ltp-list — General Discussions

Re: [LTP] Syscall/io_* tests failing to compile

[Cyril H. <ch...@su...>]

Hi!
> I haven't modified any of the config files at all. Even i cross
> checked on couple machines and faced similar issues.

Hmm, you are cross compiling right? Have you installed the libaio-dev on
the host machine? Because that will not work since that contains native
binaries and you cannot link against them.

If that is the case you can either cross compile the libaio and set
right path for linker etc. or set up an arm virtual machine or chroot
with qemu arm emulation and compile LTP there.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] Syscall/io_* tests failing to compile

[$<ri...@nt...>]

I haven't modified any of the config files at all. Even i cross
checked on couple machines and faced similar issues.

On Thu, Sep 3, 2015 at 4:45 PM, Cyril Hrubis <ch...@su...> wrote:
> Hi!
>> I am seeing below errors while compiling syscalls from the latest
>> available LTP tar ball from sourceforge. My system is having latest
>> libaio-dev library as well.
>>
>> the same issue even observed for other io syscalls(io_getevents etc..)
>> as well. Do i need to setup any other environment to compile it
>> successfully.
>>
>> /usr/bin/arm-linux-gnueabihf-gcc -static -g -O2 -fno-strict-aliasing
>> -pipe -Wall -W -static -Wold-style-definition
>> -I/home/LTP/ltp-full-20150420/testcases/kernel/include
>> -I../../../../include -I../../../../include -static -pthread
>> -L../../../../lib  io_destroy01.c   -lltp  -o io_destroy01
>> /tmp/cc0f3RO7.o: In function `main':
>> /home/LTP/ltp-full-20150420/testcases/kernel/syscalls/io_destroy/io_destroy01.c:77:
>> undefined reference to `io_destroy'
>> collect2: ld returned 1 exit status
>> make: *** [io_destroy01] Error 1
>
> The -laio is missing from the command line, but since you are seeing the
> failure the HAVE_LIBAIO_H must be defined in config.h.
>
> So it's looks like you have hand edited config.h (which is wrong thing
> to do) after installing the library instead of reruning configure.
>
> --
> Cyril Hrubis
> ch...@su...



-- 
Thanks & Regards,
M.Srikanth Kumar.

Re: [LTP] Syscall/io_* tests failing to compile

[Cyril H. <ch...@su...>]

Hi!
> I am seeing below errors while compiling syscalls from the latest
> available LTP tar ball from sourceforge. My system is having latest
> libaio-dev library as well.
> 
> the same issue even observed for other io syscalls(io_getevents etc..)
> as well. Do i need to setup any other environment to compile it
> successfully.
> 
> /usr/bin/arm-linux-gnueabihf-gcc -static -g -O2 -fno-strict-aliasing
> -pipe -Wall -W -static -Wold-style-definition
> -I/home/LTP/ltp-full-20150420/testcases/kernel/include
> -I../../../../include -I../../../../include -static -pthread
> -L../../../../lib  io_destroy01.c   -lltp  -o io_destroy01
> /tmp/cc0f3RO7.o: In function `main':
> /home/LTP/ltp-full-20150420/testcases/kernel/syscalls/io_destroy/io_destroy01.c:77:
> undefined reference to `io_destroy'
> collect2: ld returned 1 exit status
> make: *** [io_destroy01] Error 1

The -laio is missing from the command line, but since you are seeing the
failure the HAVE_LIBAIO_H must be defined in config.h.

So it's looks like you have hand edited config.h (which is wrong thing
to do) after installing the library instead of reruning configure.

-- 
Cyril Hrubis
ch...@su...

[LTP] Syscall/io_* tests failing to compile

[$<ri...@nt...>]

Hi All,


I am seeing below errors while compiling syscalls from the latest
available LTP tar ball from sourceforge. My system is having latest
libaio-dev library as well.

the same issue even observed for other io syscalls(io_getevents etc..)
as well. Do i need to setup any other environment to compile it
successfully.

/usr/bin/arm-linux-gnueabihf-gcc -static -g -O2 -fno-strict-aliasing
-pipe -Wall -W -static -Wold-style-definition
-I/home/LTP/ltp-full-20150420/testcases/kernel/include
-I../../../../include -I../../../../include -static -pthread
-L../../../../lib  io_destroy01.c   -lltp  -o io_destroy01
/tmp/cc0f3RO7.o: In function `main':
/home/LTP/ltp-full-20150420/testcases/kernel/syscalls/io_destroy/io_destroy01.c:77:
undefined reference to `io_destroy'
collect2: ld returned 1 exit status
make: *** [io_destroy01] Error 1

-- 
Thanks & Regards,
M.Srikanth Kumar.

