clisp-devel

screen -> module

[Sam S. <sd...@gn...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

why is not screen a module?
the answer always have been: "for the same reason so sockets have to be
in stream.d: modules cannot create new built-in streams".
this is a valid reason for sockets which must be as fast as file and
pipe streams and thus cannot be gray streams, but why can't window
streams be gray streams?

PS. actually, it is not clear why modules cannot create new kinds of
streams. they already create objects, why can't they create
pseudofunctions? and what else is needed for new built-in streams
created by modules?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHNH/CPp1Qsf2qnMcRApiFAJ9f8sR7576p5YASlR5e2UYyRVFYMgCfdeD5
ebZLNCYWhiLk7E0Y+T75u3Y=
=EGx1
-----END PGP SIGNATURE-----

clisp 2.43 Solaris 9 build failure

[Michael A. <Mic...@fs...>]

Hello Sam,

on Solaris 9 compilation of current CVS fails since there is no stdint.h:

gcc  -g -O2 -W -Wswitch -Wcomment -Wpointer-arith -Wimplicit -Wreturn-type
-Wmissing-declarations -Wno-sign-compare -O2 -fno-schedule-insns -fno-gcse
-DUNIX_BINARY_DISTRIB -DUNICODE -DDYNAMIC_FFI -DNO_GETTEXT -DNO_SIGSEGV
-I. -c sequence.c
In file included from sequence.d:6:
lispbibl.d:8967: warning: volatile register variables don't work as you
might wish
lispbibl.d:9100: warning: register used for two global register variables
In file included from sequence.d:6:
lispbibl.d:11447: warning: register used for two global register variables
lispbibl.d:11466: warning: register used for two global register variables
lispbibl.d:11672: warning: register used for two global register variables
rm -f streq.h
ln -s gllib/streq.h streq.h
rm -f cjk.h
ln -s gllib/uniwidth/cjk.h cjk.h
rm -f uniwidth.h
ln -s gllib/uniwidth.h uniwidth.h
rm -f unitypes.h
ln -s gllib/unitypes.h unitypes.h
rm -f uniname.h
ln -s gllib/uniname.h uniname.h
rm -f uninames.h
ln -s gllib/uniname/uninames.h uninames.h
gcc  -g -O2 -W -Wswitch -Wcomment -Wpointer-arith -Wimplicit -Wreturn-type
-Wmissing-declarations -Wno-sign-compare -O2 -fno-schedule-insns -fno-gcse
-DUNIX_BINARY_DISTRIB -DUNICODE -DDYNAMIC_FFI -DNO_GETTEXT -DNO_SIGSEGV
-I. -c charstrg.c
In file included from charstrg.d:8:
lispbibl.d:8967: warning: volatile register variables don't work as you
might wish
lispbibl.d:9100: warning: register used for two global register variables
In file included from charstrg.d:8:
lispbibl.d:11447: warning: register used for two global register variables
lispbibl.d:11466: warning: register used for two global register variables
lispbibl.d:11672: warning: register used for two global register variables
In file included from uniwidth.h:20,
                 from charstrg.d:378:
unitypes.h:21:20: stdint.h: No such file or directory
make: *** [charstrg.o] Error 1

Solaris 10 does have a working sdtint.h

Cheers,

Michael

Re: clisp 2.43 Solaris 9 build failure

[Sam S. <sd...@gn...>]

Hi Michael,

alas, I cannot get cvs to work on heron:
$ cvs up
? ffcall/build-aux/.new.config.sub
cvs [update aborted]: writing ffcall/build-aux/config.sub: No such file or directory

so I am afraid we are up to a "debug by proxy" session. :-(

> * Michael Abshoff <Zvpunry.Nofubss@sfzngu.zngurzngvx.hav-qbegzhaq.qr> [2007-11-09 22:50:30 +0100]:
>
> rm -f streq.h
> ln -s gllib/streq.h streq.h
> rm -f cjk.h
> ln -s gllib/uniwidth/cjk.h cjk.h
> rm -f uniwidth.h
> ln -s gllib/uniwidth.h uniwidth.h
> rm -f unitypes.h
> ln -s gllib/unitypes.h unitypes.h
> rm -f uniname.h
> ln -s gllib/uniname.h uniname.h
> rm -f uninames.h
> ln -s gllib/uniname/uninames.h uninames.h
> gcc  -g -O2 -W -Wswitch -Wcomment -Wpointer-arith -Wimplicit -Wreturn-type
> -Wmissing-declarations -Wno-sign-compare -O2 -fno-schedule-insns -fno-gcse
> -DUNIX_BINARY_DISTRIB -DUNICODE -DDYNAMIC_FFI -DNO_GETTEXT -DNO_SIGSEGV
> -I. -c charstrg.c
> In file included from charstrg.d:8:
> lispbibl.d:8967: warning: volatile register variables don't work as you
> might wish
> lispbibl.d:9100: warning: register used for two global register variables
> In file included from charstrg.d:8:
> lispbibl.d:11447: warning: register used for two global register variables
> lispbibl.d:11466: warning: register used for two global register variables
> lispbibl.d:11672: warning: register used for two global register variables
> In file included from uniwidth.h:20,
>                  from charstrg.d:378:
> unitypes.h:21:20: stdint.h: No such file or directory
> make: *** [charstrg.o] Error 1

1. is configure happy with the system-supplied stdint.h?

$ grep stdint build/config.cache

does it say something like
gl_cv_header_working_stdint_h=${gl_cv_header_working_stdint_h=yes}
??

2. does it create a build/gllib/stdint.h from src/gllib/stdint.in.h?

$ ls -l build/gllib/

3. does the appended patch fix the problem?

> Solaris 10 does have a working sdtint.h

where?
I guess gnulib is not happy with it...

-- 
Sam Steingold (http://sds.podval.org/) on Fedora release 7 (Moonshine)
http://pmw.org.il http://jihadwatch.org http://mideasttruth.com
http://israelunderattack.slide.com http://camera.org
Apathy Club meeting this Friday.  If you want to come, you're not invited.


Index: makemake.in
===================================================================
RCS file: /cvsroot/clisp/clisp/src/makemake.in,v
retrieving revision 1.699
diff -u -w -p -u -b -w -i -B -r1.699 makemake.in
--- makemake.in	7 Nov 2007 15:20:54 -0000	1.699
+++ makemake.in	11 Nov 2007 01:39:09 -0000
@@ -631,6 +631,8 @@ case $# in
      ;;
 esac
 
+CFLAGS=${CFLAGS}' -Igllib'
+
 if [ "${with_gmalloc}" = yes ]; then GMALLOC=gmalloc; else GMALLOC=""; fi
 
 if [ $CROSS = true ] ; then
@@ -1956,7 +1958,7 @@ echol "X_LIBS = ${X_LIBS}"
 X_LIBS="\$(X_LIBS)"
 if [ "${with_dynamic_modules}" != no ]; then
   echol 'MODULE_CPPFLAGS = $(CPPFLAGS)'
-  echol 'MODULE_CFLAGS = $(CFLAGS) -Igllib '"${XCC_PICFLAG}"
+  echol 'MODULE_CFLAGS = $(CFLAGS) '"${XCC_PICFLAG}"
   echol 'MODULE_CLFLAGS = $(CLFLAGS) '"${XCC_DLOPENABLE_EXE_FLAG}"
   MODULE_CPPFLAGS_VAR='MODULE_CPPFLAGS'
   MODULE_CFLAGS_VAR='MODULE_CFLAGS'


Diffs between working revision and workfile end here.

Re: screen -> module

[Bruno H. <br...@cl...>]

Sam asked:
> why is not screen a module?
> the answer always have been: "for the same reason so sockets have to be
> in stream.d: modules cannot create new built-in streams".
> this is a valid reason for sockets which must be as fast as file and
> pipe streams and thus cannot be gray streams, but why can't window
> streams be gray streams?

The speed argument here too, and a question of ease of implementation.

> PS. actually, it is not clear why modules cannot create new kinds of
> streams. they already create objects, why can't they create
> pseudofunctions?

It's because pseudocode_tab and pseudodata_tab are fixed arrays. If you
were to change it into an extensible array of arrays, modules could register
their share of it.

> and what else is needed for new built-in streams created by modules?

There are 4 instances of 'case strmtype_window' in stream.d. These would
have to be replaced with C function calls, through function pointers.

Bruno