jul 31 12:21:12 2010
+++ tests/test-readlink.h Sat Jul 31 12:20:44 2010
@@ -51,7 +51,7 @@
ASSERT (errno == EINVAL);
errno = 0;
ASSERT (func (BASE "file/", buf, sizeof buf) == -1);
- ASSERT (errno == ENOTDIR);
+ ASSERT (errno == ENOTDIR || errno == EINVAL); /* AIX yields EINVAL */
/* Now test actual symlinks. */
if (symlink (BASE "dir", BASE "link"))
