[PATCH] tests: add chmod.test

[Dmitry V. Levin]

On Sat, Mar 05, 2016 at 11:11:26AM +0530, Anchit Jain wrote:
> 
> 
> 05.03.2016, 03:12, "Dmitry V. Levin" <ldv at altlinux.org>:
> > On Fri, Mar 04, 2016 at 08:45:48PM +0000, Anchit Jain wrote:
> > [...]
> >>  +#include "tests.h"
> >>  +#include <assert.h>
> >
> > As you no longer use assert, no need to include <assert.h>.
> >
> >>  +#include <fcntl.h>
> >>  +#include <stdio.h>
> >>  +#include <unistd.h>
> >>  +#include <errno.h>
> >>  +#include <sys/syscall.h>
> >>  +
> >>  +#if defined __NR_chmod
> >
> > The only two headers that have to be included before __NR_chmod test
> > are "tests.h" and <sys/syscall.h>.
> >
> > In case you are in pedantic mood, all the rest could be included
> > after __NR_chmod test, like in tests/alarm.c and many similar tests.
> >
> >>  +
> >>  +int
> >>  +main(void)
> >>  +{
> >>  + static const char fname[] = "chmod_test_file";
> >>  + int fd = open(fname, O_CREAT|O_RDONLY, 0400);
> >>  +
> >>  + if (fd == -1)
> >>  + perror_msg_and_fail("open");
> >>  +
> >>  + if (syscall(__NR_chmod, fname, 0600) != 0)
> >>  + {
> >
> > The opening brace should be placed on "if" line.
> >
> >>  + if (errno == ENOSYS)
> >>  + perror_msg_and_fail("chmod ENOSYS");
> >
> > If chmod has failed with ENOSYS, it might be a normal condition
> > on some architectures, so it shouldn't be treated as a test error.
> > Just print it using
> >
> >         printf("chmod(\"%s\", 0600) = -1 ENOSYS (%m)\n", fname);
> 
> But the test should not continue after this,

Why not?  It just shouldn't test for ENOENT decoding.


-- 
ldv
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20160305/b3a17758/attachment.bin>