Hugs still (again?) has a thread scheduling behavior that breaks Fran immediately. The required Prelude tweak is described at http://research.microsoft.com/~conal/Fran/news.htm under the "1.15" heading. Will this Prelude change get made at some point? - Conal
Maybe (I have to think through the consequences a bit first), but until that time you may want to insert calls to 'yield' immediately after a call to 'forkIO' -- that should have the same effect as the tweak you refer to. --sigbjorn ----- Original Message ----- From: Conal Elliott To: hugs-bugs@haskell.org Sent: Tuesday, January 08, 2002 10:04 Subject: hugs thread scheduling Hugs still (again?) has a thread scheduling behavior that breaks Fran immediately. The required Prelude tweak is described at http://research.microsoft.com/~conal/Fran/news.htm under the "1.15" heading. Will this Prelude change get made at some point? - Conal
Conal:
Hugs still (again?) has a thread scheduling behavior that breaks Fran immediately. The required Prelude tweak is described at http://research.microsoft.com/~conal/Fran/news.htm under the "1.15" heading. Will this Prelude change get made at some point?
Sigbjorn:
Maybe (I have to think through the consequences a bit first), but until that time you may want to insert calls to 'yield' immediately after a call to 'forkIO' -- that should have the same effect as the tweak you refer to.
Changing the thread scheduling behaviour could break HGL so please be very careful. The testsuite has a number of tests for the scheduling behaviour. Some of these are simply to detect changes - they indicate that something might break not that something necessarily has. At least one is to check that the consumer-producer pattern used in HGL (and elsewhere) works. [Hmmm, I thought I'd mailed Conal about the change when I made it - I certainly meant to.] -- Alastair Reid reid@cs.utah.edu http://www.cs.utah.edu/~reid/
participants (3)
-
Alastair David Reid -
Conal Elliott -
Sigbjorn Finne