Re: [COMMIT] Cleanup HP-UX 10.20 support such that it builds again

[Mark Kettenis <kettenis at gnu dot org>]

   Date: Mon, 22 Nov 2004 09:31:40 -0800
   From: Randolph Chung <randolph@tausq.org>

   > 	* hppa-hpux-nat.c: New file.

   Mark, is there any particular reason why we can't fix hppah-nat.c
   instead of introducing a new file altogether? Is hpux10.20 really so 
   different from hpux11i?

Well, with software, everything is possible.  But to be honest, I
really think fixing hppah-nat.c would really be a lot of work, and I
think the end-result still won't be pretty.  To answer your question:
Yes, HP-UX 10.20 is radically different from HP-UX 11.xx.  If you
strip away all the layers of goo that the HP people have added, you'll
see that HP-UX 10.20 is a rather traditional ptrace(2) target, while
HP-UX 11.xx uses the HP-UX specific ttrace(2) function.  Actually,
most of the contents of hppah-nat.c, infptrace.c, inftarg.c and
infttrace.c is goo to make ttrace(2) look like ptrace(2).  Together
that goo adds up to more lines of code than the new code in
hppa-hpux-nat.c.

   if possible I'd like to avoid having so many pa files to maintain :)

Hopefully my next step can accomplish that.  I'm currently working on
a replacement for infttrace.c.  Actually it's already working pretty
well for 32-bit HP-UX 11.00, although follow-fork/vfork/exec and
"hardware" watchpoints don't work yet.  I'll probably check something
in today or tomorrow.  Then after I've re-implemented the missing
features we can throw the switch and get rid of hppah-nat.c,
infttrace.c and inftarg.c.  The end-result will be significantly less
code that will be much easier to maintain than the current code.

The major benefit of this all (and the emajor goal) is to make HP-UX
considerably less special than it is now.  I think this is essential
for the future of the HP-UX GDB port.  Without major surgery, it just
won't survive the changes being made to the core parts of GDB.

Cheers,

Mark