Re: [LTP] ltp: thp04.c test case

[Cyril H. <ch...@su...>]

Hi!
There was no response for about three months. If nobody stands up and
fixes the testcase I would go for dropping it since it does not work
reliably as it is.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Cyril H. <ch...@su...>]

Hi!
I've just pushed last patch that has been requested to be included in
the relese. Unless anybody objects I will proceed with tagging the git,
uploading the tarballs, etc. tomorrow.

If you haven't tested latest git yet, now is the last chance to do so.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] Test Anything Protocal output

[Cyril H. <ch...@su...>]

Hi!
> > > Well, if your re-writing it, than TAP could be added after that. Is this re-write
> > > a long term thing, or something your planning to complete in the next couple months?
> > 
> > The tst_res.c is one of the last files under the lib/ that hasn't been
> > cleaned up yet and I certainly want to do that before next release,
> > which would be in aprox. three months.
> 
> Ok, I can wait for that. Are you planning on doing it in steps, or just one big update
> for the file ? I ask only to see if there's something I could look at prior to the final release.

Just now I'm still in the planning and designing phase. I will likely
want to discuss the changes before they get commited to LTP, either as
patches on ML or as separate git repo. We can wire up TAP output at that
point.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Guangwen F. <fen...@cn...>]

Hi!

On 2015/09/01 23:35, Cyril Hrubis wrote:
> Hi!
>>> Now the only problem that that blocks the release on my side are
>>> failures of the newly rewritten netns testcases that were discussed
>>> last week.
>>>
>>
>> In RHEL5.11GA, ns_ifmove.c gives compilation error since 'IFLA_NET_NS_PID'
>> undeclared in the kernel.
> 
> Should be fixed in latest git, please test (don't forget to regenerate
> configure script).
> 

It's OK for compilation on RHEL5.11GA now, thanks!


Best Regards,
Guangwen Feng

Re: [LTP] [PATCH v2] ltp/pounder: Don't let the testing log output to /dev/tty

[Li W. <li...@re...>]

Hi,

On Tue, Sep 1, 2015 at 9:44 PM, Cyril Hrubis <ch...@su...> wrote:

> Hi!
> > +     out = fdopen(3, "w+");
>
> This is just complicated way of doing out = stderr;
>
> And I doubt that this will fail under normal circumstances.
>
> So what about doing just out = stderr; or out = stdout instead?
>


Totally agree.  actually I want do that at first. :)

Feel free to modify the patch.  "out = stdout"  is good.



>
> Otherwise it looks fine and I'm OK with applying this before the
> release.
>

Thanks!


>
> --
> Cyril Hrubis
> ch...@su...
>



-- 
Regards,
Li Wang
Email: li...@re...

Re: [LTP] Test Anything Protocal output

[<dw...@fi...>]

On Tue, Sep 01, 2015 at 05:42:29PM +0200, Cyril Hrubis wrote:
> Hi!
> > It looks like it could be changes in lib/tst_res.c line 404 in function tst_print().
> > There's other components to TAP but the change above it the biggest I think.
> 
> Getting 90% working would be easy enough, the real problem is making it
> work for all corner cases, which is the reason I would rather want to do
> this for a cleaner code.

Ok .

> > > I've started to work on this just recently. Just now I'm experimenting
> > > with rewriting it from scratch keeping most of the API unchanged (since
> > > we cannot change easily the API for > ~1000 testcases).
> > > 
> > > Then there is another problem, so far LTP testcases are executed using
> > > the ltp-pan test driver, which executes testcases, catches segfaults,
> > > etc. We would have to consider how that would interact with the output
> > > generated from testcases.
> > > 
> > > One option would be to get rid of most of the pan functionality by
> > > moving the test driver code into the test library, so that each testcase
> > > would fork first, then run the real test in it's own virtual terminal,
> > > etc.
> > 
> > Well, if your re-writing it, than TAP could be added after that. Is this re-write
> > a long term thing, or something your planning to complete in the next couple months?
> 
> The tst_res.c is one of the last files under the lib/ that hasn't been
> cleaned up yet and I certainly want to do that before next release,
> which would be in aprox. three months.

Ok, I can wait for that. Are you planning on doing it in steps, or just one big update
for the file ? I ask only to see if there's something I could look at prior to the final release.

Daniel 

Re: [LTP] Test Anything Protocal output

[Cyril H. <ch...@su...>]

Hi!
> It looks like it could be changes in lib/tst_res.c line 404 in function tst_print().
> There's other components to TAP but the change above it the biggest I think.

