I think I've found a regression in Haskell Platform, but what do I do next?

Hi, Recently I had to switch from ghc 6.12.3 to ghc 7.0.1 because of the bug #4235. This switch has introduced a problem with one of third-party libraries I use ('satchmo', an interface for SAT solvers). The problem is that writing to stdin of a process started by System.Process.runInteractiveCommand silently fails provided it's stderr is not read. If stderr of the spawned process is redirected to /dev/null or fully read by the parent process, then problem does not come up. The troubled code is available at http://bit.ly/dNwrIV (lines 43-45). Changes between versions of the 'process' library (1.0.1.3 -> 1.0.1.4) bundled with ghc 6.12.3 and 7.0.1 only include one bug fix, which is not the source of the problem described, as far as I can understand. The problem is not present with ghc 6.12.* at all. Something has definitely changed between these two ghc versions. Since I've spent considerable amount of time on finding a workaround for the problem, I'd like to file a real bug report, but I'm not completely sure where to do it because there are several pieces of software involved (ghc itself, 'process', 'satchmo' and 'minisat2' SAT solver). Could somebody guide me please?

On Tue, Feb 22, 2011 at 7:08 AM, Nick Frolov
Something has definitely changed between these two ghc versions. Since I've spent considerable amount of time on finding a workaround for the problem, I'd like to file a real bug report, but I'm not completely sure where to do it because there are several pieces of software involved (ghc itself, 'process', 'satchmo' and 'minisat2' SAT solver). Could somebody guide me please?
There were a few I/O manager related bugs, one related to forking processes. These should be fixed in GHC 7.0.2 (and the next HP release), due soon. -- Johan

On Tue, 2011-02-22 at 07:12 -0800, Johan Tibell wrote:
On Tue, Feb 22, 2011 at 7:08 AM, Nick Frolov
wrote: Something has definitely changed between these two ghc versions. Since I've spent considerable amount of time on finding a workaround for the problem, I'd like to file a real bug report, but I'm not completely sure where to do it because there are several pieces of software involved (ghc itself, 'process', 'satchmo' and 'minisat2' SAT solver). Could somebody guide me please?
There were a few I/O manager related bugs, one related to forking processes. These should be fixed in GHC 7.0.2 (and the next HP release), due soon.
I've just reproduced the bug with GHC 7.0.2. It seems that it still has to be reported, but, again, I have no idea if it's caused by a possible regression in GHC. It could be that one of the libraries relies on the wrong behaviour of GHC 6.12.*, right?
participants (2)
-
Johan Tibell
-
Nick Frolov