(had to resent this email as it got rejected the first time by the mailing list) On Wed, 2011-11-16 at 19:41 +0000, Ian Lynagh wrote:
On Tue, Nov 15, 2011 at 11:07:31PM +0000, Duncan Coutts wrote:
On Tue, 2011-11-15 at 18:12 +0000, Ian Lynagh wrote:
On Mon, Nov 14, 2011 at 11:15:17AM +0000, Duncan Coutts wrote:
On Thu, 2011-11-03 at 12:40 +0000, Simon Peyton-Jones wrote:
Don, Duncan: were you wanting to get these changes into 7.4?
Yes. They're now in the upstream repo for bytestring along with various other changes. So I hope that'll flow through into ghc's mirror.
I tried pulling them a few days ago, but validate didn't go through.
Due to other libs requiring bytestring == 0.9.*, or any other more interesting reason? If you don't recall, nm.
I don't think I looked into the failure in detail.
fyi, I've been trying to reproduce your validate observation, and I got quite a few -Werror caused errors in the process (which have been fixed in the upstream bytestring repo) This morning I ran a "validate" on a freshly checked out GHC source (which uses bytestring-0.9.2) tree, and got the following result: OVERALL SUMMARY for test run started at Thu Nov 17 12:30:14 CET 2011 3121 total tests, which gave rise to 10479 test cases, of which 0 caused framework failures 7673 were skipped 2733 expected passes 70 expected failures 0 unexpected passes 3 unexpected failures Unexpected failures: concurrent/should_run 5238 [exit code non-0] (normal) perf/compiler T5030 [stat not good enough] (normal) perf/compiler parsing001 [stat not good enough] (normal) Then I applied the modifications (see the two patches attached) to the ghc source-tree and the haskeline library required to make ghc compile with bytestring-0.10 instead of bytestring-0.9.2, made sure the bytestring library was really updated to the 0.10 version, cleaned the source trees, and finally re-ran "validate" Now I'm pleased to report, that the overall summary reported the exact same numbers with bytestring-0.10 linked into GHC instead of bytestring-0.9.2 :-) hth, hvr