| 41bb9ed1b29567b541675412d40730b7ca817bdc |
18-Feb-2016 |
Jan Stancek <jstancek@redhat.com> |
open_posix: add SAFE_PFUNC macro
Simple macro to exit test and print error if a pthread_* function
returned non-zero.
Acked-by: Cyril Hrubis <chrubis@suse.cz>
Signed-off-by: Jan Stancek <jstancek@redhat.com>
afe_helpers.h
|
| e2d32e4c57b4a83841288c4010a653885fdb9606 |
12-Nov-2015 |
Jan Stancek <jstancek@redhat.com> |
open_posix_testsuite: add set_affinity_single()
Couple testcases rely on set_affinity(), to limit usage of CPUs
to a single one. However they hardcode CPU number to "0", which
is causing problems on systems where CPU0 is not available (for
example it's offline, etc.)
This patch adds set_affinity_single() (which sets affinity to
first single online CPU) and replaces all calls to set_affinity(0).
Signed-off-by: Jan Stancek <jstancek@redhat.com>
Acked-by: Cyril Hrubis <chrubis@suse.cz>
ffinity.h
|
| 520f02a0cb9200cca5b837b78079be7e4727004b |
29-Oct-2015 |
Cyril Hrubis <chrubis@suse.cz> |
include/affinity.h: define _GNU_SOURCE on Linux
Otherwise the affinity cannot be used.
This is better than setting _GNU_SOURCE in the testcases
unconditionally, the affinity.h header must be include first though.
Another option would be compiling the affinity function into object file
and linking it with testcases (which wouldn't need to expose the GNU
extensions in the testcases).
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
ffinity.h
osixtest.h
|
| f45c69c30ed6a5a0813df80d61d52d05d8baa181 |
21-Oct-2015 |
Cyril Hrubis <chrubis@suse.cz> |
openposix: sched_setparam: Extract get_ncpu()
Move the get_ncpu() function into the file in include/ so that we have
only one copy of the code and the code can be reused.
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
cpu.h
|
| 338d209d9aa1b58ecc9849dd5ac919af232c6ecd |
09-Oct-2014 |
Zeng Linggang <zenglg.jy@cn.fujitsu.com> |
sigaddset: fix ARRAY_SIZE undefined
open_posix_testsuite/conformance/interfaces/sigaddset/4-1.c compiles failed
since b863a0bd, because of ARRAY_SIZE is undefined in open_posix_testsuite.
Signed-off-by: Zeng Linggang <zenglg.jy@cn.fujitsu.com>
Acked-by: Cyril Hrubis <chrubis@suse.cz>
osixtest.h
|
| 4d1552df78b4a6d69b77e97d0f3ef7371c80c7f8 |
03-Jul-2014 |
Garrett Cooper <yanegomi@gmail.com> |
Improve *BSD port again
- Use setpgid instead of setpgrp
- Add *BSD support
- Mute some -Wuninitialized warnings
- Mute some clang warnings
- Add a simple script for printing out the overall run summary
Signed-off-by: Garrett Cooper <yanegomi@gmail.com>
Sponsored-by: EMC / Isilon Storage Division
oatime.h
|
| 1c84cdb09319a8ca80ee59fb886cc3532846a0e6 |
14-May-2014 |
Cyril Hrubis <chrubis@suse.cz> |
lib, openposix: Fix mount flags detection code.
The problem is that there are two records for root filesystem in the
/proc/mounts. One real and one dummy that looks like:
rootfs / rootfs rw 0 0
Now the problem is that the code matched the first one that has
incomplete flags and returned wrong results.
Previous attempt to fix this was to ignore all pseudo filesystem that
does have a path to a device in mnt_fsname.
That however excludes tmpfs that is commonly used for /tmp filesystem
nor it works for nfs mounts that starts with IP address.
This patch changes it to skip only the problematic entry with "rootfs"
in the mnt_fsname.
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
oatime.h
|
| 4ba1c46742f4b272d14b042a8e286a1a518f539f |
27-Mar-2013 |
Cyril Hrubis <chrubis@suse.cz> |
open_posix_testsuite/include/noatime: Fix.
* Remove unused variable.
* Add parenthesis
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
oatime.h
|
| 6de744ecba6043f5fb79f618bdea3592d006f31d |
31-Oct-2012 |
Cyril Hrubis <chrubis@suse.cz> |
open_posix_testsuite/.../mmap/13-1.c: Add check.
This commit creates a function for checking if
path is on a filesystem that is mounted noatime
as well as it adds this check to the correspoinding
mmap test.
