bpt/guile.git
12 years agognulib update
Andy Wingo [Thu, 16 Jun 2011 16:22:50 +0000 (18:22 +0200)]
gnulib update

Added pipe2 and open modules, to get O_CLOEXEC.

* libguile/Makefile.am (guile_filter_doc_snarfage$(EXEEXT)): Add gnulib
  here, in the native build case, for rpl_fflush if needed.

foo

12 years agofix initial values of reallocated fluids
Andy Wingo [Thu, 16 Jun 2011 17:35:14 +0000 (19:35 +0200)]
fix initial values of reallocated fluids

* libguile/threads.h:
* libguile/threads.c (scm_i_reset_fluid): New internal function, resets
  the binding of a fluid for all threads.  Needed for fluid GC.

* libguile/fluids.c (new_fluid): Call scm_i_reset_fluid here.

12 years agomore NEWS
Andy Wingo [Thu, 16 Jun 2011 13:59:19 +0000 (15:59 +0200)]
more NEWS

* NEWS: More updates.

12 years agoupdate NEWS
Andy Wingo [Thu, 16 Jun 2011 11:29:17 +0000 (13:29 +0200)]
update NEWS

* NEWS: Update for 2.0.2.

12 years agoadd docs for shift and reset
Andy Wingo [Thu, 16 Jun 2011 11:01:43 +0000 (13:01 +0200)]
add docs for shift and reset

* doc/ref/api-control.texi (Prompt Primitives): Break call-with-prompt
  and abort-to-prompt out into a subsubsection.
  (Shift and Reset): New subsubsection.

12 years agofix hash-set! on weak tables
Andy Wingo [Thu, 16 Jun 2011 10:06:43 +0000 (12:06 +0200)]
fix hash-set! on weak tables

* test-suite/tests/weaks.test: Add tests.
* libguile/hashtab.c (scm_hash_fn_set_x): Fix updates to weak-value hash
  tables to not deadlock inside the alloc lock.

12 years agoFix likely crash in `stable-sort!'
Andreas Rottmann [Thu, 9 Jun 2011 20:11:02 +0000 (22:11 +0200)]
Fix likely crash in `stable-sort!'

* libguile/sort.c (scm_stable_sort_x): Properly handle zero-length
  vectors.
* test-suite/tests/sort.test ("stable-sort"): Add test for this case.

12 years agodoc: Mention the former name of `guild'.
Ludovic Courtès [Wed, 8 Jun 2011 21:19:37 +0000 (23:19 +0200)]
doc: Mention the former name of `guild'.

* doc/ref/scheme-using.texi (Using Guile Tools): Add a footnote
  mentioning that `guild' used to be `guile-tools'.

* doc/ref/tools.texi (Executable Modules): Nitpicks.

12 years agoadd guild docs
Andy Wingo [Tue, 31 May 2011 20:36:52 +0000 (22:36 +0200)]
add guild docs

* doc/ref/scheme-using.texi (Using Guile Tools): Add some optimistic
  guild docs.

12 years agorename `guile-tools' to `guild'
Andy Wingo [Mon, 30 May 2011 21:02:03 +0000 (23:02 +0200)]
rename `guile-tools' to `guild'

* configure.ac: Look for ln -s.  Write out `guild' instead of
  `guile-tools'.

* meta/Makefile.am (install-data-hook): Link the installed `guild' to
  the backward-compatible `guile-tools' name.
  (bin_SCRIPTS, EXTRA_DIST): Fix up for guild change.

* meta/guild.in: Moved here from `guile-tools.in'.

* doc/ref/Makefile.am (autoconf-macros.texi):
* doc/ref/api-evaluation.texi (Compilation):
* doc/ref/autoconf.texi (Autofrisk, Using Autofrisk):
* doc/ref/mod-getopt-long.texi (getopt-long Reference):
* doc/ref/tools.texi (Miscellaneous Tools, Executable Modules): Minimal
  doc update.

* .gitignore:
* am/guilec (.scm.go):
* libguile/Makefile.am (snarf2checkedtexi):
* module/Makefile.am (ice-9/psyntax-pp.go): Update makefiles, etc.

* module/scripts/README:
* module/scripts/lint.scm:
* module/scripts/list.scm: Update commentaries.

12 years agowrite-objcode uses target-endianness, target-word-size
Andy Wingo [Mon, 30 May 2011 20:18:48 +0000 (22:18 +0200)]
write-objcode uses target-endianness, target-word-size

* libguile/_scm.h (SCM_OBJCODE_ENDIANNESS_OFFSET):
  (SCM_OBJCODE_WORD_SIZE_OFFSET): New defines.
* libguile/objcodes.c (scm_write_objcode): Use target-endianness and
  target-word-size when writing the objcode cookie.

12 years agocompile-bytecode uses target-endianness
Andy Wingo [Fri, 27 May 2011 11:29:45 +0000 (13:29 +0200)]
compile-bytecode uses target-endianness

* module/language/assembly/compile-bytecode.scm (compile-bytecode):
  Use target-endianness, from (system base target).

12 years agoadd (system base target)
Andy Wingo [Mon, 30 May 2011 19:28:30 +0000 (21:28 +0200)]
add (system base target)

* module/Makefile.am:
* module/system/base/target.scm: Add a minimal module to parameterize
  the target system type and inspect properties on it like cpu, vendor,
  os, endianness, and word size.

12 years agofix subtle and bad scm_internal_hash_fold bug for weak tables
Andy Wingo [Mon, 30 May 2011 09:33:10 +0000 (11:33 +0200)]
fix subtle and bad scm_internal_hash_fold bug for weak tables

* libguile/hashtab.c (scm_internal_hash_fold): Don't try to unlink
  deleted weak pairs.  Our previous code was buggy (`prev' should have
  only been updated in the case of a successful traversal), but more
  than that, we're not in the alloc lock.

  Thanks very much to Michael Wells for the report, and the debugging!

12 years agoTweak R6RS transcoded ports flushing
Andreas Rottmann [Fri, 27 May 2011 14:09:04 +0000 (16:09 +0200)]
Tweak R6RS transcoded ports flushing

* libguile/r6rs-ports.c (tp_flush): Only operate on the underlying port
  when it is open.

12 years agornrs io ports: fix port encoding when opening file ports
Andreas Rottmann [Fri, 27 May 2011 13:32:01 +0000 (15:32 +0200)]
rnrs io ports: fix port encoding when opening file ports

* module/rnrs/io/ports.scm (open-file-input-port)
  (open-file-output-port): Ensure the resulting ports are binary when no
  transcoder is specified.

