Re: Hugs on Windows, FFI

http://blogs.msdn.com/oldnewthing/archive/2003/11/04/55560.aspx Disabling DEP is definately a workaround, if you call VirtualProtect with the appropriate constants then it doesn't need DEP disabling. This is definatley a better solution! I am running an Intel machine at work, but I don't have admin so figuring out whether DEP is enabled or not, and changing it, is probably not going to be easy. Thanks Neil On 9/11/05, Brian Smith wrote:

On 9/9/05, Neil Mitchell wrote:

...

Hi,

Running "make check" I get the following, dropping most of the begining since it seems sensible enough. Apparently this should show if FFI works on Windows?

I noticed that your test results didn't include failures for Sparc.hs/Callback.hs/Sin.hs. After more investigation, I found that these tests only fail when Windows Data Execution Prevention (DEP) [1] is enabled. When I disable DEP for hugs.exe, these tests pass; when I enabled DEP then the tests fail. So, apparently the code that builds the thunks used for callbacks does not correctly set the executable/non-writable flags on the thunk after it is created. There was a very similar bug in GHCi/runghc that was fixed just a few weeks ago.

I think the solution to this problem is probably simple for somebody familiar with the GHCi solution, or somebody familiar with the Windows memory management API's. But, failing a fix, we should at least note this problem in the release notes, and the installer should automatically disable DEP for hugs.exe, runhugs.exe, and WinHugs.exe.

[1] http://support.microsoft.com/kb/875352

- Brian

_______________________________________________ Cvs-hugs mailing list Cvs-hugs@haskell.org http://www.haskell.org/mailman/listinfo/cvs-hugs