On May 15, 2005, at 3:54 AM, Bob Ippolito wrote:
> I got access to the sourceforge compile farm, which runs Mac OS X
> 10.2 server, so now I have some way of testing Jaguar (minus any
> tests that need WindowServer access, like many of the AppKit tests).
> I did fix one bug with str bridging, but it looks like there are
> issues with protocols on Mac OS X 10.2. Other than that, we seem
> to be OK. Merely importing objc.test.protocol crashes in
> __objcInit (not sure why.. don't care that much right now).
> Skipping test_protocol, the rest of objc.test works.
> There are also "bugs" in the tests for PyObjCTools.KeyValueCoding
> that assume the Objective-C implementation of valueForKeyPath:
> should support array operators. Failing these on Mac OS X 10.2 are
> Similarly, this is also expected to fail, because valueForKey: on
> Mac OS X 10.2 doesn't do the array stuff:
> FAIL: test_TestArrayCoding (objc.test.test_ctests.CTests)
I've narrowed it down, importing objc.test.protocol crashes because
there is a bug in the Objective-C runtime in 10.2, it can't define
protocols in MH_BUNDLE, so that test suite is now skipped if not
Everything in PyObjCTools.test and objc.test that currently fail on
Mac OS X 10.2 are expected to fail and will always be expected to
fail, as they're testing functionality that didn't exist at the time.
I haven't and can't do any of the AppKit tests, but I don't remember
anything meaningful being in that suite anyway.
... so, in other words, svn trunk is Jaguar compatible (and has been
since I fixed that str bug).