Asynchronous exceptions and laziness bugs in Control.Concurrent.QSem/QSemN

[Bas van Dijk]

I just created a ticket for this so it's not forgotten:

http://hackage.haskell.org/trac/ghc/ticket/3944

Bas

On Thu, Mar 18, 2010 at 9:54 PM, Bas van Dijk <v.dijk.bas at gmail.com> wrote:
> Hello,
>
> after discovering the bugs regarding asynchronous exceptions and
> laziness in Data.Unique and Control.Concurrent.SampleVar I became a
> bit suspicious about the other concurrency mechanisms.
>
> It looks like Control.Concurrent.Chan is safe because it uses
> modifyMVar_ internally which performs the necessary block.
>
> Control.Concurrent.QSem and QSemN are not alright however! They both
> suffer the same bugs:
>
> * Potential dead-lock due to not blocking asynchronous exceptions.
> * Potential space-leak because of lazily putting numeric expression in a MVar.
>
> The attached patch fixes these bugs.
>
> regards,
>
> Bas
>