[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Nmh-workers] Volunteer Capacity.

From: Ralph Corderoy
Subject: Re: [Nmh-workers] Volunteer Capacity.
Date: Sat, 22 Jul 2017 13:23:48 +0100

Hi Bob,

> Out of curiosity, would it help to have "micro-volunteers" to update
> the testing?

I'd have thought so, though David is, I think, the main person behind
the test suite.

> namely having (sufficient) test coverage,
> If enough of us do that, maybe we could significantly improve
> coverage?

Yes, I think so, even if it's just a few plugging away steadily.

I played about with gcov(1) coverage testing on nmh a while back and
should do so again and write up how to do it this time.  Because I was
just trying to get more lines executed by the tests, I added
command-invocation tests that didn't check standard output or error,
just the exit status.  An example is

Knowing the tests cover more lines makes running them under valgrind(1)
more useful.

The idea was to chase the low-hanging fruit by looking at gcov's
annotated source and spot the easy bits to get coverage on.  It may seem
that it's only an if statement and an error message, but they're often
calling routines and so have a knock-on effect for coverage.  I thought
once that "easy" outer veneer was exercised we could look again at
gcov's results and pick a chunk of "library" routines on the next level
down and work backwards to see what caller we can run.  At this point,
we'll probably be more interested in the command output matching
expectations, i.e. did the library routines not just run but work
correctly, so the added tests would be more like the existing ones.

> Of course, one of the main developers still needs to "approve" any
> commits, so that doesn't completely eliminate work on testing.

I think a few canned git(1) commands, for those that don't know it,
would get you a long way, and they can prepare a patch for email that
those with commit rights can easily apply whilst still attributing the
original author.  Once they've got the hang of it and want to keep going
then the tester can sign up on nongnu.org themselves.

Cheers, Ralph.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]