The "symlink ownership not supported" condition is expected to be indicated by EOPNOTSUPP, as opposed to the "no support for ownership", which is indicated by ENOSYS.
The latter condition is checked for earlier in this test on a directory (rather than on a symlink). And if ENOSYS is raised for a symlink, but not for a directory, this would be against the expected behavior and should make the test fail. This change helps this test to work correctly (i.e., to be skipped) if gnulib's own implementation of lchown is used. Such convention concerning the meaning of EOPNOTSUPP was also stated in The Open Group Base Specifications Issue 6 IEEE Std 1003.1, 2004 Edition (http://pubs.opengroup.org/onlinepubs/009604599/functions/lchown.html); however, the EOPNOTSUPP case is removed in Issue 7, 2018 Edition (http://pubs.opengroup.org/onlinepubs/9699919799/functions/lchown.html). gnulib's own implementation in lchmod.c follows this convention since a62be9f4039b4499cfbb76e394cad2259d03fa84 (2004-08-07): int lchown (const char *file, uid_t uid, gid_t gid) { # if HAVE_CHOWN # if ! CHOWN_MODIFIES_SYMLINK struct stat stats; if (lstat (file, &stats) == 0 && S_ISLNK (stats.st_mode)) { errno = EOPNOTSUPP; return -1; } # endif return chown (file, uid, gid); # else /* !HAVE_CHOWN */ errno = ENOSYS; return -1; # endif } --- tests/test-lchown.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/test-lchown.h b/tests/test-lchown.h index bc10a5038..b1a7d4b12 100644 --- a/tests/test-lchown.h +++ b/tests/test-lchown.h @@ -124,7 +124,7 @@ test_lchown (int (*func) (char const *, uid_t, gid_t), bool print) return 77; } result = func (BASE "dir/link2", -1, -1); - if (result == -1 && errno == ENOSYS) + if (result == -1 && errno == EOPNOTSUPP) { ASSERT (unlink (BASE "dir/file") == 0); ASSERT (unlink (BASE "dir/link2") == 0); -- 2.17.1
