Chris Dornan <[EMAIL PROTECTED]> writes:
> ghc has started getting quite chatty in 2.0x; some of the chatter is very
> useful. However, could I suggest that the warning messages provided by default
> be deployed sparingly. I would be inclined to provide warnings only if a
> warnings flag is provided -- it is nice to have large builds proceed with only
> the important stuff hitting the screen. While developing, the warnings would
> be enabled of course.
By default, you get a standard set of warnings with GHC. These are:
-fwarn-incomplete-patterns and -fwarn-overlapping-patterns. There are
also -fwarn-name-shadowing and -fsignatures-required, which aren't
enabled by default.
I take your point that this isn't very consistent: there should be a
way to turn off all warnings easily. What do other people think?
The options are:
* have all warnings off by default, a standard set of warnings
being available by adding the -W command line option.
* have a standard set of warnings on by default, with all
warnings being turned off by the -Wnot (or something) flag.
This is in addition to each warning being selectable independently via
the relevant command-line flag.
> Here are some of the diagnostics with my comments:
>
> foo.lhs:492: No explicit method nor default method for `bar'
> in an instance declaration for `PriorityQueue'
This is a warning, and should say so.
> foo.lhs:155:
> Warning: Possibly incomplete patterns
> in the definition of function `foobar'
The incomplete pattern detector is currently sound but not complete (I
believe), but this is being worked on.
> NOTE: Simplifier still going after 4 iterations; bailing out.
> ghc: module version changed to 4; reason: exports changed
These messages come from us wanting to know exactly what's going on -
any hint of trouble could point to a possible bug. I agree this stuff
should be turned off by default.
Cheers,
Simon
--
Simon Marlow [EMAIL PROTECTED]
University of Glasgow http://www.dcs.gla.ac.uk/~simonm/
finger for PGP public key