Getting 90% working would be easy enough, the real problem is making it
work for all corner cases, which is the reason I would rather want to do
this for a cleaner code.

> > I've started to work on this just recently. Just now I'm experimenting
> > with rewriting it from scratch keeping most of the API unchanged (since
> > we cannot change easily the API for > ~1000 testcases).
> > 
> > Then there is another problem, so far LTP testcases are executed using
> > the ltp-pan test driver, which executes testcases, catches segfaults,
> > etc. We would have to consider how that would interact with the output
> > generated from testcases.
> > 
> > One option would be to get rid of most of the pan functionality by
> > moving the test driver code into the test library, so that each testcase
> > would fork first, then run the real test in it's own virtual terminal,
> > etc.
> 
> Well, if your re-writing it, than TAP could be added after that. Is this re-write
> a long term thing, or something your planning to complete in the next couple months?

The tst_res.c is one of the last files under the lib/ that hasn't been
cleaned up yet and I certainly want to do that before next release,
which would be in aprox. three months.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Cyril H. <ch...@su...>]

Hi!
> > Now the only problem that that blocks the release on my side are
> > failures of the newly rewritten netns testcases that were discussed
> > last week.
> > 
> 
> In RHEL5.11GA, ns_ifmove.c gives compilation error since 'IFLA_NET_NS_PID'
> undeclared in the kernel.

Should be fixed in latest git, please test (don't forget to regenerate
configure script).

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] [PATCH v2] ltp/pounder: Don't let the testing log output to /dev/tty

[Cyril H. <ch...@su...>]

Hi!
> +	out = fdopen(3, "w+");

This is just complicated way of doing out = stderr;

And I doubt that this will fail under normal circumstances.

So what about doing just out = stderr; or out = stdout instead?

Otherwise it looks fine and I'm OK with applying this before the
release.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Cyril H. <ch...@su...>]

Hi!
> another solution could be to let ns_ifmove.c as it is (so it won't build)
> and only add a check for binaries:
> tst_check_cmds ns_create ns_exec ns_ifmove

We would have to ingore the return value from the ns_ifmove build, which
means custom rule in the Makefile, which is ugly.

> to the netns test code. On the other hand, this will produce unclear
> output from the netns testcases, you will only know that ns_ifmove
> binary is unavailable, but won't know why.
> 
> The other way, as you pointed out, would be to check for kernel version
> and propagate TCONF to the netns_helper.sh. The patch which adds what we
> need ('IFLA_NET_NS_PID') to the kernel is:
> http://lists.linuxfoundation.org/pipermail/containers/2007-September/007113.html

Kernel version is completly orthogonal problem here.

The compilation fails since the IFLA_NET_NS_PID is not present in
linux/if_link.h. I've just finished a patch that adds autoconf check for
the constant, adds ifdefs into the ns_ifmove and propagates TCONF from
ns_ifmove to the helper. This may still fail if somebody runs the test
on recent enough userspace with older kernel. But I doubt that this will
be the case, at least for any released distribution.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Matus M. <mma...@re...>]

Hi,

another solution could be to let ns_ifmove.c as it is (so it won't build)
and only add a check for binaries:
tst_check_cmds ns_create ns_exec ns_ifmove

to the netns test code. On the other hand, this will produce unclear
output from the netns testcases, you will only know that ns_ifmove
binary is unavailable, but won't know why.

The other way, as you pointed out, would be to check for kernel version
and propagate TCONF to the netns_helper.sh. The patch which adds what we
need ('IFLA_NET_NS_PID') to the kernel is:
http://lists.linuxfoundation.org/pipermail/containers/2007-September/007113.html

Matus


----- Original Message -----
From: "Cyril Hrubis" <ch...@su...>
To: "Guangwen Feng" <fen...@cn...>
Cc: ltp...@li..., jja...@re...
Sent: Tuesday, September 1, 2015 1:42:30 PM
Subject: Re: [LTP] LTP release

Hi!
> > I've did my release testing and pushed a few fixes, mosty minor ones.
> > 
> > Now the only problem that that blocks the release on my side are
> > failures of the newly rewritten netns testcases that were discussed
> > last week.
> > 
> 
> In RHEL5.11GA, ns_ifmove.c gives compilation error since 'IFLA_NET_NS_PID'
> undeclared in the kernel.

Looks like we need a configure check for the enum constant and way to
propagate TCONF from the ns_ifmove to the netns_helper.sh.

I will look into this.

-- 
Cyril Hrubis
ch...@su...

