Any estimates on the difficulty of changing Linker.c to be able to use standard dynamic link calls (dlopen, ...) rather than having to be customized for every target platform. I'm considering making this work on powerpc-linux, but I'm thinking it might be a more useful task to make the linker more portable, rather than just hacking in another platform. There are a few platforms (linux alpha, linux powerpc) where the compiler wasn't very difficult to get working, but ghci doesn't work because of the linker problem. Dave
On Wed, 2004-06-02 at 17:44, David Brown wrote:
Any estimates on the difficulty of changing Linker.c to be able to use standard dynamic link calls (dlopen, ...) rather than having to be customized for every target platform.
I asked Simon M about this recently: http://www.mail-archive.com/glasgow-haskell-users@haskell.org/msg06189.html He has some notes at: http://www.haskell.org/~simonmar/shared-libs Duncan
On Wed, Jun 02, 2004 at 06:19:00PM +0100, Duncan Coutts wrote:
On Wed, 2004-06-02 at 17:44, David Brown wrote:
Any estimates on the difficulty of changing Linker.c to be able to use standard dynamic link calls (dlopen, ...) rather than having to be customized for every target platform.
I asked Simon M about this recently:
http://www.mail-archive.com/glasgow-haskell-users@haskell.org/msg06189.html
Actually, this isn't the question that I have. This is about ghci's dynamic loading of C objects (or other similar language). There is a handrolled linker in the ghci code. I'm interested in replacing this handrolled linker with dlopen... so that it is easier to port. Dave
participants (2)
-
David Brown -
Duncan Coutts