Re: Re: Zope tests: 4 OK, 2 Failed

[Tres Seaver]
...
> The other failure, in 'testRunIgnoresParentSignals', is the one which
> Tim despairs of ever fixing :(

And there are two reasons for that.  First, it's part of the zdaemon
tests, and nobody feels they "own" zdaemon, so nobody ever looks at
it.  Second, Tim normally runs on Windows, where zdaemon is a big fat
no-op (Windows doesn't have Unixish signals, so the zdaemon tests are
skipped on Windows); whenever I've tried this test on a Linux box, it
passed.

According to all the signal esoterica I recall from my Unix days,
testRunIgnoresParentSignals _should_ pass on every platform that
supports Unixy signals.  So on boxes where it fails, it's never going
to get fixed unless & until people on such boxes do some work to
figure out _why_ it's failing.  It can't be that hard (it's a very
short & simple test), it seems more that just nobody can be convinced
to bother trying.

I can just note that like many tests involving multiple processes,
it's vulnerable to platform timing accidents.  As a comment in the
test says,

        time.sleep(2) # race condition possible here

If you're running the test on a box so feeble or overburdened that it
takes more than 2 seconds for the line before that to get a new
spawned process running, the test will fail.  For someone on a box
where this test does fail, that's the first thing to determine:  is
the test failing because spawnvp is pig-slow on that box, or for some
other reason (and if so, what?)?