lewie@EE.ECN.PURDUE.EDU (Jeff Lewis) (05/06/88)
Well I've gotten a lot of mail about this COFF issue (surprise, surprise!). Michael Tiemann and Steve Simmons pointed out the clincher: BSD stabs (even tho' a hack) is a big win because of the flexibility. I hadn't thought about it, but taking a closer look, COFF symbol tables are strongly oriented towards C. Short of resorting to other hacks, they leave little room for the additional types in G++ (or in K - but more on that later...). There are rumours in the air of COFF being upgraded in the future by AT&T/Sun - perhaps this issue will come up again then. On pure simplicity (but not elegance) - I'm going to opt for RMS's original method with the exception that the header be pure native (BSD or COFF) whilst the symbol table be BSD all the way. I've done a lot with COFF format recently, so I will try my hand at COFF library translation and whatever small changes that would be required to the loader and gdb. I hope to have this done by/on this Saturday for other reasons, so I'll let everyone know how it turns out. One obvious problem: this may work fine for me, but those who have SysV, but not a 68000 aren't helped at all. To whomever would know: is the ld.c in the g++ distribution the one true GNU loader? --Jeff Lewis (lewie@ee.ecn.purdue.edu, pur-ee!lewie)