Hi all,

Long story... gcc aix dxexec can work. need -bbigtoc flag to ld.  need to skip
optimization on certain files.  if interested, read on...

Bill hinted at gcc problems the other day, but I thought what the hey, so I 
kicked off
a gcc 2.95.2 /aix build before I left the other day.  The exercise refreshed my
memory.  I couldn't even build dxexec... ld gave a TOC overflow (and no 
dxexec).  I
looked around newsgroups via deja.com and found the claim that I need to 
include the
flag -bbigtoc for ld to finish.  OK, now I have a dxexec from gcc, and I 
experience the
displayx.c line 4775 i=256 segfault.

Poking around a little more, it looks as if the gcc I am using needs to be built
differently.  One of my colleagues here at Watson has a read-the-directions 
type post
pointing out that building gcc according to the directions with "make boostrap"
addresses the TOC issue.  I'll drop the folks who own those filesystems a 
nastygram
(kidding!) and see if we have gcc built and installed correctly.  Maybe others 
are in
this same boat.

BUT back on the subject... a segfault on 4775 displayx.c is improbable and the 
code
isn't even remotely suspicious... even neighboring lines are a stretch.  I'll 
claim
this is an optimization error.  So I compiled displayx.c with gcc but without
optimization and relinked libLIBDX.a then dxexec.  No displayx.c segfault.  
dxexec
runs, but I am not confident that there aren't other bad optimizations lurking 
in
dxexec from this version
of gcc.  Through gcc -v junk.c I find I'm actually running "GNU C version 2.95.3
19991030 (prerelease) (powerpc-ibm-aix4.3.2.0) ".  I assume from prior posts 
that you
(Randy) are at 2.95.2 from 10/27/99.  Can anyone (Dave?) with a prior version 
of gcc on
aix build optimized without this bug, and thereby help narrow this down?

Regards,
Pete

Peter Daniel Kirchner wrote:

> Randy-
>
> I can't repeat your server disconnect on aix 4.3.2, current(2/15), and 8 bit
> display built with xlc.  With 1/26-ish source, image window #2 did get very 
> few
> colors with software rendering, which must be a now-fixed bug, just not the
> coredump variety.  hardware rendering --in this case, IBM's software opengl 
> to a 2d
> card-- works fine for 1/26 and current.
>
> btw
> I wouldn't recommend replacing gcc with xlc via editing makefiles.  Rather, 
> when I
> "configure" dx on aix I use:
>
> CC=xlc
> CXX=xlC
> LDFLAGS= -L /usr/local/lib
> CPPFLAGS= -I /usr/local/include
> and I have an alias for configure which resolves to
> ./configure --with-cosmojar-path=/homes/ppc4/daniel/classes/npcosmop211.jar
> --with-java40jar_path=/homes/ppc4/daniel/classes/java40.jar
>
> If (per Dave's explanation) you are "stuck" at aix 4.2, do you have some 
> standard
> set of ptf's that you put on your systems?
>
> Regards,
> Pete
>
> Randy Polson wrote:
>
> > One additional comment:
> >
> > I tried running dx on one machine, and displaying on another.  Disconnect
> > from server at the same line 4775 in displayx.c with cmapsize=i=256.
> >
> > On Mon, 14 Feb 2000, David L. Thompson wrote:
> >
> > > A couple of questions. 1) Are you on an 8 bit display system? Which
> > > version are you compiling (cvs current)?
> > >
> > > David
> > >
> > > >Hello-
> > > >First, before I forget, thanks Pete and Daivd for you tips for getting 
> > > >the
> > > >java part to compile.  I changed the line in /src/uipp/java/Makefile from
> > > >cc=gcc to cc=xlc an everything build OK.  I changed the jni_mdh.h file,
> > > >and next build I'll tell
> > > >you how it goes.
> > > >
> > > >Next, I am having difficulties displaying multiple images within the same
> > > >net for AIX 4.2.  For an example, I chose AutoColor2d from the tutorial.
> > > >It runs just
> > > >fine.  If I add a second image module, it disconnects from server.  The
> > > >output from dbx is below.
> > > >
> > > >-Randy
> > > >
> > > >Segmentation fault in getOneMapTranslation at line 4775 in file
> > > >"/u/res3/rpolson
> > > >/OpenDX/dx-4.0.10/dx/src/exec/libdx//displayx.c"
> > > >  4775       for (i = 0; i < cmapsize; i++)
> > > >(dbx) where
> > > >getOneMapTranslation(dpy = 0x2010c328, d = 0x206126b0, force = 0), line
> > > >4775 in
> > > >"displayx.c"
> > > >createTranslation(couldn't resolve reference
> > > >(dbx)
> > > >
> > > >(dbx) print cmapsize
> > > >256
> > > >(dbx) print i
> > > >256
> > > >(dbx)
> > >
> > > --
> > > .............................................................................
> > > David L. Thompson                          The University of Montana
> > > mailto:[EMAIL PROTECTED]                 Computer Science Department
> > > http://www.cs.umt.edu/u/dthompsn           Missoula, MT  59812
> > >                                             Work Phone : (406)257-8530
> > >

Reply via email to