[Gcl-devel] Re: your mail

[Camm Maguire]

Greetings!

Tuukka Toivonen <address@hidden> writes:

> On Thu, 20 Dec 2001, Tuukka Toivonen wrote:
> 
> >Maxima initialization routine. In fact, I think I have already made a
> >patch for that. I'll dig it up for you, tomorrow.
> 
> I couldn't find anything, probably I haven't done it. However, I found
> some other fixes or workarounds for GCL, that caused me trouble when
> compiling gcl-2.4.0 (and older) with glibc-2.0.6 and sed-3.02:
> (I don't know if you're interested supporting such an old libc, but
> since the fixes are small, I think it would be a good idea)
> 

Thanks for these!  See notes below.  Would you consider registering at
savannah, and allow me to set you up as a gcl developer?  Patches such
as these could then be submitted there (as well as here, if desired)
for central reference in case someone else might wish to work on it.
If so, please let me know what your savannah id is/will be.


> (1) ---------------------------------------------------------------
> 
> gcc -pipe -fwritable-strings  -DVOL=volatile -I/cdimage/gcl-2.4.0/o
> -fsigned-char -c -O  -I../gcl-tk -I../h/ -O4   sfasl.c
> In file included from ../h/ext_sym.h:7,
>                  from sfaslelf.c:23,
>                  from sfasl.c:40:
> /usr/include/elf.h:42: warning: `mode' attribute ignored
> /usr/include/elf.h:43: warning: `mode' attribute ignored
> /usr/include/elf.h:44: warning: `mode' attribute ignored
> /usr/include/elf.h:45: warning: `mode' attribute ignored
> /usr/include/elf.h:49: warning: `mode' attribute ignored
> /usr/include/elf.h:53: warning: `mode' attribute ignored
> /usr/include/elf.h:61: warning: `mode' attribute ignored
> sfaslelf.c: In function `relocate':
> In file included from sfasl.c:40:
> sfaslelf.c:433: `R_386_NUM' undeclared (first use in this function)
> sfaslelf.c:433: (Each undeclared identifier is reported only once
> sfaslelf.c:433: for each function it appears in.)
> make[1]: *** [sfasl.o] Error 1
> make[1]: Leaving directory `/cdimage/gcl-2.4.0/o'
> make: *** [all] Error 2
> 
> FIX (maybe #ifdef the other constants too):
> 

OK, applied.  In 2.5.0, this stuff will hopefully be handled by the
bfd library, so I have not applied there.


> diff -PruN gcl-2.4.0-orig/o/sfaslelf.c gcl-2.4.0/o/sfaslelf.c
> --- gcl-2.4.0-orig/o/sfaslelf.c Mon Apr 16 05:04:38 2001
> +++ gcl-2.4.0/o/sfaslelf.c      Sun Jul 22 13:20:50 2001
> @@ -430,9 +430,11 @@
>        FEerror("Unsupported ELF type R_386_GOTPC");
>        break;
> 
> +#ifdef R_386_NUM
>      case     R_386_NUM:
>        FEerror("Unsupported ELF type R_386_NUM");
>        break;
> +#endif
> 
>  #else
>      case     R_SPARC_WDISP30:
> 
> (2) ---------------------------------------------------------------
> 
> make[2]: Entering directory `/cdimage/gcl-2.4.0/elisp'
> if test -d "./configure: emacs: command not found" ; then cp *.el
> ./configure: emacs: command not found ; fi
> if test -f "./configure: emacs: command not found/default.el" ; then \
> cat ./configure: emacs: command not found/default.el | sed -e '/BEGIN
> gcl/,/END gcl/d' > xxx ; \
> mv ./configure: emacs: command not found/default.el ./configure: emacs:
> command not found/default.el.prev ; \
>   rm -f  ./configure: emacs: command not found/default.elc ; \
>           cat add-default.el >> xxx ; cp  xxx ./configure: emacs: command
> not found/default.el ; \
>   rm -f xxx ; else \
> cp  add-default.el ./configure: emacs: command not found/default.el ; fi
> cp: copying multiple files, but last argument (found/default.el) is not a
> directory
> Try `cp --help' for more information.
> make[2]: *** [install] Error 1
> make[2]: Leaving directory `/cdimage/gcl-2.4.0/elisp'
> make[1]: *** [install1] Error 2
> make[1]: Leaving directory `/cdimage/gcl-2.4.0'
> make: *** [install] Error 2
> 
> (I don't have emacs installed) WORKAROUND:
> 

I did something slightly different, but I think the install should
work without emacs being present.  Please check it out!


