[Gcl-devel] Bugs in gcl cause maxima build failures

[James Amundson]

Hi,

Since Ray added his slatec updates to Maxima, I have been unable to get
Maxima to build with GCL because of bugs in GCL itself. Of course, I
would like to see these bugs found and squashed in GCL, but I would also
like to find workarounds because I do not want Maxima to rely on the
bleeding edge of GCL.

I will apologize in advance: these are not good bug reports. I am
sending them along in hopes they are useful. I will try to find the time
to track down the bugs in detail later.

Versions: 

Maxima: cvs as the time I write this message.
GCL: Any recent version. I tried current cvs 2.5.0 as of yesterday, cvs
2.5.0 as of March 26, and 2.4.1, all with the same result.

Bug #1:

GCL bails out with an error when trying to compile f2cl-lib.lisp in the
src/numerical directory. The simplest way to see the bug is simply to
run gcl and type
        (compile-file "f2cl-lib.lisp")

Here is what happens:

---------------------------------------
>(compile-file "f2cl-lib.lisp")

Compiling f2cl-lib.lisp.
; (DEFUN COL-MAJOR-INDEX ...) is being compiled.
;; The variable HI is undefined.
;; The compiler will assume this variable is a global.
;; The variable LO is undefined.
;; The compiler will assume this variable is a global.
; (DEFUN IFIX ...) is being compiled.
;; Warning: INT (in lisp package) is called as a function--not yet
defined
; (DEFUN IDFIX ...) is being compiled.
;; Warning: INT (in lisp package) is called as a function--not yet
defined
; (DEFUN MAX1 ...) is being compiled.
;; Warning: INT (in lisp package) is called as a function--not yet
defined
Warning: The declaration specifier INTEGER4 is unknown.
; (DEFUN PROCESS-IMPLIED-DO ...) is being compiled.
;; The variable START is undefined.
;; The compiler will assume this variable is a global.
;; The variable END is undefined.
;; The compiler will assume this variable is a global.
;; The variable &OPTIONAL is undefined.
;; The compiler will assume this variable is a global.
;; The variable STEP is undefined.
;; The compiler will assume this variable is a global.
;; The variable INDEX-VAR is undefined.
;; The compiler will assume this variable is a global.
;; The variable DATA-VARS is undefined.
;; The compiler will assume this variable is a global.
; (DEFUN EXECUTE-FORMAT ...) is being compiled.
;; The variable WHILE is undefined.
;; The compiler will assume this variable is a global.
;; The variable DO is undefined.
;; The compiler will assume this variable is a global.
; (DEFUN D1MACH ...) is being compiled.
;; The variable LEAST-POSITIVE-NORMALIZED-DOUBLE-FLOAT is undefined.
;; The compiler will assume this variable is a global.
; (DEFUN R1MACH ...) is being compiled.
;; The variable LEAST-POSITIVE-NORMALIZED-SINGLE-FLOAT is undefined.
;; The compiler will assume this variable is a global.
End of Pass 1.  
; (DEFUN LUN->STREAM ...) is being compiled.
;; Warning:  LUN->STREAM is proclaimed but not in *inline-functions* 
T1defun could not assure suitability of args for C call
End of Pass 2.  
f2cl-lib.c: In function `L65':
f2cl-lib.c:4460: `GEN' undeclared (first use in this function)
f2cl-lib.c:4460: (Each undeclared identifier is reported only once
f2cl-lib.c:4460: for each function it appears in.)
f2cl-lib.c:4460: parse error before `V154'
f2cl-lib.c:4461: parse error before `V155'
f2cl-lib.c:4463: `V154' undeclared (first use in this function)
f2cl-lib.c:4463: `V154alloc' undeclared (first use in this function)
f2cl-lib.c:4464: `V155' undeclared (first use in this function)
f2cl-lib.c:4464: `V155alloc' undeclared (first use in this function)

Correctable error: (SYSTEM "(cd . ;gcc -pipe -fwritable-strings 
-DVOL=volatile -I/home/amundson/devel/gcl/o -fsigned-char  -O -c -I.
f2cl-lib.c -w)") returned a non-zero value 0.
Signalled by UNLESS.
If continued: Continues anyway.
Broken at CERROR.  Type :H for Help.
F2CL-LIB>>

---------------------------------------

It looks to me like some preprocessor macros are not properly defined in
the resulting C file. Beyond that, I don't know what is going on.

I can work around this problem by simply loading the file instead of
compiling it. However, I am then hit with

Bug #2:

GCL fails when trying to load the compiled version of xersve.lisp in the
src/numerical/slatec directory. The simplest way to see the bug is to
follow the session below:

----------------------------------------
|addiator>gcl
GCL (GNU Common Lisp)  Version(2.5.0) Tue Apr 30 16:40:55 CDT 2002
Licensed under GNU Library General Public License
Contains Enhancements by W. Schelter

>(load "../f2cl-package.lisp")

Loading ../f2cl-package.lisp
Finished loading ../f2cl-package.lisp
T

>(load "../f2cl-lib.lisp")

Loading ../f2cl-lib.lisp
Finished loading ../f2cl-lib.lisp
T

>(compile-file "xersve.lisp")

Compiling xersve.lisp.
End of Pass 1.  
End of Pass 2.  
OPTIMIZE levels: Safety=0 (No runtime error checking), Space=0, Speed=3
Finished compiling xersve.lisp.
#p"xersve.o"

>(load "xersve")

Loading xersve.o

Error: Cannot expand the SETF form (F2CL-LIB::ROW-MAJOR-AREF #:G1082
F2CL-LIB::K).
Fast links are on: do (si::use-fast-links nil) for debugging
Error signalled by SETF.
Broken at SYSTEM::SETF-HELPER.  Type :H for Help.
SLATEC>>
---------------------------------------

I don't have a workaround for this problem. I am stuck.

I would be happy to send the individual files in private email if it
would help.

Best,
Jim