2.1.0 build glitches on Solaris 8

[David Mathog]

I just tried installing 2.1.0 on Solaris 8.  After unpacking it in the
top level did:

# ./configure --with-png
# make

and it blew up in the libtool link of plplot complaining about not
finding "yes/lib".
Turns out plplot/Makefile had 

CPPFLAGS =  -Iyes/include -Iyes
LDFLAGS =  -Lyes/lib

which is seriously wrong.  Tried again with gmake and had the same
problem.  Oops,
just realized that make clean doesn't remove Makefile.  Take them all
out with:

# find . -name Makefile -exec rm {} \;
# ./configure --with-png

and this time plplot/Makefile  shows:
CPPFLAGS =  -I/usr/include -I/usr
LDFLAGS =  -L/usr/lib

and now make for plplot works.    Bizarre - some sort of first run
initialization problem exists. Unfortunately all is still not working
right.. When it hits ajacd.o it blows up with:

ajacd.c: In function `acdBadVal':
ajacd.c:2827: `__builtin_va_alist' undeclared (first use in this
function)
ajacd.c:2827: (Each undeclared identifier is reported only once
ajacd.c:2827: for each function it appears in.)
ajacd.c: In function `acdLog':
ajacd.c:13323: `__builtin_va_alist' undeclared (first use in this
function)
ajacd.c: In function `acdPretty':
ajacd.c:13359: `__builtin_va_alist' undeclared (first use in this
function)
*** Error code 1
make: Fatal error: Command failed for target `ajacd.lo'
Current working directory /usr/local/src/EMBOSS-2.1.0/ajax
*** Error code 1
make: Fatal error: Command failed for target `all-recursive'

This was using gcc version 2.95.3 20010315 (release).   I've seen
problems with
that builtin before, but usually it's the other way around - conflicting
definitions when
gcc uses certain includes.  Here it looks like those includes were left
out.

Try once more using the native (Forte) compiler (which is how I built
the 1.12.0 version)

# make clean
# find . -name Makefile -exec rm {} \;
# export CC=cc
# echo $CC
cc
# ./configure --with-png

RATS - it rejects the png stuff this time with:
checking if png driver is wanted... yes
checking for inflateEnd in -lz... yes
checking for png_destroy_read_struct in -lpng... no
No png driver will be made due to librarys missing/old.

continue on anyway to see if this resolves the builtin problem...
Yes, apparently using the Forte compiler takes care of the va_builtin
problem - at the cost of losing the PNG device.  The Forte compiler is:

Sun WorkShop 6 update 1 C 5.2 2000/09/11

Somebody must have this working on Solaris.  What is the magic
incantation to
get a full build with PNG enabled???

Thanks,

David Mathog
mathog at caltech.edu
Manager, Sequence Analysis Facility, Biology Division, Caltech