> diff -PruN gcl-2.4.0-orig/elisp/makefile gcl-2.4.0/elisp/makefile
> --- gcl-2.4.0-orig/elisp/makefile       Sun May  6 09:18:49 2001
> +++ gcl-2.4.0/elisp/makefile    Sun Jul 22 13:20:50 2001
> @@ -111,13 +111,13 @@
> 
>  install:
>         if test -d "${EMACS_SITE_LISP}" ; then cp *.el ${EMACS_SITE_LISP} ;
> fi
> -       if test -f "${EMACS_DEFAULT_EL}" ; then \
> +       -if test -f "${EMACS_DEFAULT_EL}" ; then \
>         cat ${EMACS_DEFAULT_EL} | sed -e '/BEGIN gcl/,/END gcl/d' > xxx ; \
>         mv ${EMACS_DEFAULT_EL} ${EMACS_DEFAULT_EL}.prev ; \
>           rm -f  ${EMACS_DEFAULT_EL}c ; \
>            cat add-default.el >> xxx ; cp  xxx ${EMACS_DEFAULT_EL} ; \
>           rm -f xxx ; else \
>         cp  add-default.el ${EMACS_DEFAULT_EL} ; fi
> -       chmod a+r ${EMACS_DEFAULT_EL}
> +       -chmod a+r ${EMACS_DEFAULT_EL}
> 
> (3) ---------------------------------------------------------------
> 
> if [  -d "/usr/info/" ] ; then  cp *.info* /usr/info/ ; fi
> cp: cannot create regular file `/usr/info/gcl-si.info': Read-only file system
> cp: cannot create regular file `/usr/info/gcl-si.info-1.gz': Read-only file 
> system
> cp: cannot create regular file `/usr/info/gcl-si.info-2.gz': Read-only file 
> system
> cp: cannot create regular file `/usr/info/gcl-si.info-3.gz': Read-only file 
> system
> 
> (I gave --prefix=/tmp for testing but nevertheless GCL wanted to install
> info pages in /usr)
> 

OK, this one I've run into as well, and is a bit awkward, as it will
take a rethinking of the makefiles to really do properly.  The info
files, and the .el files, are installed in directories determined at
configure time, according to the output of emacs run on some testfile
if present.  They are not installed in preset locations under some
"prefix".  We should probably change this, but what I did for now is
to leave the default install alone, but add a "DESTDIR" make variable,
which will simply prepend some path to *every* file gcl will install.
I.e. what prefix really should be, but is not in the current setup.

> (4) ---------------------------------------------------------------
> 
> This is a bug in libtool. I submitted patch, and recent libtool releases
> should be fixed:
> 
> checking for file... /usr/bin/file
> checking for ranlib... ranlib
> checking for strip... strip
> updating cache ./config.cache
> ./ltconfig: /home/cyborg/bin/: is a directory
> ./ltconfig: exec: /home/cyborg/bin/: cannot execute: Permission denied
> configure: error: libtool configure failed
> 
> (this happens since ltconfig tries to execute exec "", i.e. with
> zero-length argument. I'd say bash has a bug in error reporting too) FIX:

Applied.

> 
> --- gcl-2.4.0-orig/gmp/ltconfig Mon Mar 20 14:45:00 2000
> +++ gcl-2.4.0/gmp/ltconfig      Sun Jul 22 13:21:17 2001
> @@ -117,7 +117,7 @@
>        exec "$CONFIG_SHELL" "$0" --no-reexec ${1+"$@"}
>      else
>        # Try using printf.
> -      echo='printf "%s\n"'
> +      echo='printf %s\n'
>        if test "X`($echo '\t') 2>/dev/null`" = 'X\t' &&
>          echo_testing_string=`($echo "$echo_test_string") 2>/dev/null` &&
>          test "X$echo_testing_string" = "X$echo_test_string"; then
> 
> (5) ---------------------------------------------------------------
> 
> This happens while compiling Maxima-5.6:
> 
> rm -f `echo  opers.o utils.o sumcon.o sublis.o runtim.o merror.o mformt.o
> mutils.o outmis.o ar.o misc.o comm.o comm2.o mlisp.o mmacro.o buildq.o
>   [...deleted similar lines...]
> psolve.o algsys.o polyrz.o cpoly.o mtrace.o scs.o asum.o fortra.o optim.o
> marray.o mdot.o irinte.o series.o numth.o laplac.o pade.o homog.o combin.o
> mstuff.o ratpoi.o pois2.o pois3.o  mdebug.o | sed -e "s:[.]o:.lisp"`
> sed: -e expression #1, char 12: Unterminated `s' command
> 
> FIX: change sed -e "s:[.]o:.lisp" to sed -e "s:[.]o:.lisp:g"
> (or maybe just sed -e "s:[.]o:.lisp:")
> 

This is a bug in the maxima makefile, I think the one under the src
directory.

Take care,

> 
> 

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