Re: Any one using Compaq(DEC) Tru64 UNIX?

[Marcus G. Daniels]

(Bill Northcott's new patch bounced, due to size.  I've copied them to
ftp://ftp.swarm.org/pub/swarm/src/testing/OSF5-Tru64 for now.)

BN> make check fails on only 4 tests

Sigh, I can see you're having fun.  ;-)

BN> collections:archiver    (HDF5Object.m :800   'differing source and target
BN> types')
BN> defobj:archiver (HDF5Object.m :800   'differing source and target types')

I bet it has to do with a lack of typing by GCC for 64 bit types of
some kind.  It should print out too integers after that..

BN> defobj:Forwarding    (Just what is going on around L506 of
BN> libobjc/sendmsg.c where it mentions fragile code??)

GCC has a half-working means by which you can indicate arguments
and construct a call on the fly.  On some systems it doesn't work, though, and
figuring out where values land in a call frame must done by hand.
You might try the optimistic __builtin_return default and see
if the tests pass.


Here's the text of his message:
----------------

I have worked on this a lot more.  I am still getting up to speed on this.

The computer is a Compaq XP1000 with an Alphaev6 CPU and 4Gb of DRAM.  The
OS is Tru64 UNIX v5.0
The compiler is gcc 2.95.2 compiled with two small patches which are
attached.  These remove errors due to the EV6 CPU not being recognised.

We are using the following defines for the build of the Swarm libraries
-D_HIDDEN_VA_LIST -D_XOPEN_SOURCE_EXTENDED -D__LANGUAGE_C__
The first of these fixes what appears to be a mistake in the gcc compiler
install script.  It changes _HIDDEN_VA_LIST in stdio.h to
_HIDDEN_DUMMY_VA_LIST but fails to do the same in va_list.h.  This results
in a preprocessor error if stdio.h is included before stdargs.h.
The second and third defines appear to be the correct ones for the Compaq
C headers.

We also applied some small patches a copy of which are attached.  These
remove errors due to the EV6 cpu not being recognised.

make check fails on only 4 tests
collections:archiver    (HDF5Object.m :800   'differing source and target
types')
defobj:archiver (HDF5Object.m :800   'differing source and target types')
defobj:Forwarding    (Just what is going on around L506 of
libobjc/sendmsg.c where it mentions fragile code??)
simtools:AppendFileTest

The HDF5 libraries pass all their tests.

I can compile and run the ArtStkMkt code, but it will core dump after 1000
iterations with a 'Floating exception'.

Any suggestions?  If you would like more information please ask.

Bill Northcott
School of Banking and Finance
UNSW


                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.