* test-suite/tests/r6rs-ports.test: Remove superfluous global change of
  `%default-port-encoding' and accompanying comment.
  ("7.2.7 Input Ports"): Add test ensuring `open-file-input-port' opens
  a binary port when no transcoder is specified.
  ("8.2.10 Output ports"): Strengthen existing `open-file-output-port'
  binary-ness test by setting `%default-port-encoding' to "UTF-8".

12 years agoAdd some tests for the R6RS I/O libraries
Andreas Rottmann [Fri, 27 May 2011 12:48:31 +0000 (14:48 +0200)]
Add some tests for the R6RS I/O libraries

* test-suite/tests/r6rs-ports.test
  (call-with-bytevector-output-port/transcoded): New helper procedure.
  ("8.2.6 Input and output ports"): Use that helper procedure.
  (encoding-error-predicate): New helper procedure.
  ("8.2.12 Textual Output"): Add tests for `put-char' and `put-string'
  exception behavior on encoding errors.

12 years agoDoc for getopt-long's new stop-at-first-non-option option
Neil Jerram [Thu, 26 May 2011 21:02:16 +0000 (22:02 +0100)]
Doc for getopt-long's new stop-at-first-non-option option

* doc/ref/mod-getopt-long.texi (getopt-long): Mention optional keyword
  parameters.

  (getopt-long Reference): Document #:stop-at-first-non-option.

12 years agoRemove unused script-getopt.texi
Neil Jerram [Thu, 26 May 2011 20:30:05 +0000 (21:30 +0100)]
Remove unused script-getopt.texi

* doc/ref/Makefile.am (guile_TEXINFOS): Remove script-getopt.texi.

* doc/ref/script-getopt.texi: Deleted.  This file wasn't included in
  the manual, and its content is now duplicated identically in
  scheme-scripts.texi.

12 years agoEmit a 1-based line number in error messages
Neil Jerram [Thu, 12 May 2011 22:16:05 +0000 (23:16 +0100)]
Emit a 1-based line number in error messages

* module/ice-9/boot-9.scm (exception-printers): Add 1 to the 0-based
  line number.

12 years agoReveal guile-tools's inner simplicity...
Neil Jerram [Sun, 8 May 2011 21:51:07 +0000 (22:51 +0100)]
Reveal guile-tools's inner simplicity...

...by not using its own-rolled getopt, and moving the `list' function
to a separate script

* meta/guile-tools.in: Use (ice-9 getopt-long).

  (directory-files, strip-extensions, unique, find-submodules,
  list-scripts): Deleted (and moved to new `list.scm' file).

  (getopt): Deleted.

  (main): Use getopt-long.  Default to calling the `list' script if no
  script is specified.

* module/scripts/list.scm: New script.

* module/Makefile.am (SCRIPTS_SOURCES): Add list.scm.

12 years agoImplement #:stop-at-first-non-option option for getopt-long
Neil Jerram [Sun, 8 May 2011 21:21:51 +0000 (22:21 +0100)]
Implement #:stop-at-first-non-option option for getopt-long

(For use by guile-tools)

* module/ice-9/getopt-long.scm: Use (ice-9 optargs) so we can use
  define*.

  (process-options): Add stop-at-first-non-option parameter.  When
  this is true, stop processing when we hit a non-option (so long as
  that non-option isn't something that resulted from the unclumping of
  a short option group).

  (getopt-long): Add #:stop-at-first-non-option keyword; pass it on to
  process-options.

* test-suite/tests/getopt-long.test ("stop-at-first-non-option"): New
  test (for the above).

12 years agoHandle short option unclumping progressively, instead of all upfront
Neil Jerram [Thu, 26 May 2011 16:38:41 +0000 (17:38 +0100)]
Handle short option unclumping progressively, instead of all upfront

This is needed as a prerequisite for the following change that
introduces the stop-at-first-non-option option, because when that
option is used we don't know upfront how far through the command
line we should proceed with unclumping.

* module/ice-9/getopt-long.scm (expand-clumped-singles): Delete.

  (process-options): Add a loop variable to indicate how many elements
  at the start of `argument-ls' are known not to be clumped.  When we
  see a short option and this variable is <= 0, perform unclumping
  (using code that used to be in expand-clumped-singles) and loop with
  the variable > 0.

  (getopt-long): Don't call expand-clumped-singles upfront here.

12 years agoSimplify getopt-long handling of option values, esp with multiple occurrences
Neil Jerram [Sun, 8 May 2011 20:52:01 +0000 (21:52 +0100)]
Simplify getopt-long handling of option values, esp with multiple occurrences

Basically, accumulate values in the `process-options' loop variables,
instead of using set-option-spec-value!

* module/ice-9/getopt-long.scm (option-spec): Delete the `value' slot.

  (process-options): Delete `val!loop' and just use `loop' everywhere
  instead.  When adding an option spec to `found', add the
  corresponding value too; hence `found' becomes an alist, where it
  was previously a list of specs.

  (getopt-long): Use assq-ref to get values out of `found'.  Remove
  unhittable error condition for detecting an option that requires an
  explicit value, where a value wasn't supplied.  This condition is
  actually caught and handled in `process-options'.  Rewrite the end
  of the procedure much more simply.

12 years agoFix "occurrances" typos in getopt-long code and test
Neil Jerram [Sun, 8 May 2011 20:36:54 +0000 (21:36 +0100)]
Fix "occurrances" typos in getopt-long code and test

* module/ice-9/getopt-long.scm (process-options, getopt-long): Change
  to "occurrences".

* test-suite/tests/getopt-long.test ("multiple occurrences"): Same
  again.

12 years agoleave guile when reading signal pipe
Andy Wingo [Thu, 26 May 2011 16:29:05 +0000 (18:29 +0200)]
leave guile when reading signal pipe

* libguile/scmsigs.c (read_signal_pipe_data, signal_delivery_thread):
  Leave guile when reading from the signal pipe.  Hopefully that lets GC
  know not to wake up this thread.

12 years agolazily init futures worker pool
Andy Wingo [Thu, 26 May 2011 16:14:32 +0000 (18:14 +0200)]
lazily init futures worker pool

* module/ice-9/futures.scm (%workers, %create-workers!)
  (create-workers!): Define a mechanism to spawn off the future threads
  only when the first future is created.
  (make-future): Call create-workers! here.

