[Haskell-cafe] IO and State

[Jan-Willem Maessen]

Marcin 'Qrczak' Kowalczyk wrote:

>"Iavor S. Diatchki" <diatchki at cse.ogi.edu> writes:
>
>  
>
>>I find the argument a bit disturbing, as it seems to imply that it
>>is OK for the compiler to produce code without any context switches
>>at all
>>    
>>
>
>Note that in this case if the main program doesn't explicitly block
>on MVars, I/O nor timeout, then finalizers will never be run and would
>be kept in memory despite garbage collection. So such implementation
>would not be able to run certain programs which create lots of
>finalized objects, even if almost all of them become dead soon after
>creation.
>
This is little different from the situation with GC-based finalization 
in other languages (cf Java).  This is why finalizers are generally 
viewed as a backstop to clean up resources which were accidentally left 
un-freed, rather than a universal solution to cleanup.

-Jan-Willem Maessen