------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
Ltp...@li...
https://lists.sourceforge.net/lists/listinfo/ltp-list

Re: [LTP] LTP release

[Cyril H. <ch...@su...>]

Hi!
> as you mentioned, the problem in ns_exec was setns'ing the "user" namespace,
> I think it would be sufficient to just comment/remove this line:
> rf |= open_ns_fd(argv[1], "user");
> from ns_exec.c or even all calls to open_ns_fd() function except for the network
> namespace as it is used in netns testcases. I will rewrite ns_create and ns_exec
> after the release.

I was trying to be more defensive and remove everything that is not
needed, but I can live with this solution as well.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Cyril H. <ch...@su...>]

Hi!
> > I've did my release testing and pushed a few fixes, mosty minor ones.
> > 
> > Now the only problem that that blocks the release on my side are
> > failures of the newly rewritten netns testcases that were discussed
> > last week.
> > 
> 
> In RHEL5.11GA, ns_ifmove.c gives compilation error since 'IFLA_NET_NS_PID'
> undeclared in the kernel.

Looks like we need a configure check for the enum constant and way to
propagate TCONF from the ns_ifmove to the netns_helper.sh.

I will look into this.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Matus M. <mma...@re...>]

Hi,

as you mentioned, the problem in ns_exec was setns'ing the "user" namespace,
I think it would be sufficient to just comment/remove this line:
rf |= open_ns_fd(argv[1], "user");
from ns_exec.c or even all calls to open_ns_fd() function except for the network
namespace as it is used in netns testcases. I will rewrite ns_create and ns_exec
after the release.

Matus Marhefka

----- Original Message -----
From: "Cyril Hrubis" <ch...@su...>
To: ltp...@li...
Cc: "Jiri Jaburek" <jja...@re...>, "Matus Marhefka" <mma...@re...>
Sent: Monday, August 31, 2015 1:23:20 PM
Subject: Re: [LTP] LTP release

Hi!
I've did my release testing and pushed a few fixes, mosty minor ones.

Now the only problem that that blocks the release on my side are
failures of the newly rewritten netns testcases that were discussed
last week.

Jiri: What about changing the ns_exec to join only net namespace for
      now and rewriting it after the release?

Everybody: If you have issues with the latest git please report them.
           And if you haven't tested it yet please do so ASAP.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Guangwen F. <fen...@cn...>]

On 2015/08/31 19:23, Cyril Hrubis wrote:
> Hi!
> I've did my release testing and pushed a few fixes, mosty minor ones.
> 
> Now the only problem that that blocks the release on my side are
> failures of the newly rewritten netns testcases that were discussed
> last week.
> 

In RHEL5.11GA, ns_ifmove.c gives compilation error since 'IFLA_NET_NS_PID'
undeclared in the kernel.

> Jiri: What about changing the ns_exec to join only net namespace for
>       now and rewriting it after the release?
> 
> Everybody: If you have issues with the latest git please report them.
>            And if you haven't tested it yet please do so ASAP.
> 

Re: [LTP] [PATCH v2] ltp/pounder: Don't let the testing log output to /dev/tty

[Li W. <li...@re...>]

ping, is there possible to push this before the newest LTP released?


On Fri, Aug 21, 2015 at 6:13 PM, Li Wang <li...@re...> wrote:

> v1-->v2:
>         - FILE *tty_fp --> FILE *out
>         - Using out = stdout to replace the fdopen() operate;
>
> Signed-off-by: Li Wang <li...@re...>
> Signed-off-by: Cyril Hrubis <ch...@su...>
> ---
>  tools/pounder21/Install              |  2 +-
>  tools/pounder21/fancy_timed_loop.c   | 29 ++++++++++++-----------------
>  tools/pounder21/infinite_loop.c      | 25 ++++++++++---------------
>  tools/pounder21/pounder              | 14 +++++++-------
>  tools/pounder21/test_scripts/memtest |  2 +-
>  tools/pounder21/timed_loop.c         | 27 +++++++++++----------------
>  6 files changed, 42 insertions(+), 57 deletions(-)
>
> diff --git a/tools/pounder21/Install b/tools/pounder21/Install
> index bf13528..6fb3a8f 100755
> --- a/tools/pounder21/Install
> +++ b/tools/pounder21/Install
> @@ -65,7 +65,7 @@ done
>  # Set up optdir
>  mkdir -p "$POUNDER_OPTDIR"
>  if [ ! -d "$POUNDER_OPTDIR" ]; then
> -       echo "Could not create $POUNDER_OPTDIR; aborting." > /dev/tty
> +       echo "Could not create $POUNDER_OPTDIR; aborting."
>         exit 1
>  fi
>
> diff --git a/tools/pounder21/fancy_timed_loop.c
> b/tools/pounder21/fancy_timed_loop.c
> index cf8580e..e2ca1dc 100644
> --- a/tools/pounder21/fancy_timed_loop.c
> +++ b/tools/pounder21/fancy_timed_loop.c
> @@ -32,13 +32,13 @@
>  static int res = 0;
>  static char *progname;
>  static pid_t test_pgrp;
> -static FILE *tty_fp;
> +static FILE *out;
>
>  static int the_signal = SIGTERM;
>
>  static void int_func(int signum)
>  {
> -       pounder_fprintf(tty_fp,
> +       pounder_fprintf(out,
>                         "%s: Killed by interrupt.  Last exit code = %d.\n",
>                         progname, res);
>         kill(-test_pgrp, the_signal);
> @@ -47,7 +47,7 @@ static void int_func(int signum)
>
>  static void alarm_func(int signum)
>  {
> -       pounder_fprintf(tty_fp, "%s: Killed by timer.  Last exit code =
> %d.\n",
> +       pounder_fprintf(out, "%s: Killed by timer.  Last exit code =
> %d.\n",
>                         progname, res);
>         kill(-test_pgrp, the_signal);
>         exit(res);
> @@ -90,14 +90,9 @@ int main(int argc, char *argv[])
>                 }
>         }
>
> -       tty_fp = fdopen(3, "w+");
> -       if (tty_fp == NULL) {
> -               tty_fp = fopen("/dev/tty", "w+");
> -               if (tty_fp == NULL) {
> -                       perror("stdout");
> -                       exit(2);
> -               }
> -       }
> +       out = fdopen(3, "w+");
> +       if (out == NULL)
> +               out = stdout;
>
>         if (use_max_failures) {
>                 progname = rindex(argv[7], '/');
> @@ -142,11 +137,11 @@ int main(int argc, char *argv[])
>                 gid = atoi(argv[3]);
>         }
>
> -       pounder_fprintf(tty_fp, "%s: uid = %d, gid = %d, sig = %d\n",
> +       pounder_fprintf(out, "%s: uid = %d, gid = %d, sig = %d\n",
>                         progname, uid, gid, the_signal);
>
>         while (1) {
> -               pounder_fprintf(tty_fp, "%s: %s loop #%d.\n", progname,
> +               pounder_fprintf(out, "%s: %s loop #%d.\n", progname,
>                                 start_msg, revs++);
>                 pid = fork();
>                 if (pid == 0) {
> @@ -196,22 +191,22 @@ int main(int argc, char *argv[])
>                 }
>                 // interrogate it
>                 if (WIFSIGNALED(stat)) {
> -                       pounder_fprintf(tty_fp, "%s: %s on signal %d.\n",
> +                       pounder_fprintf(out, "%s: %s on signal %d.\n",
>                                         progname, fail_msg,
> WTERMSIG(stat));
>                         res = 255;
>                 } else {
>                         res = WEXITSTATUS(stat);
>                         if (res == 0) {
> -                               pounder_fprintf(tty_fp, "%s: %s.\n",
> progname,
> +                               pounder_fprintf(out, "%s: %s.\n", progname,
>                                                 pass_msg);
>                         } else if (res < 0 || res == 255) {
> -                               pounder_fprintf(tty_fp,
> +                               pounder_fprintf(out,
>                                                 "%s: %s with code %d.\n",
>                                                 progname, abort_msg, res);
>                                 exit(-1);
>                                 // FIXME: add test to blacklist
>                         } else {
> -                               pounder_fprintf(tty_fp,
> +                               pounder_fprintf(out,