12 years agoremove scm_newcell_count, scm_newcell2_count
Andy Wingo [Thu, 26 May 2011 13:58:42 +0000 (15:58 +0200)]
remove scm_newcell_count, scm_newcell2_count

* libguile/inline.h:
* libguile/gc.c: Remove declaration and definition of unused
  "scm_newcell_count" and "scm_newcell2_count".  Since SCM_INTERNAL is
  "extern", these symbols were not externally visible anyway, at least
  under Linux or Windows.

12 years agodeprecate scm_immutable_{double_,}cell
Andy Wingo [Thu, 26 May 2011 10:34:30 +0000 (12:34 +0200)]
deprecate scm_immutable_{double_,}cell

* libguile/inline.h:
* libguile/deprecated.h:
* libguile/deprecated.c (scm_immutable_cell, scm_immutable_double_cell):
  Deprecate these, as the GC_STUBBORN API doesn't do anything any more.

* libguile/strings.c (scm_i_c_make_symbol): Change the one use of
  scm_immutable_double_cell to scm_double_cell.

13 years agoClarify docs for SCM_UNSPECIFIED: used sometimes but not always
Mark H Weaver [Wed, 25 May 2011 16:31:21 +0000 (12:31 -0400)]
Clarify docs for SCM_UNSPECIFIED: used sometimes but not always

* doc/ref/data-rep.texi (Immediate objects): Clarify the description
  of SCM_UNSPECIFIED.  It is returned by some (but not all) expressions
  whose value is unspecified by the Scheme standard.

13 years agoout-of-tree build fix
Andy Wingo [Wed, 25 May 2011 08:27:46 +0000 (10:27 +0200)]
out-of-tree build fix

* test-suite/standalone/Makefile.am (GUILE_AUTO_COMPILE): Add srcdir to
  the environment.
* test-suite/standalone/test-import-order: Look for the modules in the
  srcdir.  Fixes out-of-tree builds.

13 years agothreadsafe addition to the goops vtable_class_map
Andy Wingo [Tue, 24 May 2011 20:52:39 +0000 (22:52 +0200)]
threadsafe addition to the goops vtable_class_map

* libguile/goops.c (scm_i_define_class_for_vtable): Lock around
  additions to the vtable_class_map.  Use is unlocked though; solving
  GOOPS parallel-initialization would be a good thing.

13 years agothreadsafe access to source properties weak hash table
Andy Wingo [Tue, 24 May 2011 20:46:09 +0000 (22:46 +0200)]
threadsafe access to source properties weak hash table

* libguile/srcprop.c: Add a lock around scm_source_whash, and use it.

13 years agoread + source properties simplification
Andy Wingo [Tue, 24 May 2011 19:25:11 +0000 (21:25 +0200)]
read + source properties simplification

* libguile/srcprop.h: Remove internal scm_source_whash declaration.
* libguile/srcprop.c (scm_i_set_source_properties_x)
  (scm_i_has_source_properties): New helpers.
  (scm_source_whash): Make static.

* libguile/read.c (scm_read_sexp): Remove register declarations here;
  let's trust the compiler.  Remove code to incrementally build up a
  copy; instead let's let scm_i_set_source_properties_x handle copying
  the expression if needed.
  (scm_read_quote, scm_read_syntax): Use scm_i_set_source_properties_x.
  (recsexpr): Remove this helper from 1996.
  (scm_read_sharp_extension): Instead of trying to recursively label
  sharp-read subforms with source properties, just label the outside
  form and rely on the macro-expander to propagate it down.

13 years agosafely access the trampoline weak map
Andy Wingo [Tue, 24 May 2011 19:09:24 +0000 (21:09 +0200)]
safely access the trampoline weak map

* libguile/smob.c (scm_i_smob_apply_trampoline): Protect the trampoline
  weak map with a mutex.

13 years agoset-procedure-property! threadsafety issue
Andy Wingo [Mon, 23 May 2011 20:38:13 +0000 (22:38 +0200)]
set-procedure-property! threadsafety issue

* libguile/procprop.c (scm_set_procedure_property_x): Fix a threadsafety
  bug pointed out by Ken Raeburn.

13 years agoreally threadsafe access to symbol table
Andy Wingo [Mon, 23 May 2011 20:24:27 +0000 (22:24 +0200)]
really threadsafe access to symbol table

* libguile/symbols.c (symbols_lock): Rename from intern_lock.
  (lookup_interned_symbol, lookup_interned_latin1_symbol): Instead of
  faith-based programming, just use the mutex.  Though I haven't seen
  this break, Ken is right!

13 years agoDon't call scm_lock_mutex and scm_unlock_mutex via pointer of wrong type
Mark H Weaver [Sun, 22 May 2011 19:23:27 +0000 (15:23 -0400)]
Don't call scm_lock_mutex and scm_unlock_mutex via pointer of wrong type

* libguile/threads.c (lock_mutex_return_void, unlock_mutex_return_void):
  New static functions that simply call scm_lock_mutex and
  scm_unlock_mutex, respectively, but return void instead of SCM.

  (scm_dynwind_lock_mutex): Pass unlock_mutex_return_void to
  scm_dynwind_unwind_handler_with_scm, and lock_mutex_return_void to
  scm_dynwind_rewind_handler_with_scm.  Previously, we passed
  scm_unlock_mutex and scm_lock_mutex (which return SCM), but the
  scm_dynwind_* functions expect pointers to functions which return
  void.  When SCM is of type union, this changes the calling conventions
  of the functions on some platforms (e.g. GCC 4.5.2 and 4.5.3 on x86).

13 years agofix define-module ordering
Andy Wingo [Sat, 21 May 2011 16:29:03 +0000 (18:29 +0200)]
fix define-module ordering

* module/ice-9/boot-9.scm (define-module): Fix to load the #:use-module
  clauses in the order in which they appear in the define-module form.
  Thanks to Jan Nieuwenhuizen for the report.

* test-suite/standalone/test-import-order: Add new test that
  define-module and use-modules resolve the interface in the right
  order.

* test-suite/standalone/Makefile.am:
* test-suite/standalone/test-import-order-a.scm:
* test-suite/standalone/test-import-order-b.scm:
* test-suite/standalone/test-import-order-c.scm:
* test-suite/standalone/test-import-order-d.scm: Aux files.

