[erlang-questions] common test hiding errors?

[Tim Watson]

On 14 August 2011 16:29, Joel Reymont <joelr1@REDACTED> wrote:
> I'm testing a system composed of various processes.
>
>> Once you've detected that there is a test failure, you should be able to re-run the test manually (say, from command line) to debug it.
>
> Will I get the error from the crashing process in the error log or in the terminal?
>
> It's a pointless exercise if I do not.
>

Hi Joel. Usually I try to find a way of making failures detectable, so
that my tests can do proper assertions about them. Sometimes I'll do
this by noting a pid once a process starts up, and then after some
prodding and poking around, I'll assert that it's still alive by
sneakily asking its supervisor or looking for it with whereis or gproc
or whatever. If I can't make sense of other peoples code and need to
test it I might turn on tracing (in the init_per_testcase/2 function)
and inspect that.

I'm not really sure why you're not seeing crashes in the test log, but
I think the offending code is in either test_server_ctl or the
handler(s) it installs, but again memory fails me as to how this can
be fudged back into working order. Sorry that's not much to go on -
it's kinda late here. :)