>                                                 "%s: %s with code %d.\n",
>                                                 progname, fail_msg, res);
>                                 if (max_failures > 0) {
> diff --git a/tools/pounder21/infinite_loop.c
> b/tools/pounder21/infinite_loop.c
> index d78b442..3a9d68c 100644
> --- a/tools/pounder21/infinite_loop.c
> +++ b/tools/pounder21/infinite_loop.c
> @@ -32,11 +32,11 @@
>  static int res = 0;
>  static char *progname;
>  static pid_t test_pgrp;
> -static FILE *tty_fp;
> +static FILE *out;
>
>  static void int_func(int signum)
>  {
> -       pounder_fprintf(tty_fp,
> +       pounder_fprintf(out,
>                         "%s: Killed by interrupt.  Last exit code = %d.\n",
>                         progname, res);
>         kill(-test_pgrp, SIGTERM);
> @@ -75,14 +75,9 @@ int main(int argc, char *argv[])
>                 }
>         }
>
> -       tty_fp = fdopen(3, "w+");
> -       if (tty_fp == NULL) {
> -               tty_fp = fopen("/dev/tty", "w+");
> -               if (tty_fp == NULL) {
> -                       perror("stdout");
> -                       exit(2);
> -               }
> -       }
> +       out = fdopen(3, "w+");
> +       if (out == NULL)
> +               out = stdout;
>
>         if (use_max_failures) {
>                 progname = rindex(argv[3], '/');
> @@ -110,7 +105,7 @@ int main(int argc, char *argv[])
>          * loop test and descendants easily */
>
>         while (1) {
> -               pounder_fprintf(tty_fp, "%s: %s loop #%d.\n", progname,
> +               pounder_fprintf(out, "%s: %s loop #%d.\n", progname,
>                                 start_msg, revs++);
>                 pid = fork();
>                 if (pid == 0) {
> @@ -149,22 +144,22 @@ int main(int argc, char *argv[])
>                 }
>                 // interrogate it
>                 if (WIFSIGNALED(stat)) {
> -                       pounder_fprintf(tty_fp, "%s: %s on signal %d.\n",
> +                       pounder_fprintf(out, "%s: %s on signal %d.\n",
>                                         progname, fail_msg,
> WTERMSIG(stat));
>                         res = 255;
>                 } else {
>                         res = WEXITSTATUS(stat);
>                         if (res == 0) {
> -                               pounder_fprintf(tty_fp, "%s: %s.\n",
> progname,
> +                               pounder_fprintf(out, "%s: %s.\n", progname,
>                                                 pass_msg);
>                         } else if (res < 0 || res == 255) {
> -                               pounder_fprintf(tty_fp,
> +                               pounder_fprintf(out,
>                                                 "%s: %s with code %d.\n",
>                                                 progname, abort_msg, res);
>                                 exit(-1);
>                                 // FIXME: add test to blacklist
>                         } else {
> -                               pounder_fprintf(tty_fp,
> +                               pounder_fprintf(out,
>                                                 "%s: %s with code %d.\n",
>                                                 progname, fail_msg, res);
>                                 if (max_failures > 0) {
> diff --git a/tools/pounder21/pounder b/tools/pounder21/pounder
> index c3defcf..caf60d3 100755
> --- a/tools/pounder21/pounder
> +++ b/tools/pounder21/pounder
> @@ -201,8 +201,8 @@ shift `expr $OPTIND - 1`
>
>  # Are we already running?
>  if [ -f "$POUNDER_PIDFILE" ]; then
> -       echo "File $POUNDER_PIDFILE exists; pounder may already be
> running." > /dev/tty
> -       echo "Either run 'pounder -k' to stop all tests, or remove it." >
> /dev/tty
> +       echo "File $POUNDER_PIDFILE exists; pounder may already be
> running."
> +       echo "Either run 'pounder -k' to stop all tests, or remove it."
>         exit 1
>  fi
>
> @@ -225,21 +225,21 @@ fi
>  # Set up log directory
>  mkdir -p "$POUNDER_LOGDIR" 2> /dev/null
>  if [ ! -d "$POUNDER_LOGDIR" ]; then
> -       echo "Could not create $POUNDER_LOGDIR; aborting." > /dev/tty
> +       echo "Could not create $POUNDER_LOGDIR; aborting."
>         exit 1
>  fi
>
>  # Set up dir for optional components
>  mkdir -p "$POUNDER_OPTDIR"
>  if [ ! -d "$POUNDER_OPTDIR" ]; then
> -       echo "Could not create $POUNDER_OPTDIR; aborting." > /dev/tty
> +       echo "Could not create $POUNDER_OPTDIR; aborting."
>         exit 1
>  fi
>
>  # Set up tmpdir
>  mkdir -p "$POUNDER_TMPDIR"
>  if [ ! -d "$POUNDER_TMPDIR" ]; then
