patch applied (testsuite): Accept output
mnislaih at gmail.com
Tue Sep 11 12:11:49 EDT 2007
On 11/09/2007, at 17:39, Simon Marlow wrote:
> I'm slightly worried at the idea of a special case for TypeRep.
> That doesn't seem right to me - I realise that we're showing the
> full internal details of TypeRep here, but the debugger does that
> for every type. It's a debugger, it's supposed to show you what's
> really going on. If we start adding special cases to get
> abstraction back again, when do we stop?
Yes, perhaps I was being too user-friendly in my appreciation.
The debugger shouldn't try to hide too many things.
> I think break024 is doing the right thing, we just need to
> construct a test that gives reliable results. Perhaps just
> avoiding throwDyn is good enough?
> Which reminds me - I think the printer defaults to using 'show'
> when it knows enough about the type, doesn't it? That behaviour
> worries me too. I don't like the idea that it magically switches
> from raw printing to 'show' when enough information about the type
> becomes available.
It switches to show when enough type information and evaluatedness
are available; the term needs to be completely evaluated before
switching to show.
We could definitely add a flag for this.
I can see myself wanting the current behaviour in GHCi, but
preferring more information when
using the debugger under a GUI, for instance.
More information about the Cvs-ghc