guile environment patch

[Jost Boekemeier <jostobfe@calvados.zrz.TU-Berlin.DE>]

the following two patches implement first class top-level environments
as described by Jim Blandy in his environment proposal
(http://www.gnu.org/software/guile/docs/env.html) and a simple module
system on top of them (no documentation yet).

environments:
ftp://ftp.tfh-berlin.de/pub/incoming/guile-environment.patch.gz

module system:
ftp://ftp.tfh-berlin.de/pub/incoming/guile-environment-test.patch.gz


Changelog entry:

	* alist.c, arbiters.c, async.c, backtrace.c, boolean.c chars.c,
	  continuations.c, debug.c, dynl.c, dynwind.c eq.c, eval.c,
	  evalext.c, feature.c, filesys.c, fluids.c, fports.c, gc.c,
	  gdbint.c, gsubr.c, guardians.c, hash.c, hashtab.c, init.c,
	  ioext.c, iselect.c, keywords.c, list.c, load.c, macros.c,
	  mallocs.c, net_db.c, numbers.c, objects.c, objprop.c, options.c,
	  pairs.c, ports.c, posix.c, print.c, procprop.c, procs.c,
	  ramap.c, random.c, read.c, readline.c, regex-posix.c, root.c,
	  scmsigs.c, script.c, simposc, socket.c, sort.c, stackchk.c,
	  stacks.c, stime.c, strings.c, strop.c, strorder.c, strports.c,
	  struct.c, symbols.c, tag.c, threads.c, throw.c, unif.c,
	  vectors.c, version.c, vports.c, weaks.c: 
	changed `void scm_init_<name>()' to `SCM scm_init_<name>(env)'.
	All init functions take an environment argument and return
	SCM_UNSPECIFIED.  For example `scm_init_struct (scm_root_environment)'.
	

	* async.c, backtrace.c, debug.c, feature.c, filesys.c, fports.c,
	gh_funcs.c, gsubr.c, load.c, macros.c, net_db.c, objects.c,
	ports.c, posix.c, procs.c, read.c, readline.c, regex-posix.c,
	root.c, scmsigs.c, script.c, snarf.h, socket.c, srcprop.c,
	cpp_cnvt.awk, stime.c, struct.c: changed scm_sysintern() to
	scm_environment_intern(env, <string>, <value>)
	  
	* debug.c removed procedures make-gloc and gloc?

	macro scm_procedure_environment returns the current environment
	for tc7_cclo
	
	environment argument to scm_memcons is now required

	changed SCM_CAR(scm_sysintern(...)) to 
	   scm_permanent_object (SCM_CAR (scm_intern ("<name>")))

	* dynl.c : 
	removed module registry, struct moddata, scm_register_module_xxx(), 
	scm_registered_modules(), scm_clear_registered_modules())

	scm_dynamic_call() takes an environment as its third argument which
	is passed on to the init function
	
	* eval.c : 

	Convention:  ENVIRONMENT names environment smobs
	             ENV names environments which have the following structure:
	             (environment-smob . local-alist)
	
	Macros EVALCELLCAR, EVALCAR, SIDEVAL take an environment as 2nd
	argument

	Macro EXTEND_ENV turned into a function that cons'es a new
	(environment-smob . local-alist) pair

	scm_ilookup(), scm_lookupcar(), scm_lookupcar1(), scm_unmemocar(),
	scm_ceval(), scm_unmemocopy(): environment argument is a pair
	(environment-smob . local-alist)

	lookupcar1(): removed top_thunk, changed scm_sym2vcell to 
	SCM_ENVIRONMENT_CELL(), memoization is done through 
	scm_eval_environment_memoize_cell_internal()

	scm_unmemocar(): unmemoize gloc via
	eval_environment_update_memoized()

	scm_m_define, unmemocopy, scm_ceval: changed SCM_TOP_LEVEL(env) to
	SCM_TOP_LEVEL_ENVP (env)  

	scm_m_define(): scm_sym2vcell(...) changed to
	SCM_ENVIRONMENT_DEFINE()

	scm_ceval(): all environment manipulations changed to
	 {
	   SCM local_env = SCM_CDR (env);
	   <do something with local_env>
	 }
	 

	scm_eval_3() : removed scm_system_transformer

	scm_eval(...)      -> scm_eval(..., env) env must be an environment
	                      smob 
	scm_eval_x(...)  dto.

	* evalext.c turned scm_defined_p (defined?) into syntax.  Symbol
	is looked up in the local environment, the environment-smob and
	all its ancestors.

	scm_make_named_hook(...) -> scm_make_named_hook(..., env)

	* gc.c scm_gc_mark() msymbols don't have slots anymore.

	scm_must_malloc(), scm_must_realloc(), scm_must_free():
	changed char* to void*

	scm_init_storage(): removed scm_symhash

	* gh.h, gh_*.c
	gh_eval_str_*(...)  -> gh_eval_str_*(..., env) 
	gh_eval_file_*     dto.
	gh_load            dto.
	gh_new_procedure*  dto.
	gh_define          dto.

	disabled gh_lookup() and gh_module_lookup() which aren't implemented
	
	* gsubr.c : scm_make_gsubr(...) -> scm_make_gsubr(..., env)

	* guardians.c : scm_init_guardian(): create "guardian" subr in
	the current env.

