> On a slightly related topic: I was a bit surprised to notice that
> 5.02.2 appears to produce binaries that are incompatible with 5.02.1,
> given that it was only a minor version number change. It's
> not a huge
> deal (since this at least produced link errors), but it was a bit
> surprising. Was that intentional?
It turns out that it's hard to maintain binary compatibility, because of
the high level of inter-module optimisation that GHC does. When we ship
a new patchlevel, it's often the case that the libraries will export a
slightly different interface than before (even if nothing changed in the
source code), so any module compiled against the old libraries won't
work with the new ones.
In the case of 5.02.2 we also made a change to the virtual machine which
caused a binary incompatibility with older object files (this was to fix
the oversize heap check problem).
We should have emitted a "heads up", sorry about that. Please assume
that different compiler versions produce incompatible code unless we
claim otherwise.
Cheers,
Simon
_______________________________________________
Glasgow-haskell-bugs mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