> -       echo "Could not create $POUNDER_TMPDIR; aborting." > /dev/tty
> +       echo "Could not create $POUNDER_TMPDIR; aborting."
>         exit 1
>  fi
>
> @@ -259,8 +259,8 @@ fi
>  export > "$POUNDER_LOGDIR/environment"
>
>  echo "Starting $POUNDER_VERSION"
> -echo "STARTING TESTS." > /dev/tty
> -echo "To kill all tests, run 'pounder -k' or press ^C."> /dev/tty
> +echo "STARTING TESTS."
> +echo "To kill all tests, run 'pounder -k' or press ^C."
>
>  # Handle the duration thing...
>  function kill_after {
> diff --git a/tools/pounder21/test_scripts/memtest
> b/tools/pounder21/test_scripts/memtest
> index c040147..e212bde 100755
> --- a/tools/pounder21/test_scripts/memtest
> +++ b/tools/pounder21/test_scripts/memtest
> @@ -22,7 +22,7 @@
>
>  # Can we find the script?
>  if [ ! -f "$POUNDER_OPTDIR/memtest.sh" ]; then
> -       echo "Can't find memtest.sh; did you run Install?" > /dev/tty
> +       echo "Can't find memtest.sh; did you run Install?"
>         exit -1
>  fi
>
> diff --git a/tools/pounder21/timed_loop.c b/tools/pounder21/timed_loop.c
> index 966c08a..a7413c3 100644
> --- a/tools/pounder21/timed_loop.c
> +++ b/tools/pounder21/timed_loop.c
> @@ -32,11 +32,11 @@
>  static int res = 0;
>  static char *progname;
>  static pid_t test_pgrp;
> -static FILE *tty_fp;
> +static FILE *out;
>
>  static void int_func(int signum)
>  {
> -       pounder_fprintf(tty_fp,
> +       pounder_fprintf(out,
>                         "%s: Killed by interrupt.  Last exit code = %d.\n",
>                         progname, res);
>         kill(-test_pgrp, SIGTERM);
> @@ -45,7 +45,7 @@ static void int_func(int signum)
>
>  static void alarm_func(int signum)
>  {
> -       pounder_fprintf(tty_fp, "%s: Killed by timer.  Last exit code =
> %d.\n",
> +       pounder_fprintf(out, "%s: Killed by timer.  Last exit code =
> %d.\n",
>                         progname, res);
>         kill(-test_pgrp, SIGTERM);
>         exit(res);
> @@ -92,14 +92,9 @@ int main(int argc, char *argv[])
>                 }
>         }
>
> -       tty_fp = fdopen(3, "w+");
> -       if (tty_fp == NULL) {
> -               tty_fp = fopen("/dev/tty", "w+");
> -               if (tty_fp == NULL) {
> -                       perror("stdout");
> -                       exit(2);
> -               }
> -       }
> +       out = fdopen(3, "w+");
> +       if (out == NULL)
> +               out = stdout;
>
>         if (use_max_failures) {
>                 progname = rindex(argv[4], '/');
> @@ -136,7 +131,7 @@ int main(int argc, char *argv[])
>         alarm(secs);
>
>         while (1) {
> -               pounder_fprintf(tty_fp, "%s: %s loop #%d.\n", progname,
> +               pounder_fprintf(out, "%s: %s loop #%d.\n", progname,
>                                 start_msg, revs++);
>                 pid = fork();
>                 if (pid == 0) {
> @@ -175,22 +170,22 @@ int main(int argc, char *argv[])
>                 }
>                 // interrogate it
>                 if (WIFSIGNALED(stat)) {
> -                       pounder_fprintf(tty_fp, "%s: %s on signal %d.\n",
> +                       pounder_fprintf(out, "%s: %s on signal %d.\n",
>                                         progname, fail_msg,
> WTERMSIG(stat));
>                         res = 255;
>                 } else {
>                         res = WEXITSTATUS(stat);
>                         if (res == 0) {
> -                               pounder_fprintf(tty_fp, "%s: %s.\n",
> progname,
> +                               pounder_fprintf(out, "%s: %s.\n", progname,
>                                                 pass_msg);
>                         } else if (res < 0 || res == 255) {
> -                               pounder_fprintf(tty_fp,
> +                               pounder_fprintf(out,
>                                                 "CHECK %s: %s with code
> %d.\n",
>                                                 progname, abort_msg, res);
>                                 exit(-1);
>                                 // FIXME: add test to blacklist
>                         } else {
> -                               pounder_fprintf(tty_fp,
> +                               pounder_fprintf(out,
>                                                 "%s: %s with code %d.\n",
>                                                 progname, fail_msg, res);
>                                 if (max_failures > 0) {
> --
> 1.8.3.1
>
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Ltp-list mailing list
> Ltp...@li...
> https://lists.sourceforge.net/lists/listinfo/ltp-list
>



