Some thoughts about coding plan

[Zhibin Li]

Hi guys,

I've been doing research about testing stuff and trying to discover more
about strace myself. As you guys mentioned on the website, branches of
parsers are way far from being covered so I think I'd better start with
coding for branches coverage. More specifically, I've got some basic
knowledge about the metric of branch coverage and besides, I noticed that
there is an option "--enable-code-coverage" which is based on gcov and lcov
and shows line/function/branch coverage in number/percent.
I think I could use this coverage report as my guidance and strart to code
for those branches which are not yet covered. For example, fcntl parser has
a lot of options and flags in the switch...case segment that need coverage
and of course other parsers may be not like it, e.g. more if else
condition. To become more familiar with these syscalls I think manual pages
will help a lot. I can refer to the categories and usages in the manual
considering there are part of syscalls I don't know about.
Concerns: Dmitry said the kernel the test suite is running is relatively
old but the branch coverage report generated by gcov is based on the kernel
on my own desktop/laptop so there must be some differences. Should I just
compile and install the older kernel and use some other platforms to figure
it out? I don't know much about the discrepancies among platforms as well
as versions of kernels (which I'm still learning). I have to say I couldn't
come out with any excellent ideas to significantly enhance the test suite
for now so what do you think if I start with the more basic stuff mentioned
above? You must've noticed by reviewing the patches I sent that I lack
experience so if you guys have any extra suggestions, I'd be appreciated!

Thanks,
Zhibin Li
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20180509/976523ae/attachment.html>