Building Python 2.3 on HP-UX 10.20 - pthread failure

[Dan Cescato]

When I try to build Python 2.3 on HP-UX 10.20 using their cc compiler,
I get a compile failure. It comes down to Python/thread_pthread.h
deciding that my machine is an unsupported system, as documented in
the README file. ("Systems defining PY_PTHREAD_D4..."). Apparently,
Python 2.3 doesn't like their implementation of pthreads. However, in
the "Platform specific notes" section, there is a note about HP-UX
10.20 which makes me believe that someone has gotten it to work.

The application we're writing in Python needs pthreads. Is there an
upgrade patch available from HP that will get this to build? Our
technology director has ruled out gcc. Has anyone gotten this to
build?

 

[Martin v. =?iso-8859-15?q?L=F6wis?=]

When I try to build Python 2.3 on HP-UX 10.20 using their cc compiler,
I get a compile failure. It comes down to Python/thread_pthread.h
deciding that my machine is an unsupported system, as documented in
the README file. ("Systems defining PY_PTHREAD_D4..."). Apparently,
Python 2.3 doesn't like their implementation of pthreads. However, in
the "Platform specific notes" section, there is a note about HP-UX
10.20 which makes me believe that someone has gotten it to work.

The application we're writing in Python needs pthreads. Is there an
upgrade patch available from HP that will get this to build?

Please read http://www.python.org/peps/pep-0011.html. Python 2.3 might
work just fine on your system, however, we (actually, I) believe that
the this specific aspect of Python is little used. So unless you act
as described in the PEP, the existing support for the feature will be
*removed* rather than the problem being corrected.

So far, I had no information that it is HP-UX 10.20 which defines
PY_PTHREAD_D4; I was believing that the code dealing with it is dead.
Restoring the feature is trivial at the moment, but it takes a
volunteer to actually do so.

Regards,
Martin

 

[Dan Cescato]

Actually, what sets PY_PTHREAD_D4 is a combination of the existence of
__hpux and _DECTHREADS_ (defined in cma.h). Based on what I see in
the PEP doc, I assume the code to support this is still in Python
2.3.1. I was planning to comment out the #error that brings the show
to a halt, and get on with life. As far as volunteering, let me talk
to the boss about this. (BTW, the response I was really hoping for
was "Just apply HP patch XXXX, and you'll be fine. Oh well...)

 

[Anthony Baxter]

Actually, what sets PY_PTHREAD_D4 is a combination of the existence of
__hpux and _DECTHREADS_ (defined in cma.h). Based on what I see in
the PEP doc, I assume the code to support this is still in Python
2.3.1. I was planning to comment out the #error that brings the show
to a halt, and get on with life. As far as volunteering, let me talk
to the boss about this. (BTW, the response I was really hoping for
was "Just apply HP patch XXXX, and you'll be fine. Oh well...)

It's a big patch <wink>, but the patch called "HP/UX 11.11" builds
clean, out of the box, on HP/UX. (Note, though, that if you're on
Itanium, you'll need to build without optimisation - the HP compiler
screws up the executable in a rather spectacular fashion).

I've been using HP's "test drive" systems for this - they don't appear
to have a HP/UX 10 system there, so I'm afraid I can't offer more
specific help.

 

[Martin v. =?iso-8859-15?q?L=F6wis?=]

Actually, what sets PY_PTHREAD_D4 is a combination of the existence of
__hpux and _DECTHREADS_ (defined in cma.h).

That is a useful hint. __hpux probably should be defined on the
system, but where does _DECTHREADS_ come from?

Regards,
Martin

 

[Dan Cescato]

That is a useful hint. __hpux probably should be defined on the
system, but where does _DECTHREADS_ come from?

Regards,
Martin

_DECTHREADS_ is #define'd in cma.h. It gets sucked in, eventually, by
anything that gets within a country mile of threads.

 

[Dan Cescato]

It's a big patch <wink>, but the patch called "HP/UX 11.11" builds

clean, out of the box, on HP/UX. (Note, though, that if you're on
Itanium, you'll need to build without optimisation - the HP compiler
screws up the executable in a rather spectacular fashion).

I've been using HP's "test drive" systems for this - they don't appear
to have a HP/UX 10 system there, so I'm afraid I can't offer more
specific help.

The bummer is that I work for a small software company, and some
customers are still on these creaky old systems. What we won't do for
a customer whose check clears the bank...

 

[Anthony Baxter]

The bummer is that I work for a small software company, and some
customers are still on these creaky old systems. What we won't do for
a customer whose check clears the bank...

I'm pretty sure none of the folks on python-dev have access to a HP/UX 10
system. So there's three possibilities for getting this fixed:

.. Since you have access to such a system, figure out what needs to be changed
and supply a patch to the SF tracker
.. Make a HP/UX 10 system available to someone who can fix the problem
.. Convince HP to add a HP/UX 10 system to the test drive farm.

It's virtually impossible for someone to fix this who doesn't have access
to a system to work on the problem. I have access to only HP/UX 11 systems,
so I can only make sure python builds out of the box on those systems. (For
instance, I'm following up with HP on a couple of issues on the HP/UX on
Itanium platform to get bugs on the platform fixed.)


Anthony

 

[AK]

I'm pretty sure none of the folks on python-dev have access to a HP/UX 10
system. So there's three possibilities for getting this fixed:

. Since you have access to such a system, figure out what needs to be changed
and supply a patch to the SF tracker
. Make a HP/UX 10 system available to someone who can fix the problem
. Convince HP to add a HP/UX 10 system to the test drive farm.

It's virtually impossible for someone to fix this who doesn't have access
to a system to work on the problem. I have access to only HP/UX 11 systems,
so I can only make sure python builds out of the box on those systems. (For
instance, I'm following up with HP on a couple of issues on the HP/UX on
Itanium platform to get bugs on the platform fixed.)


Anthony

It's never too late to have a happy childhood.

Love the quote.. where from?

-AK

 

[Anthony Baxter]

Love the quote.. where from?

Bloom County.