	* init.c : start_stack(): takes environment as first argument,
	removed scm_top_level_lookup_closure_var and
	scm_system_transformer

	scm_init_standard_ports(): added environment parameter

	scm_load_startup_files(): load boot_9.scm and init.scm into
	scm_root_environment

	scm_boot_guile():
	
	  scm_init_environments() creates scm_root_environment

	  scm_init_root() ... scm_init_script(): changed to scm_init_*(env)
	
	  turned scm_init_regex_posix(), .., scm_init_random() into
	  c modules

	* keywords.c
	changed scm_sym2ovcell_soft(symbol, scm_keyword_obarray) into
	SCM_ENVIRONMENT_CELL(scm_keyword_environment, symbol, 1)

	changed SCM_SETCDR (scm_sym2ovcell_soft(symbol ...), keyword)
	to SCM_ENVIRONMENT_DEFINE (scm_keyword_environment, symbol, keyword)

	* libguile.h : removed variable.h 

	* load.c : load(): changed  argument to 
	&struct load_data{ SCM port; SCM env; };

	primitive_load():           -> primitive_load(..., env)
	scm_init_load_path():       dto.
	scm_primitive_load_path() : dto.
	scm_primitive_load() :      dto.

	scm_read_and_eval_x() : 2 optional arguments: port and env.  env
	defaults to interaction-environment, not the current environment!!
	Guile has no idea what (the-environment) is unless you explicitly
	pass it as an argument.

	* macros.c scm_make_synt ()  -> scm_make_synt (..., env);

	* net_db.c, options.c, posix.c, regex-posix.c :
	scm_sysintern(...) -> scm_environment_intern(env,...)

	
	* procs.c :
	scm_make_subr_opt(...) -> scm_make_subr_opt(..., env)
	scm_make_subr(...):    -> scm_make_subr(..., env): 

	* read.c : scm_lreadr(); removed length argument to scm_intern()

	* root.c : removed scm_top_level_lookup_closure_var and
	system_transformer from scm_root_state

	moved initialization of most-positive-fixnum, most-negative-fixnum,
	bignum-radix from scm_init_storage() to scm_init_root()

	* root.h: changed scm_keyword_obarray to scm_keyword_environment.
	The keyword environment is a finite-environment and maps keyword
	symbols to values.

	changed sym_symhash to scm_root_environment.  The root-environment
	is an eval-environment.
	
	added scm_c_module_registry. A finite-environment which maps
	module symbols like 'ice-9/posix to values of type scm_tc7_subr_1
	(see init.c).

	added scm_interaction_environment.  According to r5rs this is a
	global variable that points to the environment in which the user
	*types* commands.  The r5rs procedure load evaluates expressions in
	this environment.  Note that this may lead to confusion if the load
	procedure is used whithin a script where (interaction-environment)
	is not equal (the-environment).  So it may be better to use
	scm_primitive_load() instead.
	
	* script : scm_compile_shell_switches(): evaluate "-c" expressions
	in scm_root_environment

	when interactive: load "ice-9/modules.scm" into scm_root_environment
	and (go (access (ice-9 guile)))

	* snarf.h : Makros SCM_PROC and SCM_SYNT defined to
	`scm_make_gsubr (RANAME, REQ, OPT, VAR, (SCM (*)()) CFN, env)'
	where env must be supplied by the scm_init_*() function.

	Makros SCM_VCELL, SCM_GLOBAL_VCELL, SCM_VCELL_INIT,
	SCM_GLOBAL_VCELL_INIT create bindings in the environment env,
	where env must be a local variable of the enclosing
	scm_init_*() function.  init_val defaults to SCM_BOOL_F. 
	Note that it is an error to set the value of an environment binding
	to SCM_UNSPECIFIED or SCM_UNDEFINED!  The following would have
	strange side effects: scm_environment_define(env, sym, SCM_UNDEFINED)
	If you do that, environment-bound? will tell you that sym is bound
	but environment-cell and environment-fold will tell you something
	differend.
	
	* stacks.c get_applybody(): replaced lookup for symbol
	`scm_sym_apply' by value `scm_i_apply'

	* strports.c scm_eval_0str(), scm_eval_string(): added required
	environment parameter

	* struct.c scm_make_struct_layout(): changed
	SCM_CAR(scm_intern_obarray(...)) to scm_permanent_object (SCM_CAR
	(scm_intern (...)))
	
	* symbols.c removed scm_sym2vcell(), scm_sym2ovcell(),
	scm_sym2ovcell_soft(), scm_intern_obarray_soft(),
	scm_intern_obarray(), scm_intern0(),
	scm_sysintern_0_no_module_lookup(),
	scm_can_use_top_level_lookup_closure_var, scm_sysintern(),
	scm_sysintern0(), scm_symbol_value0(), scm_symbol_value(),
	scm_string_to_obarray_symbol(), scm_intern_symbol(),
	scm_unintern_symbol(), scm_symbol_binding(),
	scm_symbol_internec(), scm_symbol_bound_p(), scm_symbol_set_x(),
	msymbolize(), scm_symbol_fref(), scm_symbol_pref(),
	scm_symbol_fset_x(), scm_symbol_pset_x(), scm_symbol_hash(),
	copy_and_prune_obarray(), scm_builtin_bindings(),
	scm_builtin_weak_bindings(), scm_symhash.

	changed scm_intern(name, len) to scm_intern(name)

	changed scm_gensym (name, obarray) to scm_gensym (name)
	
	* symbols.h
	removed all slots from msymbols (In the future msymbols
	can go away completely).

	ssymbols and msymbols now carry their hash value at position 
	(* ((SCM *)SCM_CHARS(x) - 1)))

	added macros SCM_SYMBOL_HASH(sym) and SCM_HASHCODE(sym) to 
	access a symbol's hash value (both macros do the same but
	SCM_HASHCODE() is easier to remember :>)

	* _scm.h: added #include "environments.h" as scm_environment_intern()
	(the replacement for scm_sysintern()) is used throughout the system.

	* Makefile.am: removed variables.[chx], removed kw.h, added
	environments.[ch]


Jost