🚀 go-pugleaf

In article <FRANL.93Jan27115049@draco.centerline.com> franl@centerline.com (Fran Litterio) writes:
>> Yes, you can get a modified gcc-2.3.3.u3 for PA HP's from
>> jaugar.utah.edu:/dist.
>> I just downloaded hpuxbin.tar.Z, which contained the executables
>> for gcc, gdb, etc. that worked (with symbolic debugging & all)
>> without modification.
>
>I thought that HP was refusing to "free" its debugger symbol table
>format (specifically the header file that defines all of the debugger
>symbol data structures).  That would mean no Gnu product can read or
>write HP debug info, because the FSF won't use code that isn't free,

This is correct.  The tools from jaguar (not jaugar, as above) generate
debug information using "stabs" format.  This means that gcc/g++ code
can be debugged with gdb, but you can't use gdb to debug code produced
by HP's compilers, and you can't use HP's debugger to debug gcc/g++
code.

>Which is it: Do gcc and gdb understand HP debug info or don't they?

They don't -- they use an incompatible format, the same format used
on BSDish systems.

>I too have heard about work at the University of Utah to teach gcc/gdb
>about HP debug info, but I have also heard that HP has effectively
>killed that work by refusing to free the necessary header file.

They've killed it even deader than that.  The Utah people have signed
a nondisclosure agreement, meaning they could be sued if they reveal
HP trade secrets even if they found some other way to do the job
(without HP proprietary headers).  Either HP will need to relent,
or someone else (who has not signed a nondisclosure) will need to
do the work.

I've had similar problems with HP, attempting to port a freely
redistributable system to the HP.  This system does incremental linking,
which is very tricky on the HP-PA architecture.  HP's attitude on things
like this is not like other companies.  I could get the information,
perhaps, by signing a nondisclosure (HP's rivals would simply send me
the information), but that would prevent releasing free software that
uses the ideas.

To be fair, my contact at HP found me some non-HP experts to talk to
so I could make some progress on my problem.  I'm still bothered by
the fact that it is evidently against company policy to permit me
to talk to an HP employee who, say, knows how the linker works.

Given this attitude on the part of HP, I'd have to recommend that research
organizations avoid the company until they have a change of policy -- it
shouldn't be a secret as to how to do advanced programming on your
workstation.  If it is, and you need to do this kind of programming, it
might be a good idea to buy someone else's.



--
Joe Buck	jbuck@ohm.berkeley.edu