[Crackerjack-devel] [RFC] [OLS 2008 FALLOUT] Issue # 3 (LTP-Robust)

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

Hi,

With a brief stop over in our fallout discussion, i thought of resuming
discussing the other issues i came up.

As we have proposed in our OLS 2008 paper:
http://ltp.sourceforge.net/documentation/technical_papers/ltp-ols-2008-paper.pdf,

about LTP-robust, which will be a sub-project under LTP. As we are aware
that we have huge limitation in improving code coverage in testing from
the user space alone, as , we do not have sufficient test cases to test
all the features that gets added in the kernel exponentially. The reason
for that is well known, as, we do not have enough volunteers to write
test cases and keep up testing the growing kernel.

Hence, we envisaged the concept of creation of LTP-Robust, which
basically will be a mechanism built in LTP to inject faults in the
kernel. Mechanism of LTP-Robust should be able to do the following:

1) Inject Random faults inside the kernel, when individual test cases
are being run by PAN,
2) PAN should be able to detect that test failures are because of
dynamic faults injected, rather than actual faults in the kernel,
3) Once it is able to decipher this information, it should be able to
restart that test once again with fault-less kernel.

With the above testing methods, we believe we can improve the code
coverage by also touching the error paths inside the kernel, which in
normal circumstance will not be touched by the test code. In order to
achieve the above mechanism, we need to change the PAN (ltp/pan/pan.c)
code.

I would like to know if any of you would be interested to work on this.
Before that let us discuss and sort out things which you think needs
attention before embarking.

Regards--
Subrata