13 years ago(syntax foo) -> #'foo in goops
Andy Wingo [Sat, 21 May 2011 11:12:44 +0000 (13:12 +0200)]
(syntax foo) -> #'foo in goops

* module/oop/goops.scm: Change instances of (syntax foo) to #'foo.

13 years agoclocktime freebsd portability
Andy Wingo [Fri, 20 May 2011 15:41:06 +0000 (17:41 +0200)]
clocktime freebsd portability

* libguile/stime.c (HAVE_POSIX_CPUTIME): Hack around buggy FreeBSD
  implementation of _POSIX_CPUTIME.

13 years agofix compile error in mingw fstat socket detection
Andy Wingo [Fri, 20 May 2011 10:50:08 +0000 (12:50 +0200)]
fix compile error in mingw fstat socket detection

* libguile/filesys.c (fstat_Win32) [__MINGW32__]: Apparently there is no
  _S_IFSOCK on mingw32.  Thanks to Volker Grabsch for the report.

13 years agogen-scmconfig cross-compilation fix
Andy Wingo [Fri, 20 May 2011 10:26:53 +0000 (12:26 +0200)]
gen-scmconfig cross-compilation fix

* libguile/Makefile.am (gen-scmconfig.$(OBJEXT)): Fix
  cross-compilation.  Thanks to Volker Grabsch for the report!

13 years agofix documentation for option-set! syntaxen
Andy Wingo [Fri, 20 May 2011 09:54:46 +0000 (11:54 +0200)]
fix documentation for option-set! syntaxen

* doc/ref/api-evaluation.texi (Scheme Read): Note that read-set! is
  syntax.
  (Scheme Write): Likewise for print-set!.

* doc/ref/api-io.texi (Writing): Remove reference to
  print-options-interface.

* doc/ref/repl-modules.texi (Readline Options): Update, and add entries
  for readline-options, readline-set! et al.

13 years agoImprove R6RS conformance wrt. conditions in the I/O libraries
Andreas Rottmann [Sat, 14 May 2011 17:29:26 +0000 (19:29 +0200)]
Improve R6RS conformance wrt. conditions in the I/O libraries

* module/rnrs/io/ports.scm (open-file-output-port): Handle `no-fail'
  file option.
  (with-i/o-filename-conditions): Use `with-throw-handler' instead of `catch'.
  (with-i/o-port-error,
  with-textual-output-conditions. with-textual-input-conditions): New
  exception-conversion helpers.
  (put-char, put-datum, put-string, display): Use
  `with-textual-output-conditions' instead of `with-i/o-encoding-error'
  to get proper conditions in case of write errors.
  (get-char, get-datum, get-line, get-string-all, lookahead-char):
  Likewise, for the input case.

* test-suite/tests/r6rs-ports.test (pass-if-condition, test-file,
  make-failing-port): New helpers.
  ("8.2.10 Output ports"): Add some tests for `open-file-output-port'.
  ("8.2.9 Textual Input"): Add tests read error conditions.
  ("8.2.12 Textual Output"): Add tests for write error conditions.
  ("8.3 Simple I/O"): Add tests for conditions, `call-with-input-file'
  and `call-with-output-file'.

13 years agorefactor do_thread_exit cleanup handler invocation
Andy Wingo [Fri, 13 May 2011 11:33:59 +0000 (13:33 +0200)]
refactor do_thread_exit cleanup handler invocation

* libguile/threads.c (do_thread_exit): Redo cleanup handler call so as
  not to shove a SCM into a pointer.

13 years agoi18n error return type fix
Andy Wingo [Fri, 13 May 2011 11:29:48 +0000 (13:29 +0200)]
i18n error return type fix

* libguile/i18n.c (chr_to_case, str_to_case): Return #f in error case
  instead of 0.

13 years agoPTR2SCM and SCM2PTR in inline.h
Andy Wingo [Fri, 13 May 2011 11:29:15 +0000 (13:29 +0200)]
PTR2SCM and SCM2PTR in inline.h

* libguile/inline.h (scm_cell, scm_immutable_cell):
  (scm_double_cell, scm_immutable_double_cell):
  (scm_words): Be more consistent in use of PTR2SCM and SCM2PTR.

13 years agocompile-time assertion in net_db uses constant expressions
Andy Wingo [Fri, 13 May 2011 11:25:14 +0000 (13:25 +0200)]
compile-time assertion in net_db uses constant expressions

* libguile/net_db.c: Use constant expressions for EAI_BADFLAGS and
  AI_ALL representations.

13 years agoports.c uninitialized static SCM values are #f, not 0
Andy Wingo [Fri, 13 May 2011 11:24:29 +0000 (13:24 +0200)]
ports.c uninitialized static SCM values are #f, not 0

* libguile/ports.c: (scm_current_input_port, scm_current_output_port)
  (scm_current_error_port): Use #f as the uninitialized value instead of
  0.

13 years agoscm_port_for_each fix
Andy Wingo [Fri, 13 May 2011 11:23:12 +0000 (13:23 +0200)]
scm_port_for_each fix

* libguile/ports.c (scm_port_for_each): Inline the call to
  scm_c_port_for_each, to avoid type errors.

13 years agoweak_bucket_assoc tweak
Andy Wingo [Fri, 13 May 2011 11:21:51 +0000 (13:21 +0200)]
weak_bucket_assoc tweak

* libguile/hashtab.c (weak_bucket_assoc): Change assertion to be a check
  and abort, and so that only calls GC_is_visible if the check fails.

13 years agovm_make_boot_program initializer fix
Andy Wingo [Fri, 13 May 2011 11:20:39 +0000 (13:20 +0200)]
vm_make_boot_program initializer fix

* libguile/vm.c (vm_make_boot_program): Use #f as the "I don't have a
  program" value.

13 years agocorrect use of SCM2PTR in SCM_I_REGISTER_DISAPPEARING_LINK forms
Andy Wingo [Fri, 13 May 2011 11:18:11 +0000 (13:18 +0200)]
correct use of SCM2PTR in SCM_I_REGISTER_DISAPPEARING_LINK forms

* libguile/hashtab.c (set_weak_cdr):
* libguile/vectors.c (scm_c_vector_set_x):
* libguile/weaks.c (scm_weak_car_pair, scm_weak_cdr_pair):
  (scm_doubly_weak_pair): Use SCM2PTR for the target of
  SCM_I_REGISTER_DISAPPEARING_LINK calls.

