[Mono-dev] question about System.Web.Extensions unit tests

[Noam Lampert]

I agree that the tests need modification. 
There is no intention of compelling Mono to fail, and there are plenty
of solutions that Mono should pass the tests. I described some below.

However, having an #ifdef in the tests does not mean that we (Mono and
Grasshopper) are "kept away from it INCORRECTLY because of #if
TARGET_JVM THAT DOES NOT PASS ON .NET". On the contrary, it means that a
human being reviewed the difference between Mono and .NET, clearly
marked the difference in the code of the test, and decided that this
difference is acceptable. It is far better than creating tests that pass
on Mono and fail on .NET.

Noam

-----Original Message-----
From: Atsushi Eno [mailto:atsushi at ximian.com] 
Sent: Monday, June 02, 2008 5:12 PM
To: Noam Lampert
Cc: mono-devel-list
Subject: Re: [Mono-dev] question about System.Web.Extensions unit tests

Hello,

That is very strange argument. Do you understand that your way of
testing compelling Mono to fail while you guys are kept away
from it INCORRECTLY because of #if TARGET_JVM THAT DOES NOT PASS
ON .NET? I regard it as inappropriate practice and strongly suggest
to rewrite those tests.

Atsushi Eno

Noam Lampert wrote:
> Hi Atsushi,
> 
> We care about .NET. Please do not rewrite the tests to fail on .NET.
> Passing on .NET gives strong confidence that our implementation is
> correct, and not only non-regressive.
> 
> As you can see, the order of the members in Grasshopper and .NET is
> different and hence the #if.
> 
> The best way to fix these tests is to have a strong comparison engine
> that will understand that these specific differences are not
important,
> but this is not a small task.
> 
> Perhaps a simpler workaround is to succeed on either of the strings
> (e.g. Assert.IsTrue(script == res1 || script == res2); ).
> Another alternative is to change the #if TARGET_JVM to #if !DOTNET.
> 
> Noam
> 
> -----Original Message-----
> From: Atsushi Eno [mailto:atsushi at ximian.com] 
> Sent: Monday, June 02, 2008 3:51 AM
> To: mono-devel-list at lists.ximian.com
> Subject: [Mono-dev] question about System.Web.Extensions unit tests
> 
> Hello,
> 
> I have been seeing NUnit test failures in Sys.Web.Extensions. It keeps
> the buildbot orange (warned), so today I checked the tests to see
> what should be expected.
> 
> There I found that such lines that treats TARGET_JVM as different:
> 
> --------
> #if TARGET_JVM
> 			Assert.AreEqual ("$create(My.Type, null, null,
> {\"myName2\":\"myCompId2\",\"myName1\":\"myCompId1\"},
> $get(\"Element1\"));", script);
> #else
> 			Assert.AreEqual ("$create(My.Type, null, null,
> {\"myName1\":\"myCompId1\",\"myName2\":\"myCompId2\"},
> $get(\"Element1\"));", script);
> #endif
> --------
> 
> Though I see no reason to differentiate tests for GH. To my
> understanding from our meeting at Mono summit 2006 time, Mainsoft
> does not use #if TARGET_JVM without any reasonable differences,
> so I would like to know the reason why there are such switches all
> around the tests so that they can pass on GH while they will fail
> on Mono, if any.
> 
> Actually when I ran make PROFILE=net_3_5 run-test-ondotnet, there
> was no error report. So if those tests are rewritten, they will
> fail on .NET.
> 
> If it is only about behavioral difference in generic Dictinary`2,
> then those tests should not be conditionalized to legalize only in
> TARGET_JVM and rewritten to fail only on .NET. (Who cares after all?)
> 
> Atsushi Eno
> 
>