Re: [LTP] testcases for io_cancel()

[Subrata M. <su...@li...>]

I am afraid that this new version introduced more failures than the
previous one. I have resorted to and merged the last stable patch given
by you. I have additionally included run-io_cancel.sh to check for
io_cancel01 binary availability on machines where it has not been built.
If the tests are built it will run the tests, else will show the CONF
message.

Thanks for bearing with me multiple testing cycles. This is now merged.
When you do porting in future you can look into this model as well.

I will not be expecting more patches for the porting project.

Regards--
Subrata

On Mon, 2008-06-23 at 22:24 +0900, Masatake YAMATO wrote:
> Hi,
> 
> > Thank you for resending the patch. I have again tested across various
> > architectures and kernels and had the following observation:
> > 
> > 1) It encountered a build failure on a 2.6.7 kernel instead of presence
> > of header checking. Please see whether you need to check for other
> > parameters as well before building,
> 
> I've intrduce the same technique used in 
> ltp/testcases/realtime/scripts/check_pi.sh. See attachment.
> 
> > 2) When it installed successfully, it also ran successfully on those
> > machines,
> > 3) When it did not install as it did not build, the run will give errors
> > due to absence of io_cancel* binary(s) in testcases/bin. User will see
> > error message like:
> > 
> > initiation_status="pan(): execvp of 'io_cancel01' (tag io_cancel01)
> > failed.  errno:2  No such file or directory"
> > 
> > Instead of that we should have some meaningful error message (the same i
> > proposed to Li Zefan and Vijay Kumar) like:
> > 
> > io_cancel01 0 CONF : System doesn't support execution of the test
> > 
> > For this to happen, the runtest/syscalls should have the following code
> > instead:
> > 
> > io_cancel01 if [ -f io_cancel01 ]; then io_cancel; else echo
> > "io_cancel01 0 CONF : System doesn't support execution of the test";
> > exit 0; fi
> 
> 
> Instead of editing runtest/syscalls, I've used #ifdef/#endif in 
> io_cancel01.c. See the attachment.
> 
> Signed-off-by: Masatake YAMATO <ya...@re...>
>