13 years agoscm_is_false instead of == SCM_BOOL_F; also is_null, is_true, etc
Andy Wingo [Fri, 13 May 2011 11:04:49 +0000 (13:04 +0200)]
scm_is_false instead of == SCM_BOOL_F; also is_null, is_true, etc

* libguile/deprecation.c (scm_issue_deprecation_warning)
* libguile/eval.c (CAPTURE_ENV):
* libguile/goops.c (make_dispatch_procedure, make_class_from_symbol):
  (create_smob_classes):
* libguile/guardians.c (finalize_guarded, scm_i_get_one_zombie):
* libguile/hashtab.c (scm_fixup_weak_alist, scm_internal_hash_fold):
* libguile/i18n.c (scm_nl_langinfo)
* libguile/load.c (scm_primitive_load)
* libguile/posix.c (scm_setrlimit)
* libguile/socket.c (scm_to_sockaddr):
* libguile/srcprop.c (scm_make_srcprops): Use scm_is_false / scm_is_true
  / scm_is_null instead of comparing against SCM_BOOL_F et al.

13 years agomore care regarding SCM_PACK and SCM_UNPACK
Andy Wingo [Fri, 13 May 2011 10:51:56 +0000 (12:51 +0200)]
more care regarding SCM_PACK and SCM_UNPACK

* libguile/control.c (reify_partial_continuation):
* libguile/eval.c (RETURN_BOOT_CLOSURE):
* libguile/frames.c (scm_frame_num_locals, scm_frame_local_ref)
  (scm_frame_local_set_x)
* libguile/frames.h (SCM_FRAME_SET_RETURN_ADDRESS):
  (SCM_FRAME_SET_MV_RETURN_ADDRESS, SCM_FRAME_SET_DYNAMIC_LINK):
* libguile/goops.c (scm_class_of, scm_primitive_generic_generic)
  (scm_c_extend_primitive_generic, compute_getters_n_setters)
  (scm_sys_initialize_object):
* libguile/guardians.c (finalize_guarded):
* libguile/list.c (SCM_I_CONS):
* libguile/macros.c (scm_i_make_primitive_macro)
  (scm_make_syntax_transformer):
* libguile/memoize.c (MAKMEMO, SCM_MAKE_MEMOIZER)
  (SCM_MAKE_REST_MEMOIZER):
* libguile/modules.c (scm_module_reverse_lookup)
* libguile/print.c (iprin1):
* libguile/promises.c (scm_make_promise)
* libguile/srcprop.c (scm_make_srcprops):
* libguile/vectors.c (scm_c_vector_ref):
* libguile/vm-engine.c (vm_engine)
* libguile/vm-i-scheme.c (REL, add1, sub1):
* libguile/vm-i-system.c (new_frame, call_cc)
* libguile/weaks.h (SCM_WEAK_PAIR_WORD_DELETED_P): Be more careful about
  SCM_PACK / SCM_UNPACK.

13 years agoscm_is_eq for SCM vals, not == or !=
Andy Wingo [Fri, 13 May 2011 10:42:01 +0000 (12:42 +0200)]
scm_is_eq for SCM vals, not == or !=

* libguile/bytevectors.c (scm_make_bytevector, STRING_TO_UTF)
  (UTF_TO_STRING):
* libguile/continuations.c (scm_i_check_continuation):
* libguile/expand.h (SCM_EXPANDED_P):
* libguile/fluids.c (scm_i_make_with_fluids):
* libguile/generalized-vectors.c (scm_make_generalized_vector):
* libguile/goops.c (SCM_GOOPS_UNBOUNDP, slot_definition_using_name):
  (scm_c_extend_primitive_generic, more_specificp, scm_make)
* libguile/i18n.c (SCM_VALIDATE_OPTIONAL_LOCALE_COPY):
  (scm_locale_string_to_integer)
* libguile/modules.c (resolve_duplicate_binding):
  (scm_module_reverse_lookup)
* libguile/posix.c (scm_to_resource):
* libguile/r6rs-ports.c (scm_put_bytevector):
* libguile/socket.c (scm_connect, scm_bind, scm_sendto
* libguile/stacks.c (find_prompt):
* libguile/variable.c (scm_variable_ref, scm_variable_bound_p):
* libguile/vm-engine.h (ASSERT_BOUND_VARIABLE, ASSERT_BOUND)
* libguile/vm-i-system.c (VARIABLE_BOUNDP, local_bound)
  (long_local_bound, fluid_ref): Use scm_is_eq to compare, not == / !=.

13 years agodeprecate scm_internal_dynamic_wind
Andy Wingo [Fri, 13 May 2011 10:16:56 +0000 (12:16 +0200)]
deprecate scm_internal_dynamic_wind

* libguile/dynwind.c:
* libguile/dynwind.h:

* libguile/deprecated.h (scm_t_inner):
* libguile/deprecated.c (scm_internal_dynamic_wind): Deprecate, as the
  scm_dynwind API is better, and this API encourages users to stuff SCM
  values into pointers.

13 years agoexcise scm_internal_dynamic_wind from goops.c
Andy Wingo [Fri, 13 May 2011 10:24:04 +0000 (12:24 +0200)]
excise scm_internal_dynamic_wind from goops.c

* libguile/goops.c (go_to_hell, go_to_heaven, purgatory):
  (scm_change_object_class): Rewrite to use scm_dynwind_begin instead of
  scm_dynamic_wind.

13 years agoasync.c refactor
Andy Wingo [Fri, 13 May 2011 10:15:42 +0000 (12:15 +0200)]
async.c refactor

* libguile/async.c (increase_block, decrease_block): Write more
  clearly.
  (scm_dynwind_block_asyncs, scm_dynwind_unblock_asyncs): Move
  definitions up.
  (scm_call_with_blocked_asyncs, scm_c_call_with_blocked_asyncs)
  (scm_call_with_unblocked_asyncs, scm_c_call_with_unblocked_asyncs):
  Implement in terms of scm_dynwind_{un,}block_asyncs, so that we don't
  stuff SCM values into pointers.

13 years agofix type errors
Andy Wingo [Fri, 13 May 2011 08:32:46 +0000 (10:32 +0200)]
fix type errors

* libguile/numbers.c (scm_logand): Fix a type error (comparing a SCM
  against an int, when we really wanted to compare the unpacked
  fixnum).

* libguile/ports.c (scm_i_set_conversion_strategy_x): Check
  scm_conversion_strategy_init, not scm_conversion_strategy.

