
Any chance of getting ticket #608 onto the list for the 6.6 release? Ticket #608 is "Make the NCG able to compile the RTS". Please correct me if I am wrong, but my understanding of the dependencies suggests that this is the key obstacle to dynamically linked builds on x86 platforms. My reasoning is that GCC generated position independent code conflicts with the Mangler and that the RTS is the only code that must be built with the Mangler. Reilly Hayes

Reilly Hayes wrote:
Any chance of getting ticket #608 onto the list for the 6.6 release? Ticket #608 is "Make the NCG* able to compile the RTS". *Please correct me if I am wrong, but my understanding of the dependencies suggests that this is the key obstacle to dynamically linked builds on x86 platforms. My reasoning is that GCC generated position independent code conflicts with the Mangler and that the RTS is the only code that must be built with the Mangler.
Yes, that's right. Dynamically linked libraries on x86 & x86_64 Linux would be a very nice thing to have in 6.6, at the moment I'm not sure whether we should aim for it. If we do, then #608 would be required, as you say. Last time I looked at .so's on x86_64, I almost had it working but for some difficulty convincing the linker to generate the right kind of fixups for the main program. I think Wolfgang Thaller had .so's working on Linux/x86 at one point. DLLs are easier than .so's, and would be a good way to sort out the rest of the dynamic linking infrastructure (build system, install layout etc.). If anyone is interested in working on any of this, please get in touch. I have some partial uncommitted patches for PIC on x86_64. Cheers, Simon
participants (2)
-
Reilly Hayes
-
Simon Marlow