Re: [Gcl-devel] HEAD Maxima and HEAD trad GCL

[Vadim V. Zhytnikov]

Camm Maguire ?????:
> Greetings!
>
> James Amundson <address@hidden> writes:
>
>
> > On Thu, 2004-01-08 at 13:08, Camm Maguire wrote:
> >
> > > Greetings!
> > >
> > > "Vadim V. Zhytnikov" <address@hidden> writes:
> >
> > > > Yes, but my resolution is not of kind I like.
> > > > Replace Maxima's defsystem.lisp by older version -
> > > > beginning of May 2003 or earlier.
> > >
> > > Vadim, could you please have mercy on me and explain what confluence
> > > of the newer defsystem and windows gcl is causing all this?  I must
> > > have missed it in your earlier posts.
> >
> > If there are real problems with the gcl 2.6.2 ansi build on windows and
> > the only thing that is keeping traditional gcl 2.6.2 on windows from
> > working with Maxima is defsystem.lisp, then we can *certainly* fix
> > defsystem.lisp. It is my stated desire to move Maxima to the ansi gcl
> > build as soon as possible. That does not mean I want to create an
> > artificial stumbling block over what looks like a very small problem.
> >
> > The only changes in defsystem.lisp are various portability conditionals.
> > If the May 2003 version works (I think Vadim means revision 1.9. If not,
> > he should say so), then the changes required to get the current
> > defsystem.lisp to work should be trivial.
>
>
> Thanks James!  Its good to be able to fall back on this.  I still
> would like an explanation of the problem if possible -- doesn't sound
> too difficult to fix.  Am too busy right now to diagnose it myself.
>
> Take care,

Sorry for prolonged silence (every January is a hard time to me
with respect to spare time).

Something is wrong with GCL + mingw + Maxima CVS.
Maxima GCL build stops with the following error message
===================================================
Source file 
"C:/msys/1.0/home/vadim/maxima/src/C:/msys/1.0/home/vadim/maxima/src/j/numerical/slatec/dbesi0.lisp" 
and binary file "binary-gcl/numerical/slatec/dbesi0.o" do not exist.
===================================================
See bogus /j/ in .../maxima/src/j/numerical/... source path.
This symbol may vary - it may be /l/ or /5/ or something.
Behavior is 100% reproducible with any particular GCL build
but changes if I rebuild GCL with some modifications.
In this case make will stop at other source file.
Further behavior may be also different.  You may get
succesfull Maxima build after 5-6 repetitive make
invocation (each time make compiles some more
10-15 source files and stops with similar error message
at another place) or get into loop since make stops
at the same source file.

The only "resolution" I have - replace Maxima defsystem
by older version. Then Maxima build goes flawlessly.
I'd like to know who is the real troublemaker.
Maybe we hit some subtle GCL+mingw bug.





--
     Vadim V. Zhytnikov

      <address@hidden>
     <address@hidden>