Re: Resurrecting hppa-elf

[Jeffrey A Law <law at cygnus dot com>]

  In message <Pine.LNX.4.21.0004191316090.32725-100000@front.linuxcare.com.au>y
ou write:
  > On Tue, 18 Apr 2000, Jeffrey A Law wrote:
  > 
  > > The 32bit PA elf needs a *complete and total* rewrite.  Anything based on
  > > the old code is horribly broken.
  > > 
  > > I've tried to explain the issues to HP & the puffins, but they weren't
  > > particularly interested in listening to why the old 32 PA elf code was
  > > totally inappropriate for their needs.  I sincerely hope they didn't
  > > resurrect the old stuff.
  > 
  > That seems to be what has happened.  I'm willing to spend quite a bit of
  > time to rewrite it properly.  Do you mind pointing me at an email archive
  > with the above issues you tried to talk about, other than the puffin one,
  > which I'm dredging through at the moment?
I don't think there's anything archived.

The single most important issue is to settle on a usable ABI.

The existing ABIs are:

  32bit Portable runtime -- horribly inefficient.
  32bit hpux (aka PA32) -- exists in the compiler.  Horribly complex to
               implement the necessary support in an ELF assembler/linker
               (yes, the  ABI has significant impacts on the assembler/linker).
  PA64 Exists in the compiler.  Marginal impact on the linker/assembler, but
       is only suitable for wide mode.

I believe ultimately a new ABI will have to be defined that is suitable for
a linux port and which doesn't have all the braindamage found in the 32bit
hpux ABI.  Once that's done and implemented, providing proper elf32-hppa
support isn't so bad since it'll look pretty normal.

  > How far along is your elf64-hppa work?
It's done, but I doubt it's going to help you since it's got a lot of
hpux specific implementation details.

jeff