scm_make_debugobj), eval.c (scm_makprom): SCM_DEFER/ALLOW_INTS -->
A section.
* __scm.h: Start the long-term project of moving to POSIX threads.
Phase 1: Classification of all critical sections.
(SCM_ENTER_A_SECTION, SCM_EXIT_A_SECTION): New macros: Delimiters
for A sections. (See comments in __scm.h for details.)
(SCM_ALLOW_INTS): Added thread switching code before interrupts
get re-enabled. The important effect of this is that interrupts
are blocked during thread switching so that thread data structures
don't risk getting messed up by an unfortunate signal.
(SCM_REDEFER_INTS, SCM_REALLOW_INTS): It turned out that gcc-2.8.0
seems to do more aggressive optimization which actually move
instructions around in these macros in a fatal way. Therefore:
Introduce Anthony's SCM_FENCE macro! (And I who thought he was
just superstitious...)
(SCM_TICK): Maybe do a context switch and take care of asyncs.
This macro should be used instead of SCM_ASYNC_TICK since the
latter doesn't do context switches any more.
across this. Only works on GCC. Otherwise, we hope for the best.
(SCM_DEFER_INTS, SCM_ALLOW_INTS): Use FENCE appropriately. I have
the feeling that real thread systems will not need this...
* Makefile.am: remove AWK=@AWK@.
Add a rule for generating errnos.list.
(CLEANFILES): put errnos.list here instead of in DISTCLEANFILES.
* configure.in: add AC_SUBST(AWK) and AC_SUBST(ERRNO_EXTRACT).
don't extract errnos, just set a variable (avoids the
need to recompile error.c just because configure is run.)
* unif.h: update prototypes.
* unif.c (scm_uniform_array_read,write): change the offset and
length arguments to start and end, for consistency.
* __scm.h: uncomment SCM_ARG6 and SCM_ARG7, I needed SCM_ARG6.
* ioext.h: update prototypes.
* * ioext.c (scm_read_delimited_x): replaces scm_read_line and
scm_read_line_x, it's a more general procedure using an
interface from scsh. read-line and read-line! are now defined
in boot-9.scm.
* Note that the new read-line trims the terminator
by default, previously it was appended to the returned string. An
optional argument specifies how to process the terminator (scsh
compatible). For the old behaviour: (read-line port 'concat).
scm_read_line, scm_read_line_x: deleted. (read-line port 'split)
returns a pair, but is converted to multiple values if the scsh
module is loaded.
socket.h: update prototypes.
* socket.c (scm_recvfrom): for consistency with other procedures,
take start and end as separate optional arguments.
* (scm_recv, scm_recvfrom): don't allow the second argument
to be a size, only a buffer. Change the scheme names to
recv! and recvfrom!. Don't return the buffer.
* ioext.h, posix.h: move prototypes too.
* ioext.c, posix.c (scm_read_line, scm_read_line_x, scm_write_line:
moved back from posix.c to ioext.c. Also move #includes of "genio.h"
"read.h" and "unif.h".
* ioext.c: include "chars.h"
macros: SCM_THREAD_DEFER, SCM_THREAD_ALLOW, SCM_THREAD_REDEFER,
SCM_THREAD_ALLOW_1, SCM_THREAD_ALLOW_2. Motivation: We don't want
the main code in these macros duplicated and spread over multiple
files. Renamed SCM_THREADS_SWITCHING_CODE ->
SCM_THREAD_SWITCHING_CODE.
* __scm.h: #undef SICP.
* pairs.h (SCM_EOL): Delete this definition, equating it with
SCM_BOOL_F.
* tags.h (SCM_EOL): Give it a new definition here; I think I found
the value it used to have. Doc fixes, too.
integer (assuming for now accepting an integer is a good thing).
* error.c, fports.c: replace use of %S in lgh_error args with %s.
%S will be used instead for write'ing arguments.
* unif.c (scm_transpose_array): change arguments in the SCM_WNA
asserts. fix a few other asserts.
(scm_aind, scm_enclose_array, scm_array_in_bounds_p,
scm_uniform_vector_ref, scm_array_set_x,
scm_dimensions_to_unform_array): change args in
SCM_WNA SCM_ASSERTS and change scm_wta's to scm_wrong_num_args.
strop.c (scm_substring_move_left_x, scm_substring_move_right_x,
scm_substring_fill_x): likewise.
gsubr.c (scm_gsubr_apply): likewise.
eval.c (SCM_APPLY): likewise.
* eval.c (4 places): replace scm_everr with lgh_error or
scm_wrong_num_args.
* error.c, error.h (scm_wrong_num_args, scm_wrong_type_arg,
scm_memory_error): new procedures.
