[LTP] [PATCH] open_posix_testsuite/pthread_attr_init

[Richard Palethorpe]

Hello Lucas,

Thanks for contributing, please see comments below:

Lucas Boehm writes:

> new test of checking default attr of phtread_attr_init()
>
> Linux specific implementation
>
> Signed-off-by: Lucas Boehm <lucas@opentech.at>
> ---
>  .../conformance/interfaces/pthread_attr_init/1-2.c | 126 +++++++++++++++++++++
>  1 file changed, 126 insertions(+)
>  create mode 100644 testcases/open_posix_testsuite/conformance/interfaces/pthread_attr_init/1-2.c
>
> diff --git a/testcases/open_posix_testsuite/conformance/interfaces/pthread_attr_init/1-2.c b/testcases/open_posix_testsuite/conformance/interfaces/pthread_attr_init/1-2.c
> new file mode 100644
> index 000000000..ec740e78c
> --- /dev/null
> +++ b/testcases/open_posix_testsuite/conformance/interfaces/pthread_attr_init/1-2.c
> @@ -0,0 +1,126 @@
> +/*
> + * Copyright (c) 2017, Opentech EDV Research GmbH. All rights reserved.
> + * Created by:  lucas AT opentech DOT at
> + * This file is licensed under the GPL license.  For the full content
> + * of this license, see the COPYING file at the top level of this
> + * source tree.
> +
> + * Test that pthread_attr_init()
> + *  shall initialize a thread attributes object 'attr' with the default value
> + *  for all the individual attributes used by a given implementation.
> +
> + *  Detache state:
> + *  The default setting of the detach state attribute in a newly initialized
> + *  thread attributes object is PTHREAD_CREATE_JOINABLE.
> + *  (man pthread_attr_getdetachstate)
> + *
> + *  Scope:
> + *  Linux supports PTHREAD_SCOPE_SYSTEM, but not PTHREAD_SCOPE_PROCESS.
> + *  (man pthread_attr_setscope)

I don't see a default specification for this in the POSIX manual
pages. So either value should be acceptable regardless of whether Linux
accepts it. Linux specific tests should use the normal LTP library and
do not go in open_posix_testsuite because this should also run on BSD
and other POSIX complaint systems.

> + *
> + *  Inherit Scheduler:
> + *  The default setting of the inherit-scheduler attribute in a newly
> + *  initialized thread attributes object is PTHREAD_INHERIT_SCHED.
> + *  (man phread_attr_getinheritsched)
> + *
> + *  Scheduling Policy:
> + *  SCHED_OTHER is the standard Linux time-sharing scheduler that is
> + *  intended for all threads that do not require the special real-time
> + *  mechanisms.
> + *  (man sched)

Again I don't see this in the POSIX spec?

> + *
> + * Scheduling Priority:
> + * For threads scheduled under one of the normal scheduling policies
> + * (SCHED_OTHER, SCHED_IDLE, SCHED_BATCH), sched_priority is not used
> + * in scheduling decisions (it must be specified as 0)
> + * (man sched)
> + *
> + * Steps:
> + * 1.  Initialize a pthread_attr_t object using pthread_attr_init()
> + * 2.  Using pthread_attr_getdetachstate(), test that the default value
> + *     of the detachstate is PTHREAD_CREATE_JOINABLE (the default).
> + * 3.  Using pthread_attr_getscope(), test that the default value
> + *     of the scope is PTHREAD_SCOPE_SYSTEM (the default).
> + * 4.  Using pthread_attr_getinheritsched(), test that the default value
> + *     of the Inherit scheduler is PTHREAD_INHERIT_SCHED (the default).
> + * 5.  Using pthread_attr_getschedpolicy(), test that the default value
> + *     of the sched_policy is SCHED_OTHER (the default).
> + * 6.  Using pthread_attr_getschedparam(), test that the default value
> + *     of the sched_priority is 0 (the default).
> + */
> +
> +#include <pthread.h>
> +#include <stdio.h>
> +#include "posixtest.h"
> +
> +
> +int main(void)
> +{
> +	pthread_attr_t new_attr;
> +	int detach_state;
> +	int scope;
> +	int inherit_sched;
> +	int sched_policy;
> +	struct sched_param sched_priority;
> +
> +
> +	/* Initialize attribute */
> +	if (pthread_attr_init(&new_attr) != 0) {
> +		perror("Cannot initialize attribute object\n");
> +		return PTS_UNRESOLVED;
> +	}
> +
> +	/* The test passes if the attribute object has a detachstate of
> +	 * PTHREAD_CREATE_JOINABLE, which is the default value for this
> +	 * attribute. */

No need to include inline comments in the final patch unless you are
doing something really unusual. However you can print a nice descriptive
message for the end user if the test fails.

> +	if (pthread_attr_getdetachstate(&new_attr, &detach_state) != 0) {
> +		perror("Error obtaining the detachstate of the attribute\n");
> +		return PTS_UNRESOLVED;
> +	}
> +
> +	/* The test passes if the attribute object has a scope of
> +	 * PTHREAD_SCOPE_SYSTEM, which is the default value for this
> +	 * attribute. */
> +	if (pthread_attr_getscope(&new_attr, &scope) != 0) {
> +		perror("Error obtaining the scope of the attribute\n");
> +		return PTS_UNRESOLVED;
> +	}
> +
> +	/* The test passes if the attribute object has a inherit sched of
> +	 * PTHREAD_INHERIT_SCHED, which is the default value for this
> +	 * attribute. */
> +	if (pthread_attr_getinheritsched(&new_attr, &inherit_sched) != 0) {
> +		perror("Error obtaining the inherit scheduling of the attribute\n");
> +		return PTS_UNRESOLVED;
> +	}
> +
> +	/* The test passes if the attribute object has a sched policy of
> +	 * SCHED_OTHER, which is the default value for this
> +	 * attribute. */
> +	if (pthread_attr_getschedpolicy(&new_attr, &sched_policy) != 0) {
> +		perror("Error obtaining the sched policy of the attribute\n");
> +		return PTS_UNRESOLVED;
> +	}
> +
> +	/* The test passes if the attribute object has a inherit sched of
> +	 * pthread_attr_getschedparam, which is the default value for this
> +	 * attribute. */
> +	if (pthread_attr_getschedparam(&new_attr, &sched_priority) != 0) {
> +		perror("Error obtaining the sched policy of the attribute\n");
> +		return PTS_UNRESOLVED;
> +	}
> +
> +	if (
> +		detach_state == PTHREAD_CREATE_JOINABLE && \
> +		scope == PTHREAD_SCOPE_SYSTEM && \
> +		inherit_sched == PTHREAD_INHERIT_SCHED && \
> +		sched_policy == SCHED_OTHER && \
> +		sched_priority.sched_priority == 0) {
> +		printf("Test PASSED\n");
> +		return PTS_PASS;
> +	}

The end user won't know which condition failed without using
ltrace/strace which is inconvenient.

> +	else {
> +		printf("Test FAILED\n");
> +		return PTS_FAIL;
> +	}
> +}
> --
> 2.11.0

--
Thank you,
Richard.