-- 
Regards,
Li Wang
Email: li...@re...

Re: [LTP] Test Anything Protocal output

[<dw...@fi...>]

On Mon, Aug 31, 2015 at 12:22:44PM +0200, Cyril Hrubis wrote:
> Hi!
> > I read your January 2015 article where someone asked if TAP test output would be supported. It sounded like
> > you were going to look into it. Did you make any progress?
> 
> Unfortunately not yet.
> 
> > The reason I asked if because I was wanting the same feature and I was going to review the LTP code to see
> > how hard it would be to add it, but I wouldn't want to start if someone else has already done it. I did
> > see anything in the git history tho.
> 
> Well, have a look at lib/tst_res.c, that is the place where test results
> are stored/printed. Unfortunately this code hasn't been cleaned up yet.

So this typical line in LTP,

mount02     1  TPASS  :  mount() failed expectedly: TEST_ERRNO=ENODEV(19): No such device

would have to be transformed into,

ok 1 mount02 mount() failed expectedly: TEST_ERRNO=ENODEV(19): No such device
 
After the test number "1" in this case, looks like it's free form. It's "ok" for passed
tests, and "not ok" for failed ones, and you can skip too. 

https://testanything.org/tap-specification.html

It looks like it could be changes in lib/tst_res.c line 404 in function tst_print().
There's other components to TAP but the change above it the biggest I think.

> I've started to work on this just recently. Just now I'm experimenting
> with rewriting it from scratch keeping most of the API unchanged (since
> we cannot change easily the API for > ~1000 testcases).
> 
> Then there is another problem, so far LTP testcases are executed using
> the ltp-pan test driver, which executes testcases, catches segfaults,
> etc. We would have to consider how that would interact with the output
> generated from testcases.
> 
> One option would be to get rid of most of the pan functionality by
> moving the test driver code into the test library, so that each testcase
> would fork first, then run the real test in it's own virtual terminal,
> etc.

Well, if your re-writing it, than TAP could be added after that. Is this re-write
a long term thing, or something your planning to complete in the next couple months?

Daniel

Re: [LTP] Linux 3.10

[Cyril H. <ch...@su...>]

Hi!
> hello dear:
>     Does ltp support Linux 3.10?

Yes.

> I want write a virtual usb device and I wan to reference the tusb
> module. But when I compile the module on ubuntu 12.04,I get many
> errors.

Looks like nobody has touched the test code for some quite time (there
was one minor change but apart from that nothing), it's not build by
default and basically abandoned.

The problem is that in-kernel API is not stable. So if the kernel part
of the test fails to compile you have to figure out what part of the API
has changed and fix the code accordingly. Patches for the testcases are
of course welcomed.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] LTP release

[Cyril H. <ch...@su...>]

Hi!
I've did my release testing and pushed a few fixes, mosty minor ones.

Now the only problem that that blocks the release on my side are
failures of the newly rewritten netns testcases that were discussed
last week.

Jiri: What about changing the ns_exec to join only net namespace for
      now and rewriting it after the release?

Everybody: If you have issues with the latest git please report them.
           And if you haven't tested it yet please do so ASAP.

-- 
Cyril Hrubis
ch...@su...

Re: [LTP] Test Anything Protocal output

[Cyril H. <ch...@su...>]

Hi!
> I read your January 2015 article where someone asked if TAP test output would be supported. It sounded like
> you were going to look into it. Did you make any progress?

Unfortunately not yet.

> The reason I asked if because I was wanting the same feature and I was going to review the LTP code to see
> how hard it would be to add it, but I wouldn't want to start if someone else has already done it. I did
> see anything in the git history tho.

Well, have a look at lib/tst_res.c, that is the place where test results
are stored/printed. Unfortunately this code hasn't been cleaned up yet.

I've started to work on this just recently. Just now I'm experimenting
with rewriting it from scratch keeping most of the API unchanged (since
we cannot change easily the API for > ~1000 testcases).

Then there is another problem, so far LTP testcases are executed using
the ltp-pan test driver, which executes testcases, catches segfaults,
etc. We would have to consider how that would interact with the output
generated from testcases.

One option would be to get rid of most of the pan functionality by
moving the test driver code into the test library, so that each testcase
would fork first, then run the real test in it's own virtual terminal,
etc.

-- 
Cyril Hrubis
ch...@su...

[LTP] [PATCH v6 7/7] ipv6_lib: Add it into default

[Zeng L. <zen...@cn...>]

Signed-off-by: Zeng Linggang <zen...@cn...>
Signed-off-by: Alexey Kodanev <ale...@or...>
Signed-off-by: Cyril Hrubis <ch...@su...>
---
 scenario_groups/default | 1 +
 1 file changed, 1 insertion(+)

diff --git a/scenario_groups/default b/scenario_groups/default
index 3a55702..60b1b27 100644
--- a/scenario_groups/default
+++ b/scenario_groups/default
@@ -32,3 +32,4 @@ dma_thread_diotest
 cpuacct
 can
 cpuhotplug
+ipv6_lib
-- 
1.9.3