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
oatime.h
|
| 4548c6cf9bcdd96d8303caa4130ab638b61f8a30 |
19-Oct-2012 |
Wanlong Gao <gaowanlong@cn.fujitsu.com> |
Update FSF address
find . -type f -exec sed -i 's/675 Mass Ave, Cambridge, MA 02139, USA/51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA/g' {} \;
find . -type f -exec sed -i 's/59 Temple Place, Suite 330, Boston, MA 02111-1307 USA/51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA/g' {} \;
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
ffinity.h
|
| 37550cf5a86c7ca7424a2a318dd64e550f13f5df |
18-Oct-2012 |
Chris Dearman <chris@mips.com> |
Remove whitespace at end of line
This is a scripted change done using the following command:
find . -type d -name .git -prune -o \
-type f \! -name \*\~ \! -name .\#\* \
\! -name \*.gz \! -name \*.pdf \! -name \*.patch -print0 | \
xargs -0 sed -i -e 's/[\t ]*$//'
Signed-off-by: Chris Dearman <chris@mips.com>
ffinity.h
em_pattern.h
|
| 6dc93f3a486f0be6826733a68ee800c4ec9fdd18 |
08-Jun-2012 |
Cyril Hrubis <chrubis@suse.cz> |
Fix fork/2-1.c
The allocated chunk corectness is asserted by casting it to double and
assigning a value. Later it's compared to the same value. The problem
with such approach is that in some cases the comparsion may be false,
due to implicit casts and nature of floating point operations.
So this patch removes the questionable floating point arithmetics and
instead of that fills allocated memory with pseudo random sequence which
is asserted later in the child.
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
em_pattern.h
|
| ae4f2fecbce0c4ce78c2ca0af1b47a4141ae260f |
12-Aug-2011 |
Peter W Morreale <pmorreale@novell.com> |
sched_yield 1-1 fix
The current sched_yield 1-1 test floods available processors with
SCHED_FIFO processes (nr_cpus - 1), then starts two threads that verify
the behavior of sched_yield and affinitizes these threads to the remaining
processor.
This is silly and not useful if we assume that setting affinity works
correctly (which this type of test depends upon)
So rewrite the test to merely use a single processor. We test for success
by starting a child that loops doing sched_yield()s and then have the
parent kill it. The parent can only kill it if sched_yield() works. (ok,
and affinity too...)
[Note: this patch moves the non portable affinity code into common
header that needs to be ported to your favorite UNIX]
Signed-off-by: Peter W. Morreale <pmorreale@novell.com>
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
ffinity.h
|
| 68fae61b00b160964813d3ab6e97d9308e503e1d |
25-Feb-2011 |
Cyril Hrubis <chrubis@suse.cz> |
Fix clock_getcpuclockid test.
The test was expecting that two clockid_t poiting to the same clock must
be the same, which is not required by POSIX. Now the tests, if ids are
different, check if reported time is aproximetly the same.
This also introduces timespec_nsec_diff() function, that should be used
to get nanosecond differerence of two struct timespec values.
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
imespec.h
|
| 4e69f59f171f2659bea8c22cfad44b6b6a73e340 |
15-Sep-2010 |
Garrett Cooper <yanegomi@gmail.com> |
Make errors w/ bsd and gnu code into warnings for end-users.
Signed-off-by: Garrett Cooper <yanegomi@gmail.com>
osixtest.h
|
| b90b372e4a1e06c7a1e08ca0e2f291b543592faf |
20-Jun-2010 |
Garrett Cooper <yanegomi@gmail.com> |
Revert _BSD_SOURCE vs _GNU_SOURCE check.
The _GNU_SOURCE guard in ctypes enables the kitchen sink as far as
standards are concerned, which enables _BSD_SOURCE. So let's note the
real culprit for non-compliance.
Signed-off-by: Garrett Cooper <yanegomi@gmail.com>
osixtest.h
|
| 80886b520423c015b63cff03082e274147042243 |
20-Jun-2010 |
Garrett Cooper <yanegomi@gmail.com> |
Fix POSIX compliancy of these testcases.
Most of these testcases had issues building on FreeBSD dealing with missing
headers and a few glibc ones (features.h).
There are still a few outstanding build related issues, but those will
be worked out shortly with a bit more research.
Signed-off-by: Garrett Cooper <yanegomi@gmail.com>
osixtest.h
|
| abeace112ad2b6435909b462558ceb8184feab20 |
22-Aug-2009 |
subrata_modak <subrata_modak> |
I have noticed build warnings for the following test case. This patch will fix warnings by removing unused variables in test code.
