🚀 go-pugleaf

In article <9204020432.AA09123@strawberry.cs.wwu.edu> you write:
>gcc-1.40 with Bruce's as/ld/ranlib/....
>
>  A stage2 compiler produced an exact copy of the stage2 compiler.

By the way, if people are using Bruce's as it could make sense to port
gcc 2.1 to use Bruce's as syntax - I guess it wouldn't be that hard,
the rewritten print_operand_address is a lot easier to understand and
modify than in 1.40.

>GNU ar/ld/....
>  ar and ranlib appear to have some real strange problems.  Ranlib
>  complains about bad object files when the libc.a does not have 
>  a __.SYMDEF.  After ranlib adds one, ranlib runs without complaining!

Hmm, I remember something weird about __.SYMDEF also, but not sure of
the details - anyway, it was something you could solve (perhaps a
complaint about bad archive member __.SYMDEF or something).

>gcc-1.40 with GNU as/ranlib/ld ....
>
>  I never did get this version producing working executables.

With what diffs?  If you are using the alignment Bruce is using, there
are some weird problems with gas sometimes coredumping and perhaps
also producing incorrect code.

>gcc-2.1 with GNU as/ranlib/ld ....

>  When this compiler is used to recompile itself, it does get executables
>  for most of the programs.  (GNU ld core dumped on cc1plus when compiled
>  with the gcc-2.1 compiler.)  When the gcc-2.1 compiled compiler tries
>  to do some work, it crashes fast!  (well, it says there is a syntax
>  error at a location where there is no syntax error.)  This did not
>  pass the "stage2" test!

Hmm, are you using libraries compiled with gcc 2.1?  Perhaps you
should, since the alignment is perhaps different (not sure what
exactly people talk about when they are talking about gcc-1.40-wbc).
My libraries are compiled with gcc 1.39 (with diffs which should be
functionally the same as in 2.1, I think they should be available on
nic.funet.fi).

//Jyrki