[boost] Re: Re: [test] Detecting timeout support

From: Gennadiy Rozental (gennadiy.rozental_at_[hidden])
Date: 2005-01-12 12:46:42

> I do not think I was clear enough.  Let's look at the example again...
>
>
> void infinite_loop()
> {
>     // unit test framework can break infinite loops by timeout
> #ifdef __unix  // don't have timeout on other platforms
>     BOOST_CHECKPOINT("About to enter an infinite loop!");
>     while(1);
> #else
>     BOOST_MESSAGE(
>       "Timeout support is not implemented on your platform");
> #endif
> }
>
>
>
> If I were to write a similar test, I would rather it look something
> like...
>
>
> void infinite_loop()
> {
>     // unit test framework can break infinite loops by timeout
> #ifdef BOOST_TEST_HAS_TIMEOUT
>     BOOST_CHECKPOINT("About to enter an infinite loop!");
>     while(1);
> #else
>     BOOST_MESSAGE(
>       "Timeout support is not implemented on your platform");
> #endif
> }

Example uses ifdef because:

1. It used to demonstrate that timeout support is not supported everywhere
2. We know for sure it will hang if run on a system that does not support
timeout

You as a user shouldn't in general expect it and if it does hang and your
system doesn't  support timeout you will have to interrupt it yourself. The
only reason to use ifdef is the case when you are doing portable development
and willing to ignore possible infinite loop on some configurations for the
sake of ability to run a test as a part of regression suite (without human
intervention). Is it really the case?

Gennadiy