[Haskell-cafe] multi-thread and lazy evaluation

[timothyhobbs at seznam.cz]

I'm not sure that there is anything "great" about this bug.  It seems to me 
to be a rather severe demonstration of a somewhat already known design flaw 
in the runtime :(

Could you please comment on the actual bug rather than replying here so that
the devs see that this behaviour has been confirmed?

Tim


---------- Původní zpráva ----------
Od: Corentin Dupont <corentin.dupont at gmail.com>
Datum: 25. 12. 2012
Předmět: Re: [Haskell-cafe] multi-thread and lazy evaluation

"

Great, with me compiled with ghc -threaded the bug shows up.

However, runnning "main" in ghci doesn't show the bug (it finishes 
correctly).

I have GHC 7.4.1.

 

Corentin
 


On Tue, Dec 25, 2012 at 3:34 PM, <timothyhobbs at seznam.cz
(mailto:timothyhobbs at seznam.cz)> wrote:
" 
This seems like a bug in GHC. But it has nothing to do with MVars.  I've 
narrowed this down and filed a bug report here:

http://hackage.haskell.org/trac/ghc/ticket/7528
(http://hackage.haskell.org/trac/ghc/ticket/7528)

Timothy


---------- Původní zpráva ----------
Od: Yuras Shumovich <shumovichy at gmail.com(mailto:shumovichy at gmail.com)>


Datum: 24. 12. 2012
Předmět: Re: [Haskell-cafe] multi-thread and lazy evaluation




"On Mon, 2012-12-24 at 16:16 +0100, timothyhobbs at seznam.cz
(mailto:timothyhobbs at seznam.cz) wrote:
> The real question is, does this mean that GHC is stopping the world every 
> time it puts an MVar?

No, GHC rts only locks the MVar itself.
See here:
http://hackage.haskell.org/trac/ghc/browser/rts/PrimOps.cmm#L1358
(http://hackage.haskell.org/trac/ghc/browser/rts/PrimOps.cmm#L1358)

Yuras"




_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe at haskell.org(mailto:Haskell-Cafe at haskell.org)
http://www.haskell.org/mailman/listinfo/haskell-cafe
(http://www.haskell.org/mailman/listinfo/haskell-cafe)

"


"
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.haskell.org/pipermail/haskell-cafe/attachments/20121225/8cff2210/attachment.htm>