From: Bernard V. B. <Ber...@te...> - 2002-11-22 12:01:56
|
Tomas, If the fork attributes would be removed and set to true internally, and (only) the NAntContrib tests would be run using the command line runner, would there be other problems that need to be solved to make the NUnit2 task work for real-world usage? Bernard > -----Original Message----- > From: Tomas Restrepo [mailto:to...@mv...]=20 > Sent: Friday, November 22, 2002 4:25 AM > To: nan...@li... > Subject: Re: NUnit 2 support (was: RE: [nant-dev] FAQ: The=20 > next NAnt version) >=20 >=20 > John, >=20 > > I noticed that NUnit2Task was executing very different code=20 > based on the > > fork attribute; code that looks suspiciously like the=20 > current code in > > nunit-console. >=20 > I'd like to refactor that code slightly, but that's a different thing > altogether :) >=20 > > Maybe this is a simpler workaround than using exec? >=20 > It is a good workaround for many apps. However, I'd like to=20 > point out two > things: >=20 > - With the current NUnit2 architecture, the fork=3D"false" (the = default) > option makes no sense, as it really, really, only serves to test nant > itself. In fact, fork=3D"true" should be the default, as that=20 > was how NUnit2 > was concieved: a new appdomain is created, and assemblies to=20 > test are loaded > into that appdomain from inside itself (hence all the=20 > problems we have). >=20 > -Even with fork=3D"true", nantcontrib tests still cannot run.=20 > The problem is > that nantcontrib requires both access to nantcontrib's and=20 > nants dlls, and > with the current architecture this is hard to do without some=20 > rather ugly > hacks (at least as I see it). >=20 > -- > Tomas Restrepo > to...@mv... |