Some Comments on Building Ptolemy0.5 on Sparc SunOS 4.1.X (gcc)

Some Comments on Building Ptolemy0.5 on Sparc SunOS 4.1.X (gcc)

Post by Terry Dom » Wed, 24 Aug 1994 02:12:30



Hi All - I've just picked up Ptolemy 0.5, so please excuse me if this is
redundant.  I did not find comments to this effect in the archive.

I was able to build it using gcc 2.4.5 & 2.5.8 with three small problems:

        1) src/domains/sf/matrix/stars/SDFDataCx_M.pl

                Changed input%k to Complex(input%k)

        2) src/domains/de/kernel/CalendarQueue.h

                Changed HUGE_VAL to MAXDOUBLE

        3) src/octtools/makefile

                Changed the load to use the misalign libraries.  There
                were a number of cases where I got misaligned core dumps
                when running on sparc. The load must pickup the
                crt0.o and libc from /lib/misalign.

When trying to build Ptolemy 0.5 using gcc 2.6.0, I received numerious type
clash errors (char * versus const char *).

I hope this is of some interest.

Terry Domae

 
 
 

Some Comments on Building Ptolemy0.5 on Sparc SunOS 4.1.X (gcc)

Post by Christopher Hylan » Thu, 25 Aug 1994 07:59:02


Thanks for your bug report, I've answered the points below

Quote:>    Hi All - I've just picked up Ptolemy 0.5, so please excuse me if this is
>    redundant.  I did not find comments to this effect in the archive.

>    I was able to build it using gcc 2.4.5 & 2.5.8 with three small problems:

>            1) src/domains/sf/matrix/stars/SDFDataCx_M.pl

>                    Changed input%k to Complex(input%k)

I fixed this in the development tree and the release tree, so the
change should be in the 0.5.1 release
* sdf/matrix/stars/SDFDataCx_M.pl:
        Added Complex() to: Changed input%k to Complex(input%k)
        SDFDataFix_M.pl has a similar cast.

Quote:>            2) src/domains/de/kernel/CalendarQueue.h

>                    Changed HUGE_VAL to MAXDOUBLE

I'm not sure why the change here is necessary.  The code seems to work
as is.  Can you provide me with more details?  I don't have a full
understanding of HUGE_VAL, but it seems to be present on all the
machines we are compiling under.

Quote:>            3) src/octtools/makefile

>                    Changed the load to use the misalign libraries.  There
>                    were a number of cases where I got misaligned core dumps
>                    when running on sparc. The load must pickup the
>                    crt0.o and libc from /lib/misalign.

This is fixed in the upcoming release.  Basically, the fix is to not
use the octtools memory management library -lmm.  There was also a
problem with the octtools kd library and realloc.  Both of these
problems have been fixed in the upcoming 0.5.1 release.  In 0.5,
$PTOLEMY/src/octtools/README discusses the situation a little.

Quote:>    When trying to build Ptolemy 0.5 using gcc 2.6.0, I received
>    numerious type clash errors (char * versus const char *).

We've cleaned up some of these warnings, but pigilib still has them.
Also, since we are using gcc -Wall rather than cc to compile .c files,
we get lots of warnings in pxgraph, thor and octtools.

Quote:>    I hope this is of some interest.

You bet!  We're always interested in bug reports.    

Thanks,
        Christopher

 
 
 

Some Comments on Building Ptolemy0.5 on Sparc SunOS 4.1.X (gcc)

Post by Terry Dom » Thu, 25 Aug 1994 23:23:42


Christopher -

With regards to the HUGE_VAL modification I made:

Quote:> >       2) src/domains/de/kernel/CalendarQueue.h

> >               Changed HUGE_VAL to MAXDOUBLE

> I'm not sure why the change here is necessary.  The code seems to work
> as is.  Can you provide me with more details?  I don't have a full
> understanding of HUGE_VAL, but it seems to be present on all the
> machines we are compiling under.

