Fixing #5360 and broken validate

[Simon Marlow]

On 01/08/2011 12:52, Manuel M T Chakravarty wrote:
> Simon Peyton-Jones:
>> |  As pushing patches that do not validate seems to be all the rage at the moment, I
>> |  push my fixes to #5360 although I cannot complete validate due to the appended
>> |  *unrelated* breakage.
>>
>> Yes, apologies for the original breakage which was my fault.  It was a Friday afternoon glitch.
>
> I am sorry for the sarcasm.  It wasn't so much directed at you (after all, I said, I am happy to fix that problem, which was a consequence of some of my changes), but at the fact that when I wanted to validate my fix, somebody else had introduced another entirely unrelated (and completely unnecessary) breakage.

I returned from my holiday and fired off a validate to check the state 
of the world, and I got:

Unexpected failures:
    indexed-types/should_compile  T3017 [stderr mismatch] (normal)
    indexed-types/should_fail     NoMatchErr [stderr mismatch] (normal)
    indexed-types/should_fail     T2544 [stderr mismatch] (normal)
    indexed-types/should_fail     T2627b [stderr mismatch] (normal)
    indexed-types/should_fail     T4099 [stderr mismatch] (normal)

These look like a simple case of somebody forgetting to commit updates 
to the sample output.  Could someone fix please?

And now let's try to get back to having the tree in a validate-clean 
state.  Things have been slipping lately!

Cheers,
	Simon