[m-rev.] for review: comment out castclass System.Type[]

[Peter Ross]

On Sun, Dec 16, 2001 at 05:32:35PM +1100, Fergus Henderson wrote:
> On 14-Dec-2001, Peter Ross <peter.ross at miscrit.be> wrote:
> > Peter wrote:
> > > ===================================================================
> > >
> > >
> > > Estimated hours taken: 8
> > > Branches: main
> > >
> > > System.Reflection.Assembly.GetExportedTypes returns an array of
> > > System.Type however in Beta 2 of the CLR, when you try and do a
> > > castclass to System.Type[] it fails.  The fix is not to do the
> > > castclass.
> > >
> > > compiler/ilasm.m:
> > >     Disable castclasses where the type is System.Type[], by commenting
> > >     the line out.
> > >
> > 
> > I have submitted this change, because on further investigation it appears to
> > be a bug.
> 
> What do you mean by "submitted"?
> Do you mean "committed"?
> 
Yes, I meant committed.

> And when you say "it appears to be a bug", do you mean that it appears
> to be a bug in the CLR, or a bug in the Mercury compiler?
> 
A bug in the CLR.

I tried the following

// Create an array of System.Type
ldc.i4.1
newarr class [mscorlib]System.Type

// Cast it to an object
castclass object

// Cast it back to the array
castclass class [mscorlib]System.Type[]

This ran correctly, but the following code doesn't

    // Get an array of System.Type which are exported from an Assembly.
callvirt class [mscorlib]System.Type[]
        [mscorlib]System.Reflection.Assembly::GetExportedTypes()

    // Do a redundant castclass to System.Type[]
castclass class [mscorlib]System.Type[]

Why does the first work and the second not?  I believe it is because
there is a bug in the CLR, where GetExportedTypes is returning the wrong
type information.

I believe that the System.Reflection stuff is implemented as unsafe code
so it would make sense that someone constructed the wrong type info
somewhere.

> I think it is a bug in the Mercury compiler,
> because `castclass' should not be used with array types,
> but I think your change does not properly fix the bug.
> 
What is the best way to get a definite answer on this?  Who should we
mail?

Pete
--------------------------------------------------------------------------
mercury-reviews mailing list
post:  mercury-reviews at cs.mu.oz.au
administrative address: owner-mercury-reviews at cs.mu.oz.au
unsubscribe: Address: mercury-reviews-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-reviews-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------