* libguile/read.c (recsexpr): Fix loops to avoid strange test of SCM
  values.

13 years agoallow iflags to be constant expressions with typing-strictness==2
Andy Wingo [Fri, 13 May 2011 08:19:48 +0000 (10:19 +0200)]
allow iflags to be constant expressions with typing-strictness==2

* libguile/tags.h (SCM_MAKE_ITAG8_BITS): New helper, produces a
  scm_t_bits instead of a SCM, because SCM_UNPACK is not a constant
  expression with SCM_DEBUG_TYPING_STRICTNESS==2.
  (SCM_MAKIFLAG_BITS): Remove SCM_MAKIFLAG, and replace with this, which
  returns bits.
  (SCM_BOOL_F_BITS, SCM_ELISP_NIL_BITS, SCM_EOL_BITS, SCM_BOOL_T_BITS):
  (SCM_UNSPECIFIED_BITS, SCM_UNDEFINED_BITS, SCM_EOF_VAL_BITS):
  (SCM_UNBOUND_BITS): New definitions.  Defined SCM_BOOL_F, etc in terms
  of them.
  (SCM_XXX_ANOTHER_BOOLEAN_DONT_USE_0):
  (SCM_XXX_ANOTHER_BOOLEAN_DONT_USE_1):
  (SCM_XXX_ANOTHER_BOOLEAN_DONT_USE_2):
  (SCM_XXX_ANOTHER_LISP_FALSE_DONT_USE): Be bits instead of SCM values.
  (SCM_BITS_DIFFER_IN_EXACTLY_ONE_BIT_POSITION):
  (SCM_BITS_DIFFER_IN_EXACTLY_TWO_BIT_POSITIONS): Rename from
  SCM_VALUES_DIFFER_..., and take unpacked bits as the args.

* libguile/boolean.c: Update verify block to use
  SCM_BITS_DIFFER_IN_EXACTLY_TWO_BIT_POSITIONS et al.

* libguile/debug.c (scm_debug_opts):
* libguile/print.c (scm_print_opts):
* libguile/read.c (scm_read_opts): Use iflags bits for initializers.

* libguile/hash.c (scm_hasher): Use _BITS for iflags as case labels.

* libguile/pairs.c: Nil/null compile-time check uses
  SCM_ELISP_NIL_BITS.

13 years agofix static allocation with debug_typing_strictness==2
Andy Wingo [Thu, 12 May 2011 21:29:16 +0000 (23:29 +0200)]
fix static allocation with debug_typing_strictness==2

* libguile/tags.h (SCM): For SCM_DEBUG_TYPING_STRICTNESS==2, give the
  union a tag, and use C99 compound literals to construct the value.
  This allows SCM_PACK to be a constant expression.

* libguile/snarf.h: Allow SCM_SUPPORT_STATIC_ALLOCATION for
  SCM_DEBUG_TYPING_STRICTNESS==2.
  (SCM_IMMUTABLE_STRING): Properly parenthesize the string length.
  (SCM_STATIC_PROGRAM): Fix for SCM_DEBUG_TYPING_STRICTNESS==2.

13 years agoadd missing SCM_DEPRECATED annotation
Andy Wingo [Thu, 12 May 2011 11:37:57 +0000 (13:37 +0200)]
add missing SCM_DEPRECATED annotation

* libguile/deprecated.h: Use SCM_DEPRECATED instead of SCM_API for the
  asinh, acosh, atanh, and atan2 wrappers.

13 years agoall deprecated routines emit warnings
Andy Wingo [Thu, 12 May 2011 11:07:19 +0000 (13:07 +0200)]
all deprecated routines emit warnings

* module/ice-9/deprecated.scm (substring-move-left!)
  (substring-move-right!, dynamic-maybe-call, dynamic-maybe-link)
  (try-module-linked, try-module-dynamic-link)
  ($asinh, $acosh, $atanh, $sqrt, $abs, $exp, $log, $sin, $cos)
  ($tan, $asin, $acos, $atan, $sinh, $cosh, $tanh)
  (process-define-module): Add deprecation warnings.

13 years agofix some duplication in object tables
Andy Wingo [Sun, 8 May 2011 20:55:05 +0000 (22:55 +0200)]
fix some duplication in object tables

* module/language/glil/compile-assembly.scm (build-object-table): Don't
  add the same value to an object table twice.

13 years agoAdd `vhash-fold-right'.
Ludovic Courtès [Sun, 8 May 2011 16:19:52 +0000 (18:19 +0200)]
Add `vhash-fold-right'.

* module/ice-9/vlist.scm (vhash-fold-right): New procedure.

* test-suite/tests/vlist.test ("vhash")["vhash-fold-right"]: New test.

* doc/ref/api-compound.texi (VHashes): Document `vhash-fold-right'.

13 years agoOptimize `vlist-fold-right'.
Ludovic Courtès [Sun, 8 May 2011 16:15:10 +0000 (18:15 +0200)]
Optimize `vlist-fold-right'.

* module/ice-9/vlist.scm (vlist-fold-right): Avoid `vlist-reverse' and
  instead `vlist-ref' individual elements.  The result is about twice
  faster.  Thanks Andy for suggesting it indirectly.  :-)

13 years agoadd map and for-each benchmarks
Andy Wingo [Sun, 8 May 2011 15:27:41 +0000 (17:27 +0200)]
add map and for-each benchmarks

* benchmark-suite/benchmarks/srfi-1.bm ("map", "for-each"): Add
  benchmarks.

13 years agoFix small integer return value packing on big endian machines.
Ludovic Courtès [Sun, 8 May 2011 15:21:59 +0000 (17:21 +0200)]
Fix small integer return value packing on big endian machines.

* libguile/foreign.c (pack): Add `return_value_p' parameter.  Update
  callers.
  When RETURN_VALUE_P is true, assume LOC points to an `ffi_arg', and
  cast its results to the relevant type.  This fixes packing of integer
  return values smaller than `int' on SPARC64 and PowerPC64.  Reported
  by Nelson H. F. Beebe <beebe@math.utah.edu>.

13 years agoMake the definition of `scm_read_shebang' match its declaration.
Ludovic Courtès [Sun, 8 May 2011 14:25:01 +0000 (16:25 +0200)]
Make the definition of `scm_read_shebang' match its declaration.

* libguile/read.c (scm_read_shebang): Remove the `inline' keyword.