Under the versions of g++ that I have been using, the file g++-include/math.h
includes a definition for HUGE, but not HUGE_VAL.  In g++ HUGE is defined as:

        #define HUGE MAXDOUBLE

Although on many implementations this can be defined as:

        #define HUGE (_infinity())

And, I think this exact definition was included in the 0.5pl2 (pl3?) patch and
#ifdef'd as being an SGI only fix.  Since I'm running on a Sun Sparc,
it ignores this definition.

Under gcc or cc, the definition for HUGE_VAL appears in #include <math.h> (both
in /usr/include and in the gcc include tree).  However, these files are not
used for g++.

The only explaination I can think of is that I'm running a vanilla gcc/g++ with
libg++, and that someone at Berkeley modified the math.h file for g++ to
include the HUGE_VAL definition.  Or, possibly, I have not been installing
gcc/g++ correctly somehow.

Again, sorry for the limited context I provided in my original message.  I hope
this level detail is sufficient.

Terry...

 
 
 

1. Help: Build MIT Kerberos V5 on sunos 4.1.x machine

Hi Folks,

I try to build kerberos V5 on my sunos 4.1.4 machine.
I follow the HOW_TO_BUILD instruction.  However, I
have following error message when I compile the program:

-----------  capture --------

-DKERBEROS  -DKRB_RLOGIN=\"/rlogin\"   -DLOGIN_PROGRAM=\"/krb5/sbin/login.krb5\" -DKPROGDIR=\"/krb5/bin\"  -target sun4 -c  krlogin.c
/usr/include/termios.h: 109: NL0 redefined
/usr/include/termios.h: 110: NL1 redefined
/usr/include/termios.h: 112: CR0 redefined
/usr/include/termios.h: 113: CR1 redefined
/usr/include/termios.h: 114: CR2 redefined
/usr/include/termios.h: 115: CR3 redefined
/usr/include/termios.h: 117: TAB0 redefined
/usr/include/termios.h: 118: TAB1 redefined
/usr/include/termios.h: 119: TAB2 redefined
/usr/include/termios.h: 120: XTABS redefined
/usr/include/termios.h: 123: BS0 redefined
/usr/include/termios.h: 124: BS1 redefined
/usr/include/termios.h: 129: FF0 redefined
/usr/include/termios.h: 130: FF1 redefined
/usr/include/termios.h: 162: ECHO redefined
/usr/include/termios.h: 166: NOFLSH redefined
/usr/include/termios.h: 167: TOSTOP redefined
/usr/include/termios.h: 182: FLUSHO redefined
/usr/include/termios.h: 183: PENDIN redefined
"/usr/include/sys/tty.h", line 16: syntax error at or near variable name "queue_t"
"/usr/include/sys/ptyvar.h", line 16: syntax error at or near variable name "mblk_t"
*** Error code 1
make: Fatal error: Command failed for target `krlogin.o'
Current working directory /opt/kerberos/src/appl/bsd
*** Error code 1
make: Fatal error: Command failed for target `all'
Current working directory /opt/kerberos/src/appl
*** Error code 1
make: Fatal error: Command failed for target `all'
------------ end -------------

Have anyone able to build the MIT kerberos V5 on a sunos 4.1.x
machine?  I need you help here.  

Thanks in advance.

Howard

--
Howard Yeung  - Unix System Admin,
National Library of Medicine/NIH
Tel:   (301)-402-1742

2. PA EXPO 2000 Call for Participation

3. Building Kerberos for SunOS 4.1.x - Help!

4. BOOT.INI parameters

5. lcc 4.1 build failure for sparc/solaris: 'make triple' stage

6. FS: SPARCSystem LX Desktop Workstation

7. Building lcc 4.1 with gcc on Redhat 7.2

8. Thunderclock ProCLOCK info, please

9. Distiller for SunOS 4.1.x

10. HP DAT drives on SunOS 4.1.x

11. SunOS 4.1 & vi

12. SNMP daemon for SunOS 4.1

13. So much for kerberos in SunOS 4.1 (inside the US)