On 27.06.2007 18:26, Zoe Slattery wrote:
Hi Tony

The filesystem tests are diffficult and require some compromises - it isn't easy to work out exactly where.

Well, that's what we're trying to solve here, right? =)

If we change tests so that they pass on as many OS as possible, for example by changing %d to %i, we'll be relaxing some of the strictness of the test.

Yes, but in the same time I would not like to see 10 different tests for 10 particular OSes testing the same thing.

If we do this the functional ref should make it very clear that behaviour is OS dependent and that PHP is essentially just using the underlying OS function and passing back whatever it gets.

The only problem with that is we will be failing to test that PHP really does pass back what is expected for each OS. For example the behaviour of stat() on Linux/VMware was (if I understood it correctly) a PHP problem which we only found by coding the exact behaviour that was expected on Linux.

What is the most widely used platform? I always assume that it's Linux but don't know if that's correct.

I don't have any numbers too, but I _hope_ it's Linux.

What about having a strict set of tests that were specific to Linux (or specified versions of Linux) and a less strict version of the same tests that are designed to do some level of checking (eg %i) on all OS? That would be two sets (I don't think a set for each OS is realistic either).

I guess the best way would be to divide the huge tests into a set of nano-tests 
and:
a) either skip some of them on the platforms that are known to be broken;
b) or write separate tests for those platforms.

--
Wbr, Antony Dovgal

--
PHP CVS Mailing List (http://www.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php

Reply via email to