13 years agocompile-assembly: cleanup
Andy Wingo [Sun, 8 May 2011 14:38:32 +0000 (16:38 +0200)]
compile-assembly: cleanup

* module/language/glil/compile-assembly.scm: Clean up code for
  subprograms (not needed, we just cache the glil) and object alists
  (replaced by constants tables).

13 years agocompile-assembly: use file-level constants table
Andy Wingo [Sun, 8 May 2011 14:37:47 +0000 (16:37 +0200)]
compile-assembly: use file-level constants table

* module/language/glil/compile-assembly.scm (compile-assembly): Rework
  to handle toplevel-specific code generation here, instead of in
  glil->assembly.  Specifically, here we build a global constant table,
  and arrange for it to be the objtable of the toplevel thunk.

  (compile-program): New helper, compiles a <glil-program> and returns
  just the (load-program ...) form.

  (compile-objtable): New helper, generates assembly to build an object
  table, using some other constants table, and possibly recursing to
  `compile-program' for cached GLIL programs.

  (glil->assembly): Simplify, removing the toplevel? argument, and
  replacing the object alist with an objtable computed in a previous
  pass.  Adapt to the new form of the objtable, and to use
  compile-program and compile-objtable.

13 years agocompile-assembly: add dump-constants, a new helper
Andy Wingo [Sun, 8 May 2011 14:31:18 +0000 (16:31 +0200)]
compile-assembly: add dump-constants, a new helper

* module/language/glil/compile-assembly.scm (dump-constants): New
  helper.  Generates bytecode that will result in a vector for the
  global object table being pushed on the stack.  The items in the
  global object table will share state as much as possible.

13 years agocompile-assembly: add build-constant-store, build-object-table
Andy Wingo [Sun, 8 May 2011 14:15:25 +0000 (16:15 +0200)]
compile-assembly: add build-constant-store, build-object-table

* module/language/glil/compile-assembly.scm (immediate?): New helper.
  (build-constant-store): New helper.  Walks the GLIL tree and builds up
  a constant table, as a vhash.
  (build-object-table): Another helper, builds a constant table for a
  given GLIL program.

13 years agocompile-assembly: make-meta refactor
Andy Wingo [Sun, 8 May 2011 14:13:41 +0000 (16:13 +0200)]
compile-assembly: make-meta refactor

* module/language/glil/compile-assembly.scm (make-meta): Avoid going
  through the compiler.

13 years agocompile-assembly: add traversal helpers
Andy Wingo [Sun, 8 May 2011 14:09:22 +0000 (16:09 +0200)]
compile-assembly: add traversal helpers

* module/language/glil/compile-assembly.scm (vhash-fold-right2):
  (fold2, vector-fold2): Add some traversal helpers that we'll use in
  the next commit.

13 years agofix `hash' for inf and nan
Andy Wingo [Sun, 8 May 2011 14:05:27 +0000 (16:05 +0200)]
fix `hash' for inf and nan

* libguile/hash.c (scm_hasher): Fix to work on inf and nan.
* test-suite/tests/hash.test ("hash"): Add tests.

13 years agoMake the R6RS simple I/O library use conditions
Andreas Rottmann [Sat, 7 May 2011 21:40:14 +0000 (23:40 +0200)]
Make the R6RS simple I/O library use conditions

* module/rnrs/io/ports.scm (display): Implement as an
  exception-converting wrapper around Guile's core display.
* module/rnrs/io/simple.scm: Don't export Guile's corresponding core
  procedures, but use `(rnrs io ports)' instead.  This way, we get the
  conditions required by R6RS raised.

* doc/ref/r6rs.texi (rnrs io simple): Mention that these procedures are
  supposed to raise R6RS conditions.

13 years agoFix `get_utf8_codepoint' to not consume valid starting bytes.
Ludovic Courtès [Sat, 7 May 2011 20:46:38 +0000 (22:46 +0200)]
Fix `get_utf8_codepoint' to not consume valid starting bytes.

Thanks to Mark H. Weaver for pointing this out.

* libguile/ports.c (CONSUME_PEEKED_BYTE): New macro.
  (get_utf8_codepoint): New variable `pt'.  Use
  `scm_peek_byte_or_eof'/`CONSUME_PEEKED_BYTE' pairs instead of
  `scm_get_byte_or_eof'.

* test-suite/tests/ports.test ("string ports")[#xc2 #x41 #x42, #xe0 #xa0
  #x41 #x42, #xf0 #x88 #x88 #x88]: Fix to conform to Unicode 6.0.0.
  [#xe0 #x88 #x88]: Remove test.
  [#xf0 #x80 #x80 #x41]: New test.

13 years agoAdd `scm_peek_byte_or_eof'.
Ludovic Courtès [Sat, 7 May 2011 20:41:32 +0000 (22:41 +0200)]
Add `scm_peek_byte_or_eof'.

* libguile/inline.h (scm_get_byte_or_eof): Add `SCM_UNLIKELY' for EOF.
  (scm_peek_byte_or_eof): New function.

* libguile/r6rs-ports.c (scm_lookahead_u8): Use `scm_peek_byte_or_eof'.

13 years agoMore completely document the `(rnrs io ports)' library
Andreas Rottmann [Sat, 7 May 2011 20:30:40 +0000 (22:30 +0200)]
More completely document the `(rnrs io ports)' library

* doc/ref/api-io.texi (R6RS I/O Ports): Transcribe missing parts from
  the R6RS document.

13 years agofix srfi-1 map-in-order definition
Andy Wingo [Sat, 7 May 2011 09:31:38 +0000 (11:31 +0200)]
fix srfi-1 map-in-order definition

* module/srfi/srfi-1.scm (map-in-order): As we are not extending the
  core `map' binding, actually make a new `map-in-order' alias here.
  Fixes fresh builds.

13 years agoSpecial-case UTF-8 ports to bypass `iconv' entirely.
Ludovic Courtès [Fri, 6 May 2011 15:54:09 +0000 (17:54 +0200)]
Special-case UTF-8 ports to bypass `iconv' entirely.

* libguile/ports.c (update_port_lf): Handle EOF.
  (get_utf8_codepoint, get_iconv_codepoint): New functions.
  (get_codepoint): Use them.
  (scm_i_set_port_encoding_x): Don't open conversion descriptors when
  ENCODING is "UTF-8".

* libguile/print.c (display_string_as_utf8, display_string_using_iconv):
  New functions.
  (display_string): Use them.

