Simon Peyton-Jones:
| What's the objective here?
For me the objective is to say "make validate" and come back later
to find a list of things that need fixing. Stopping at the *first*
error is unhelpful, because it means there's an hour or two (and
therefore because of other constraints a day or two) between
iterations.
The present situation wrt compile errors (eg warnings with -Werror
and Haddock) is akin to validate stopping at the *first* test that
fails. As it stands, validate helpfully runs *all* the tests, and
gives you a list of ones that fail.
In the same spirit what I'd like to see is
Unexpected warnings in modules:
simplCore/Simplify.lhs
codeGen/CgMonad.lhs
Unexpected Haddock failure ine
codeGen/CgExpr.lhs
Unexpected testsuite failures in
tc125
Do you see what I mean? I'm content that any failure above elicits
a "please fix" response rather than a "congratulations". But I'd
like to get all the way.
I have no clue how hard this is. I'm just trying to clarify what my
objective (which admittedly I did not make clear) is!
It might be tricky to achieve that, but it certainly would be very
helpful, too!
Manuel
| -----Original Message-----
| From: Ian Lynagh [mailto:[EMAIL PROTECTED]
| Sent: 23 August 2008 14:13
| To: Simon Peyton-Jones
| Cc: Manuel M T Chakravarty; [email protected]
| Subject: Re: After enjoying a working GHC HEAD build for a few
days...
|
| On Thu, Aug 21, 2008 at 08:27:44AM +0100, Simon Peyton-Jones wrote:
| > I suspect that is my fault, although I did validate. I will check
| >
| > Ian: could validate continue after a Haddock failure, and simply
report it along with other test
| failures? That would stop a Haddock failure completely borking
the validate run. Or would that be hard?
|
| What's the objective here?
|
| We could continue after a haddock error and report it at the end
of a
| build, but is that better than the developer fixing the problem
and then
| running "validate --no-clean"? If you're about to push then you
need to
| fix the problem anywy.
|
| Or do we not want to haddock as part of the validate process? This
means
| that the docs will bitrot, and the nightly docs won't build.
|
| One thing that isn't helping is that some versions of haddock are
| pickier than others over what syntax is allowed.
|
|
| Thanks
| Ian
|
_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc
