----- Original Message -----
> If these issues are concerning you, I propose we increase the level
> of testing done on the kernel PMDA (we have no pmdalinux tests using
> valgrind, IIRC, which needs to change). DSO mode makes this form of
> memory-check testing much easier too, ironically (valgrind->pminfo).
I've created test 957 which does this local context checking on the
Linux kernel PMDA. Its failing on a couple of sub-trees, if anyone
would like to take a closer look (else, I'll come back to it soon).
> I'd also like to see an increase in the kernel PMDA testing done via
> the non-/proc-prefix approach to exercise the many little parsers it
> has with canned, targeted data access (currently all pmdalinux tests
> use the running kernel as the data source, but it doesn't have to be
> that way).
This is a long-term project of course, any helpers here also would be
very much appreciated. Needs some dedicated individuals to go through
each of the /proc file parsers and begin producing test cases that will
exercise the more unusual situations. Fuzz testing might be handy here
if anyone is game to look into some of the open source fuzzers? Also,
we could begin making all new pmdalinux commits raise the bar on their
level of testing effort (whenever there's a bug, lets aim to produce a
decent set of coverage in the affected area).
cheers.
--
Nathan
|