Re: [PATCH]: NULL pointer check in timer_* functions

[Khem Raj <kraj at mvista dot com>]

Ulrich Drepper wrote:

> Khem Raj wrote:
> \> I tracked it down to a problem inside timer_* functions. Currently we do
>  
>
> > not check for timerid for NULL value inside timer_delete()
> > timer_settime() and timer_gettime() timer_getoverrun() functions.
> > This patch adds the checks to these functions.
> >    
>
> That's all bullshit.  A zero timer ID is from the outside not different
> than anything else.
0 ID is not being passed from outside but from within glibc. testcase is 
passing correct ID that it got with timer_create
it is the glibc wrapper code who is zeroing it out in 
nptl/sysdeps/unix/sysv/linux/x86_64/timer_delete.c line 40

>   If you pass an incorrect value your application
> will crash.  Fix your own bugs instead of making the implementation
> slower just because of your own inabilities.
>
>  
This is a negative testcase from ltp testsuite 
./testcases/open_posix_testsuite/conformance/interfaces/timer_delete/1-2.c
It does not segfault when linked and run with linuxthreads. It segfaults 
when linked with linuxthreads and run
with NPTL.I thought NPTL is backward compatible. Do you think this 
testcase is invalid.? if yes then that will fix my bug.

Thanks for your response

Khem