Re: Re[2]: [Haskell-cafe] How can I pass IOUArrays to FFI functions?

On Tue, Aug 21, 2007 at 12:50:22AM -0700, Ryan Ingram wrote:

Ah, sneaky. That code is fine because it uses unsafeCoerce# on "memcpy", changing memcpy from whatever type it is, into MutableByteArray# s# -> MutableByteArray# s# -> Int# -> s# -> (# s#, () #)

So as long as the GC understands MutableByteArray# it's safe; it's relying on the C calling convention being handled properly.

Which still isn't quite correct, because the code for base-2.1:Data.Array.Base.memcpy could still be perverse and trigger a GC. However, since base is version-locked to GHC, it can depend on as much undocumented behaviour as it needs. The worst that can happen is a few more testsuite failures when someone tries to change the compiler.

On 8/21/07, Bulat Ziganshin wrote:

...

Hello Stefan,

Tuesday, August 21, 2007, 10:08:59 AM, you wrote:

...

Your code is broken in a most evil and insidious way.

and this code, too? :)

freezeSTUArray :: Ix i => STUArray s i e -> ST s (UArray i e) freezeSTUArray (STUArray l u marr#) = ST $ \s1# -> case sizeofMutableByteArray# marr# of { n# -> case newByteArray# n# s1# of { (# s2#, marr'# #) -> case unsafeCoerce# memcpy marr'# marr# n# s2# of { (# s3#, () #) -> case unsafeFreezeByteArray# marr'# s3# of { (# s4#, arr# #) -> (# s4#, UArray l u arr# #) }}}}

...

Unreproducable bugs are rarely reported, but they do add to people's impression of how unstable a language/library is.

Stefan