Re: Problem with library images on Windows (again)

[Juanma Barranquero <lekktu@gmail.com>]

On 5/19/05, Eli Zaretskii <eliz@gnu.org> wrote:

> Can you set a watchpoint on the address where EBX is saved on the
> stack, and then see what code clobbers that value?

Yes. tiff_load pushes EBX into the stack, and then goes around happily
pushing arguments and calling functions from the TIFF library, which
apparently don't remove their arguments from the stack (nor does it
the tiff_load code). So the stack keeps growing and, by the time
tiff_load does the POP EBX, it gets random garbage. On returning from
tiff_load the frame pointer is restored, so the spurious stack data is
just discarded.

So it seems like a mismatch between the calling convention used to
implement the TIFF library and the one MSVC thinks the TIFF functions
are using.

I'll dig around a little more. But any help is much appreciated.

-- 
                    /L/e/k/t/u