[Ecls-list] building maxima

[Raymond Toy]

On 1/29/12 4:06 PM, Paul Bowyer wrote:
> I'm running PCLinuxOS, 32-bit, on a Pentium-4 system and I tried to 
> build maxima from source using ECL 11.1.1 and it fails during make.
>
> The configuration I used for building ECL was:
> ./configure --prefix=/usr/local/ecl32 -enable-threads=yes 
> --with-__thread=no --enable-boehm=included --with-gmp=included 
> --with-dffi=included --with-clx=builtin --enable-rpath=yes
>
> ECL completed its tests with:
> ANSI,  31 out of 21713 total tests failed:
> REGRESSIONS, 7 out of 61 total tests failed:
>
> The configuration I used for trying to build maxima was:
> export CFLAGS="-O2 -g -march=i386 -mcpu=i686 -fno-fast-math" 
> CXXFLAGS="-O2 -g -march=i386 -mcpu=i686 -fno-fast-math" && ./configure 
> --prefix=$HOME/MaximaEclTest --enable-ecl 
> --with-ecl=/usr/local/ecl32/bin/ecl
>
> I used a similar configuration to build maxima from source using sbcl, 
> cmucl, gcl, and ccl, which completed successfully and only ccl failed 
> the maxima testsuite.
>
> The output from the make process at the point of failure was:
> ;;; Emitting code for SLATEC::XERPRN.
> ;;; Finished compiling 
> /home/pfb/my-maxima-git-ecl-sandbox/maxima/src/numerical/slatec/xerprn.lisp.
> ;;;
>
> ;        - Loading binary file "binary-ecl/numerical/slatec/xerprn.fas"
> ;;; Loading 
> "/home/pfb/my-maxima-git-ecl-sandbox/maxima/src/binary-ecl/numerical/slatec/xerprn.fas"
> An error occurred during initialization:
> "$$" is not of type (SIMPLE-ARRAY CHARACTER (2))..
>
This is a bug in the translation of Fortran to Lisp.  F2cl should have
declared that as a string, not a simple-array of characters.  For now,
you can replace the declaration with string.  F2cl needs to be fixed so
that the translation is correct.

It seems that, unlike most other lisps, ecl reads in "$$" and figures
out that it can a simple-string and makes it so.  Most other lisps don't
try and just create a string (unicode).

Ray