On Sun, Jul 24, 2022 at 08:22:21PM +0000, Waldek Hebisch wrote:
> On Sun, Jul 24, 2022 at 02:27:15PM +0800, Qian Yun wrote:
> >
> >
> > On 7/24/22 10:41, Waldek Hebisch wrote:
> > >
> > >I have re-tried GCL 2.6.13_pre on Linux. I see similar error.
> > >Running offending part by hand with
> > >
> > >)lisp (si::use-fast-links nil)
> > >)set break break
> > >
> > >I got more sensible error message. Apparrently GCL is running
> > >out of memory to load code. But this is strange: GCL has allocated
> > >44G heap and there is 2.9G resident, so there should be plenty
> > >of memory. And at that stage there is about 4.5 M od code
> > >to load. There is also previously compiled code in the same image,
> > >but this should not be very big.
> > >in
> >
> > Err, did you finish this email?
> >
> > I did some lookup in this direction, and found Environment Variable
> > GCL_MEM_MULTIPLE.
> >
> > It means how many memory GCL will use. However, a smaller
> > value (0.1) can make compile go further, while bigger value (0.5) can't.
> >
> > (Of course, the effect of this variable is dependent on system memory,
> > so YMMV.)
> >
> > With GCL_MEM_MULTIPLE=0.1, now build fails at:
> >
> > ======
> > U64Int is already explicitly exposed in frame initial
> > U64Int will be automatically loaded when needed from
> > /tmp/fricas/src/algebra/U64INT.NRLIB/U64INT
> >
> > )lisp (make-databases nil nil)
> >
> >
> > "building operation.daase"
> > "building category.daase"
> > >> System error:
> >
> >
> > (1) -> mv: cannot stat 'category.daase': No such file or directory
> > mv: cannot stat 'compress.daase': No such file or directory
> > ======
> >
>
> I had to go down to GCL_MEM_MULTIPLE=0.07 (and GCL_MEM_MULTIPLE=0.02
> works the same) to get this. AFAICS the error is becuse interpsys
> is unable to evaluate BasicType. The trouble is that DATABASE property
> on BasicType is NIL. But it should be set by compilation. Apparently,
> running LOCALDATABASE by hand sets DATABASE property and compilation
> calls LOCALDATABASE, so it should be set. But it is not...
ATM only one patch is needed to build using 2.6.13pre and FriCAS
from distribution tarball:
--- ../trunk/src/lisp/fricas-lisp.lisp 2022-07-23 20:46:47.697097794 +0200
+++ ../dist/src/lisp/fricas-lisp.lisp 2022-07-25 17:25:07.328035488 +0200
@@ -986,7 +986,7 @@
(let ((ns (namestring name)))
(if (and (consp (pathname-directory name))
(eq (car (pathname-directory name))
- #-:GCL :absolute #+:GCL :root))
+ :absolute ))
ns
(concatenate 'string (get-current-directory) "/" ns))))
#+:cmu
This patch is incompatible with earlier versions of GCL, but I think
I have now way to do the same in backwards compatible way.
OTOH testsuite has several failures.
--
Waldek Hebisch
--
You received this message because you are subscribed to the Google Groups
"FriCAS - computer algebra system" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/fricas-devel/20220725162109.GA23053%40fricas.math.uni.wroc.pl.
