[Haskell-cafe] STM, newArray, and a stack overflow?

[Bas van Dijk]

On 23 March 2011 17:19, Jake McArthur <jake.mcarthur at gmail.com> wrote:
> On 03/23/2011 10:34 AM, Ketil Malde wrote:
>>
>> It works (calling the same function) from GHCi, but breaks when
>> compiled.  Also when compiling with -O0.
>
> Confirmed for GHC 7.0.2. Works fine in GHCi, but compiling it (in my case,
> with -O) and running the executable causes a stack overflow unless I run it
> with +RTS -K16m, and even then it spends 90% of its time in GC. This looks
> like it is probably a bug, to me. Maybe it should be reported in GHC's Trac?

It looks like a bug indeed.

The problem can be reduced to just:

atomically $ replicateM 1000000 (newTVar undefined)

or even simpler:

replicateM 1000000 (newIORef undefined)

Bas