Re: libffi stdcall closures

[Timothy Wall <twalljava@dev.java.net>]

Ok, thanks, I got a test that runs only for w32.  It also uncovered a  
bug in the cygwin gcc I am using (3.4.4 cygming special); it pops  
arguments for stdcall calls unless invoked in "-mno-cygwin" mode,  
which is likely to screw up terminal output.  Now I have to figure  
out a way around a) the arg popping and b) a good way to verify the  
stack is consistent before/after the call.

I'll have to check the config changes; they're probably pretty  
minor.  I know python has a couple of libffi snapshots, so they at  
least might find a standalone libffi useful.

On Nov 8, 2007, at 12:48 PM, Tom Tromey wrote:

>>>>>> "Timothy" == Timothy Wall <twall@users.sf.net> writes:
>
> Timothy> I'm working on a patch for win32/x86 stdcall closures to  
> support JNA
> Timothy> (http://jna.dev.java.net), a ctypes-like library for  
> Java.  I'm
> Timothy> working off of a libffi snapshot from the gcc trunk.
>
> Nice.  JNA looks cool.
>
> BTW, I think I read somewhere that you have some standalone configure
> changes for libffi...?  We could put those in the separate libffi
> repository if that would help you.  I occasionally think it would be
> nice to revive the standalone libffi -- but only if there's interest.
>
> Timothy> I've got the closure code working now, and I need some  
> pointers on
> Timothy> adding a win32-only stdcall callback test to the testsuite  
> (or a
> Timothy> general one if stdcall is supported elsewhere).
>
> libffi uses the dejagnu 'dg' code, so you should be able to just drop
> your test cases into the test suite.  If the test is platform-specific
> I think you can use the "target" clause to dg-do.  Grep for "target"
> in gcc/testsuite/**/*.c to see uses.
>
> Tom