* scm_everr: deleted. can use scm_wta, dropping first two args.
scm_error: convert NULL subr to SCM_BOOL_F.
* __scm.h: don't define SCM_STACK_OVFLOW, SCM_EXIT, SCM_ARG6, SCM_ARG7,
SCM_ARGERR.
* stackchk.c (scm_report_stack_overflow): use lgh_error instead
of scm_wta.
* error.c, error.h: new error keys: scm_arg_type_key,
scm_args_number_key, scm_memory_alloc_key, scm_stack_overflow_key,
scm_misc_error_key.
scm_wta: reimplement using lgh_error instead of scm_everr.
wta for range errors (ASSERT still needs work).
* error.c, error.h (scm_out_of_range): new procedure.
* error.c, error.h (scm_out_of_range_key): new key.
* posix.c (scm_sync): #else was missing.
* error.c, error.h: append _key to names scm_num_overflow and
scm_system_error.
* __scm.h (SCM_SYSMISSING, SCM_NUM_OVERFLOW): use SCM_BOOL_F instead
of consing an empty list.
(SCM_SYSERROR etc.): move into error.c, make them procedures instead
of macros, saves code and string space.
error.c, fports.c, numbers.c, posix.c, ioext.c, filesys.c, socket.c,
fdsocket.c, simpos.c: change names of SCM_SYSERROR etc., to
lower case. Rename scm_syserror_m to scm_syserror_msg.
error.h: prototypes for new procedures.
* error.c, error.h: setup scm_num_overflow key.
* __scm.h: SCM_NUM_OVERFLOW: macro for reporting numerical overflow.
Remove definition of SCM_OVSCM_FLOW.
* fports.c (scm_open_file): use SCM_SYSERROR_M.
* __scm.h: SCM_SYSERROR_M: new macro for system errors with an
explicit message and args.
* error.c, error.h, __scm.h: change system_error_sym to
scm_system_error.
* error.c (system_error_sym): remove leading %% from the Scheme name
"%%system-error".
* __scm.h (SCM_SYSMISSING): Redefine using lgh_error.
* boot-9.scm: remove leading %% from references to '%%system-error.
(%%handle-system-error): don't pass all the thrown arguments when
aborting, just the key and subr.
Remove the code to "Install default handlers for built-in errors."
Remove the definition of the syserror procedure.
Associate 'numerical-overflow with default handler.
main #include path; put most of them in a subdirectory called
'libguile'. This avoids naming conflicts between Guile header
files and system header files (of which there were a few).
* Makefile.in (pkgincludedir): Deleted.
(innerincludedir): New variable; this and $(includedir) are enough.
(INCLUDE_CFLAGS): Search for headers in "-I$(srcdir)/..".
(installed_h_files): Divide this up. Now this variable lists
those header files which should go into $(includedir) (i.e. appear
directly in the #include path), and ...
(inner_h_files): ... this new variable says which files appear in
a subdirectory, and are referred to as <libguile/mumble.h>.
(h_files): List them both.
(install): Create innerincludedir, not pkgincludedir. Put
the installed_h_files and inner_h_files in their proper places.
(uninstall): Corresponding changes.
* alist.h, append.h, arbiters.h, async.h, boolean.h, chars.h,
continuations.h, debug.h, dynwind.h, error.h, eval.h, fdsocket.h,
feature.h, fports.h, gc.h, genio.h, gsubr.h, hash.h, init.h,
ioext.h, kw.h, libguile.h, list.h, markers.h, marksweep.h,
mbstrings.h, numbers.h, options.h, pairs.h, ports.h, posix.h,
print.h, procprop.h, procs.h, ramap.h, read.h, root.h,
sequences.h, smob.h, socket.h, srcprop.h, stackchk.h, stime.h,
strings.h, strop.h, strorder.h, strports.h, struct.h, symbols.h,
tag.h, throw.h, unif.h, variable.h, vectors.h, version.h,
vports.h, weaks.h: Find __scm.h in its new location.
* __scm.h: Find scmconfig.h and tags.h in their new locations
(they're both "inner" files).
preprocessor symbol; "__scm.h" is where it might get #defined.
* __scm.h: Similar: #include <scmconfig.h> before testing
HAVE_LIMITS_H.
* __scm.h: It's HAVE_LIMITS_H, not HAVE_LIMITSH.
* __scm.h (PROTO): New macro, for declaring functions with
prototypes.
* __scm.h (SCM_SYSMISSING): When we don't have ENOSYS, don't
complain about it in the error message; the error message is
adequate without that note, and there's nothing the user can do
about it.
