[Gcl-devel] Re: [Maxima] GCL

[Camm Maguire]

Greetings!

"Vadim V. Zhytnikov" <address@hidden> writes:

> Camm Maguire wrote:
> 
> > Greetings!
> >
> > "Vadim V. Zhytnikov" <address@hidden> writes:
> >
> > > This is a multi-part message in MIME format.
> > > --------------9F7C67FF5C3294CE4BC3862C
> > > Content-Type: text/plain; charset=koi8-r
> > > Content-Transfer-Encoding: 7bit
> > >
> > >
> > >
> > > Camm Maguire wrote:
> > >
> > > > Greetings!
> > > >
> > > > 1)  What are the symptoms/problems against new glibc?  Seg faults
> > > >     perhaps?  If so my suspicion is that it is due to a fix suggested
> > > >     by Dr. Schelter to me before he passed away which for some reason
> > > >     he never put in cvs.  I hope to have a new commit ready in a few
> > > >     days with this fix in.  I know its needed for a working Debian
> > > >     gcl/maxima.
> > >
> > > My problem with gcl compilation is exactly the same as bug #108909
> > > for Debian gcl package (my error log is attached).
> >
> > OK, I've scoped this out, but the report is truncated.  No configure
> > output, etc.  May not be important.  See below.  Thanks for your
> > attached log.
> >
> > > It is claimed that the problem is fixed in latest gcl 2.4.0-3 Debian 
> > > build.
> >
> > I agree that removing Tk has nothing to do with this.
> >
> > > But my attempt to apply Debian's patch has failed.
> > > Actually I wonder whether the problem is truly fixed in Debian or not.
> >
> > I have not been able to confirm, but I believe the current state of
> > the Debian package has a few problems.  I'll be putting together a new
> > one.
> >
> > > The important point is that I'm not 100% sure that the problem is really
> > > due to glibc 2.2.4. It showed up first time after I made massive
> > > Linux upgrade (including glibc). I'm quite confident that it is
> > > not due to kernel and compiler. At present I have two Linux installation
> > > on one machine - first with glibc 2.2.3 A and second with glibc 2.2.4 B.
> > > On A I can build gcl successfully with both gcc 2.95 and 2.96.
> > > On B compilation fails with the same diagnostics independent on
> > > compiler version. Binaries built on A run on B without any trouble!
> > >
> >
> 
> Fine, I'll try you recommendation and report results to you.
> 

Great!  I'll be very interested.

> >
> > OK, here is a gut guess:
> >
> > 1) try linking statically.  Find the appropiate line in 386-linux.h or
> >    .defs, and make that the default.  If this clears up the problem,
> >    then try this with synamic linking:
> > --- ../gcl-2.4.0/h/386-linux.h  Wed May  2 16:28:58 2001
> > +++ h/386-linux.h       Wed Dec 12 21:58:29 2001
> > @@ -243,4 +243,17 @@
> >   */
> >
> >  /* End for cmpinclude */
> > +/* define if there is no _cleanup,   do here what needs
> > +   to be done before calling unexec
> > +   */
> > +#define CLEANUP_CODE \
> > +  setbuf(stdin,0); \
> > +   setbuf(stdout,0);
> > +
> > +/* the following will try to make sure that the stdin and stdout are
> > +set correctly on startup after a save */
> > +
> > +
> > +#define INIT_CORE_END terminal_io->sm.sm_object0->sm.sm_fp = 
> > stdin;terminal_io->sm.sm_object1->sm.sm_fp = stdout;
> > +
> >
> >
> 
> <snip>
> 
> > > >
> > > > 5) Please consider subscribing to the new list address@hidden  You
> > > >    can do this via the web at
> > > >    http://mail.gnu.org/mailman/listinfo/gcl-devel
> > > >
> > > > 6) It would be wonderful if you could register at
> > > >    http://savannah.gnu.org and volunteer as a gcl developer.  Your
> > > >    patches/bug reports/testing results could be posted there for
> > > >    central reference.  At the very least, you would serve a valuable
> > > >    purpose as a newest-glibc tester.  We could discuss cvs write
> > > >    access if you are interested.
> > > >
> > >
> > > I'll be glad to join the project but probably just as tester. First of all
> >
> > Testing is very valuable!  Please let me know what source savannah
> > user id is, and I'll enable you for bug/taks submissions/reporting.
> >
> > > I'm not so good in C and I feel that I can help Maxima more as Lisp
> > > coder. On the other hand I really want to help maintaining gcl.
> >
> > I very happy you feel comfortable in Lisp, as this is exactly where I
> > am weakest.  In fact, one job I posted on the site is for a
> > knowledgeable Lisp person to describe missing ANSI features from GCL,
> > and perhaps suggest a prioritization for adding new features.  This
> > and testing/port effort reporting are of great help.
> >
> 
> I've registered at Savannah as vvzhy and subscribed to the gcl-devel
> mailing list. As for ANSI CL compliance I'm sorry to disappoint you a bit :-(

OK, I've added you to the list of developers on savannah.  How about I
create a task there for the new glibc testing of both the static link
and the patch I sent you, and assign this to you?


> I have some quite good experience with other Lisp dialect which is _very_
> different from CL. I've started learning CL specific just two months ago
> so at present I'm not so good as CL compliance expert. On the other
> hand it seems to me that you should be cautious in making GCL
> Common List compliant. Adding to CL features which are missing
> in GCL is always good but there are other points where GCL works
> but works not as expected by CL standard. Consider for example
> this renown problem with APPLY. In CL standard (as far as I understand)
> the construction
>   (APPLY '(LAMBDA (U) ...) '(..))
> is not valid and this is so in clisp and cmucl. But it works on GCL and
> I do not know what shall we do this it? If we make this construction
> forbidden in GCL then immediately Maxima will stop work with it.
> 

Agreed that we should not break maxima at all costs!  But how does
maxima run then on clisp?

> Just some thoughts,
> 

For which, many thanks!

> Vadim
> 
> --
> 
> [ Vadim V. Zhytnikov  <address@hidden>  <address@hidden> ]
> 
> 
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah