[Xenomai-core] [PATCH] libnative: Avoid double-evaluation of rt_task_self()

[Xenomai-core] [PATCH] libnative: Avoid double-evaluation of rt_task_self()

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 903 bytes --]

[ can be pulled from git.xenomai.org/xenomai-jki.git queues/assorted ]

Keep the result of rt_task_self() in a local variable to avoid the
second invocation.

Signed-off-by: Jan Kiszka <jan.kiszka@domain.hid>
---

 src/skins/native/task.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/src/skins/native/task.c b/src/skins/native/task.c
index 905d366..149717b 100644
--- a/src/skins/native/task.c
+++ b/src/skins/native/task.c
@@ -247,10 +247,11 @@ int rt_task_resume(RT_TASK *task)
 
 int rt_task_delete(RT_TASK *task)
 {
+	RT_TASK *self;
 	int err;
 
 	if (task == NULL ||
-	    (rt_task_self() && task->opaque == rt_task_self()->opaque)) {
+	    ((self = rt_task_self()) && self->opaque == task->opaque)) {
 		/* Silently migrate to avoid raising SIGXCPU. */
 		XENOMAI_SYSCALL1(__xn_sys_migrate, XENOMAI_LINUX_DOMAIN);
 		pthread_exit(NULL);


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]

Re: [Xenomai-core] [PATCH] libnative: Avoid double-evaluation of rt_task_self()

[Gilles Chanteperdrix]

Jan Kiszka wrote:
> [ can be pulled from git.xenomai.org/xenomai-jki.git queues/assorted ]
> 
> Keep the result of rt_task_self() in a local variable to avoid the
> second invocation.

Maybe we could create a pure/const variant of rt_task_self() for use in
task.c only which would avoid the double evaluatino ?

-- 
					    Gilles.

Re: [Xenomai-core] [PATCH] libnative: Avoid double-evaluation of rt_task_self()

[Philippe Gerum]

On Thu, 2009-04-02 at 23:40 +0200, Gilles Chanteperdrix wrote:
> Jan Kiszka wrote:
> > [ can be pulled from git.xenomai.org/xenomai-jki.git queues/assorted ]
> > 
> > Keep the result of rt_task_self() in a local variable to avoid the
> > second invocation.
> 
> Maybe we could create a pure/const variant of rt_task_self() for use in
> task.c only which would avoid the double evaluatino ?
> 

Can't do const because of pthread_getspecific(), but pure attribute is t
hand, yes.

-- 
Philippe.

Re: [Xenomai-core] [PATCH] libnative: Avoid double-evaluation of rt_task_self()

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 674 bytes --]

Philippe Gerum wrote:
> On Thu, 2009-04-02 at 23:40 +0200, Gilles Chanteperdrix wrote:
>> Jan Kiszka wrote:
>>> [ can be pulled from git.xenomai.org/xenomai-jki.git queues/assorted ]
>>>
>>> Keep the result of rt_task_self() in a local variable to avoid the
>>> second invocation.
>> Maybe we could create a pure/const variant of rt_task_self() for use in
>> task.c only which would avoid the double evaluatino ?
>>
> 
> Can't do const because of pthread_getspecific(), but pure attribute is t
> hand, yes.
> 

At least my gcc 4.3 ignores

RT_TASK *rt_task_self(void) __attribute__ ((pure));

and keeps on inserting two calls in rt_task_delete.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]

Re: [Xenomai-core] [PATCH] libnative: Avoid double-evaluation of rt_task_self()

[Philippe Gerum]

On Sat, 2009-04-04 at 09:37 +0200, Jan Kiszka wrote:
> Philippe Gerum wrote:
> > On Thu, 2009-04-02 at 23:40 +0200, Gilles Chanteperdrix wrote:
> >> Jan Kiszka wrote:
> >>> [ can be pulled from git.xenomai.org/xenomai-jki.git queues/assorted ]
> >>>
> >>> Keep the result of rt_task_self() in a local variable to avoid the
> >>> second invocation.
> >> Maybe we could create a pure/const variant of rt_task_self() for use in
> >> task.c only which would avoid the double evaluatino ?
> >>
> > 
> > Can't do const because of pthread_getspecific(), but pure attribute is t
> > hand, yes.
> > 
> 
> At least my gcc 4.3 ignores
> 
> RT_TASK *rt_task_self(void) __attribute__ ((pure));
> 

We can't do that on rt_task_self() directly, because of
rt_task_shadow().

> and keeps on inserting two calls in rt_task_delete.
> 

What if a pure static inline helper is used internally from task.c?

> Jan
> 
-- 
Philippe.

Re: [Xenomai-core] [PATCH] libnative: Avoid double-evaluation of rt_task_self()

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1464 bytes --]

Philippe Gerum wrote:
> On Sat, 2009-04-04 at 09:37 +0200, Jan Kiszka wrote:
>> Philippe Gerum wrote:
>>> On Thu, 2009-04-02 at 23:40 +0200, Gilles Chanteperdrix wrote:
>>>> Jan Kiszka wrote:
>>>>> [ can be pulled from git.xenomai.org/xenomai-jki.git queues/assorted ]
>>>>>
>>>>> Keep the result of rt_task_self() in a local variable to avoid the
>>>>> second invocation.
>>>> Maybe we could create a pure/const variant of rt_task_self() for use in
>>>> task.c only which would avoid the double evaluatino ?
>>>>
>>> Can't do const because of pthread_getspecific(), but pure attribute is t
>>> hand, yes.
>>>
>> At least my gcc 4.3 ignores
>>
>> RT_TASK *rt_task_self(void) __attribute__ ((pure));
>>
> 
> We can't do that on rt_task_self() directly, because of
> rt_task_shadow().
> 
>> and keeps on inserting two calls in rt_task_delete.
>>
> 
> What if a pure static inline helper is used internally from task.c?
> 

Makes no difference (and defining a new helper for this purpose would be
overkill anyway).

So please pull my original version from
git://git.xenomai.org/xenomai-jki.git for-upstream

BTW, you'll find another trivial fix in that queue:

--- a/src/testsuite/switchtest/switchtest.c
+++ b/src/testsuite/switchtest/switchtest.c
@@ -1094,7 +1094,7 @@ int check_fpu(void)
 		exit(EXIT_FAILURE);
 	}

-	return (int) status;
+	return (long) status;
 }

 int main(int argc, const char *argv[])

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]