Before patch:
==========
conformance/interfaces/mq_send/5-1: build: FAILED: Compiler output: cc1: warnings being treated as errors
In file included from conformance/interfaces/mq_send/5-1.c:40:
include/mq_send.h: In function 'sync_pipe_wait_select':
include/mq_send.h:49: warning: unused variable 'buf'
conformance/interfaces/mq_send/5-1.c: In function 'main':
conformance/interfaces/mq_send/5-1.c:107: warning: unused variable 'r'
conformance/interfaces/mq_send/5-1.c:105: warning: unused variable 'act'
==========
After patch:
==========
conformance/interfaces/mq_send/5-1: build: PASS
conformance/interfaces/mq_send/5-1: link: PASS
Signed-off-by: Naresh Kamboju <naresh.kernel@gmail.com>,
q_send.h
|
| 194127e3bab719dbd1a4aee1df85146b00cde5f7 |
10-Aug-2009 |
subrata_modak <subrata_modak> |
I have noticed issues for patch applied on 12-Dec-2008. Because it is fixing the mq_send/5-1.c problem, but at the same time patch is affecting the building of few test cases in the same directory. I have listed the build log before and after the patch. So I have modified the patch such a way that it should not affect other test cases. Modifications:
1. remove function definitions from “include/posixtest.h”
2. add function definitions to new header file “include/mq_send.h”
3. add #include ”mq_send.h” line to mq_send/5-1.c test cases
I have attached modified patch and below. Please review the same. Signed-off-by: naresh kamboju <naresh.kernel@gmail.com>.
q_send.h
osixtest.h
|
| ffa73254942e0fa255978083b593a8e74101eb09 |
12-Dec-2008 |
subrata_modak <subrata_modak> |
Version 2: The test failed a few hours after I sent this off. I think it was caused by a pid collision which caused the test to use the same queue as an earlier failed test that didn't close out its queue. This version of the test uses O_EXCL to avoid this. And it cleans up the queue on every failure case. This test was racy. It relied on signals interrupting sleeps to do syncronization between processes. It also didn't cleanup the queue on the failure cases and assumed that its queue didn't already exist. This patch fixes these issues. Reported failures: https://bugzilla.novell.com/show_bug.cgi?id=375897. Instead use a syncronization pipe to have the child thread send messages to the parent thread (with reasonable timeouts since this is testing blocking calls).
# Apply to root of ltp-full-20081031 tree
$ cd testcases/open_posix_testsuite/
$ make conformance/interfaces/mq_send/5-1.test
$ conformance/interfaces/mq_send/5-1.test
Signed-off-by: Brandon Philips <bphilips@suse.de>
osixtest.h
|
| 0dc076565f772bb1953209fb69ea150b494aaa40 |
03-Jun-2005 |
robbiew <robbiew> |
Updated the open_posix_testsuite:
------------------------------------------
June 3, 2005 Version 1.5.1 Released
* PTS-1.5.1 has a plenty of AIO interface test additions and bug-fixes
by Sebastien Decugis. Also there are a lot of bug fixes by other people.
Thanks to
Sebastien Decugis and other people for continued support and
contributions.
* This release has been tested on glibc-2.3.3 and kernel 2.6.10 with
libposix-aio-0.3.
The test result for this release can be found here:
http://posixtest.sourceforge.net/testpass/PTS_1.5.1-2.6.10-PAIO_ia32.htm
* PTS Downloads and Release information:
https://sourceforge.net/project/showfiles.php?group_id=64592&release_id=
332108
PTS Notes and Changelog:
https://sourceforge.net/project/shownotes.php?release_id=332108
Thanks
Ling Yu
POSIX Test Suite maintainer
http://posixtest.sf.net
osixtest.h
|
| 19b5dc744e2b73f745491e5a081873f6c1417528 |
03-Jun-2005 |
robbiew <robbiew> |
Removed old version of Open POSIX Test Suite (OPTS).
osixtest.h
|
| 1b8a2eaac9978e1f9a829e5e38dc39c892c77c75 |
04-May-2004 |
robbiew <robbiew> |
Update to 1.4.1
osixtest.h
|
| e17215e476cd0d2b84ae8be11b1dbfb3b9016dcb |
04-May-2004 |
robbiew <robbiew> |
Update to 1.4.1
osixtest.h
|
| c911fa64e3ec4bb25012a36ebdd2ad8f747d9ea8 |
03-Jul-2003 |
robbiew <robbiew> |
Added the 1.1.0 version of the Open POSIX Test Suite.
osixtest.h
|