Hi,
* Subrata Modak <[EMAIL PROTECTED]> [2008-04-11 15:40:54]:
> On Thu, 2008-04-10 at 20:45 -0700, Darren Hart wrote:
> > On Thu, 2008-04-10 at 19:49 -0700, Andrew Morton wrote:
> > > On Thu, 10 Apr 2008 19:46:00 -0700 Andrew Morton <[EMAIL PROTECTED]>
> > > wrote:
> > >
> > > > librttest.c: In function `init_pi_mutex':
> > > > librttest.c:463: warning: implicit declaration of function
> > > > `pthread_mutexattr_setprotocol'
> > > > librttest.c:463: error: `PTHREAD_PRIO_INHERIT' undeclared (first use in
> > > > this function)
> > > > librttest.c:463: error: (Each undeclared identifier is reported only
> > > > once
> > > > librttest.c:463: error: for each function it appears in.)
> > > > librttest.c:466: warning: implicit declaration of function
> > > > `pthread_mutexattr_ge
> > >
> >
> > OK, so older glibc's (like that in FC3) won't have the necessary bits
> > for PI mutexes. Subrata, from an LTP point of view, how should this be
> > detected? I presume the "right thing" in this case is to detect the
> > fact that PTHREAD_PRIO_INHERIT doesn't exist (or perhaps just check the
> > glibc version) and not attempt to build the 'realtime' testsuite. The
> > other question is if realtime should be enabled by default - I don't
> > recall if it is or not.
>
> No. I do not think that it should be enabled by default, not only for
> realtime test cases, but also for other test cases. Just like, we had
> some problem on RT test cases compilation on some architectures and
> Chirag´s patch fixed it by avoiding compilation/run on those
> architectures. We have similar set up for features, where they check the
> kernel version on which they are built to re-assure that they are
> supported there.
> So, for old gcc version(s) also, when you feel that the tests cannot run
> below that, do a check and abort build as well as run.
>
> So, from now onwards we have the checking for gcc version as well, apart
> from the Architecture checking, Kernel Version Checking and FS type
> checking.
>
> I can ask people to do the new checking as well. But the problem is how
> all those 3000+ odd test cases can be patched for this ;-)
>
> Regards--
> Subrata
>
Before the make step, the realtime tests have a configure step that
checks exaclty for this .
It checks whether certain functionality necessary to run the test
exists or not.
Subrata,
If this is feasible, a configure step can be added to the entire ltp
testsuites
for all tests that require some special glibc/kernel support.
For now, we can probably ensure that configure is run before
make.
Thoughts?
--
Cheers,
Chirag Jog
-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
Don't miss this year's exciting event. There's still time to save $100.
Use priority code J8TL2D2.
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