* test-suite/tests/ports.test ("string ports")[#xc2 #x41 #x42]: Add a
  note that this is not the wrong behavior per Unicode 6.0.0.

13 years agoFix `foreign.test' for big endian machines.
Ludovic Courtès [Fri, 6 May 2011 15:43:37 +0000 (17:43 +0200)]
Fix `foreign.test' for big endian machines.

* test-suite/tests/foreign.test ("pointer<->bytevector")["pointer from
  bits", "dereference-pointer"]: Fix iteration order for big endian
  machines.

13 years agoavoid tls gets when handling interrupts in the vm
Andy Wingo [Thu, 5 May 2011 22:17:35 +0000 (00:17 +0200)]
avoid tls gets when handling interrupts in the vm

* libguile/__scm.h (SCM_ASYNC_TICK_WITH_CODE): Redefine to take a
  scm_i_thread* as well.  OK to do because it's within a
  BUILDING_LIBGUILE block.

* libguile/vm-engine.c (vm_engine): Cache the scm_i_thread* instead of
  the dynstate, so we can use the thread for ticks.

* libguile/vm-engine.h (VM_HANDLE_INTERRUPTS): Tick with the
  scm_i_thread* local var, to avoid excessive tls calls.

* libguile/vm-i-system.c: Fix dynstate users to use
  current_thread->dynamic_state.

13 years agomap and for-each in scheme
Andy Wingo [Thu, 5 May 2011 21:07:23 +0000 (23:07 +0200)]
map and for-each in scheme

* module/ice-9/boot-9.scm (map, for-each): Implement in Scheme instead
  of C.  There are boot versions before `cond' is defined.
  (map-in-order): Define this alias here instead of in evalext.h.

* libguile/eval.c: Stub out the map and for-each definitions to just
  call into Scheme.

* libguile/evalext.c: Remove map-in-order definition.

* module/srfi/srfi-1.scm: Replace all calls to map1 with calls to map.
  (map, for-each): Define implementations here, in Scheme, instead of in
  C.

* test-suite/tests/eval.test (exception:wrong-length, "map"): Update the
  expected exception for mapping over lists of different lengths.

* libguile/srfi-1.h:
* libguile/srfi-1.c: Remove map and for-each definitions.  Remove the
  bit that extended the core `map' primitive with another method: the
  right way to do that is with modules.

13 years agopsyntax simplification
Andy Wingo [Thu, 5 May 2011 14:28:28 +0000 (16:28 +0200)]
psyntax simplification

* module/ice-9/psyntax.scm (strip): Inline the and-map* definition to
  its one call site.

* module/ice-9/psyntax-pp.scm: Regenerate.

13 years agoscm_mem[qv] optimization
Andy Wingo [Thu, 5 May 2011 13:13:08 +0000 (15:13 +0200)]
scm_mem[qv] optimization

* libguile/list.c (scm_memq, scm_memv): Inline the tortoise/hare check
  that scm_ilength does, via SCM_VALIDATE_LIST, into the memq/memv
  bodies.  Avoids traversing these lists twice.

13 years agoGenerate `escape' and `substitute' port decoding tests.
Ludovic Courtès [Thu, 5 May 2011 15:56:47 +0000 (17:56 +0200)]
Generate `escape' and `substitute' port decoding tests.

13 years agoVM tweaks
Andy Wingo [Thu, 5 May 2011 12:04:23 +0000 (14:04 +0200)]
VM tweaks

* libguile/vm-engine.c (VM_CHECK_OBJECT, VM_CHECK_FREE_VARIABLES): Set
  to 0 for both engines.  These are really internal debugging variables,
  which don't affect user-visible features, provided that the compiler
  is correct of course.
  (VM_CHECK_UNDERFLOW): New var, also off by default: whether to check
  for stack underflow when popping values.
  (vm_engine): Don't declare object_count if we are not checking object
  table accesses.

* libguile/vm-engine.h (CACHE_PROGRAM): Don't muck with object_count
  if we are not checking object table accesses.
  (CHECK_UNDERFLOW, PRE_CHECK_UNDERFLOW): Nop out if we are not checking
  underflow.
  (POP2, POP3): New macros which check for underflow before popping more
  than one value.

* libguile/vm-i-loader.c (load_array):
* libguile/vm-i-scheme.c (set_car, set_cdr, vector_set, slot_set)
  (BV_SET_WITH_ENDIANNESS, BV_FIXABLE_INT_SET, BV_INT_SET)
  (BV_FLOAT_SET):
* libguile/vm-i-system.c (partial_cont_call, fix_closure, prompt)
  (fluid_set): Use POP2 / POP3.
  (local_set, long_local_set): Pop to locals instead of using values on
  the stack then dropping; allows for underflow to be checked before the
  value is accessed.
  (BR): Don't NULLSTACK or DROP after the operation.
  (br_if, br_if_not, br_if_eq, br_if_not_eq, br_if_null)
  (br_if_not_null): Pop to locals before doing the compare and jump.

13 years agosrfi-1 `member' in scheme, inlines to memq / memv in some cases
Andy Wingo [Thu, 5 May 2011 10:59:07 +0000 (12:59 +0200)]
srfi-1 `member' in scheme, inlines to memq / memv in some cases

* libguile/srfi-1.c:
* libguile/srfi-1.h (scm_srfi1_member): Move implementation to Scheme.

* module/srfi/srfi-1.scm (member): Implement here, with the inlining
  cases for eq? and eqv?.  Speeds up a compiled bootstrap of
  psyntax.scm, because lset-adjoin inlines to the memq case.
  (lset<=): Reindent.

  (lset-adjoin, lset-union): If the comparator is eq? or eqv?, just pass
  it through to `member', so we inline to memq / memv.  Use something
  closer to the reference implementations.

13 years agospeed up compile-bytecode
Andy Wingo [Thu, 5 May 2011 09:22:42 +0000 (11:22 +0200)]
speed up compile-bytecode

* module/language/assembly/compile-bytecode.scm (compile-bytecode):
  Rewrite to fill a bytevector directly, instead of using bytevector
  ports.  `write-bytecode' itself is still present and almost the same
  as before; it's just that `write-byte' et al now inline the effect of
  writing a byte to a binary port.

* test-suite/tests/asm-to-bytecode.test (comp-test): Refactor to use
  public interfaces.

13 years agosilly "optimization" in (language assembly)
Andy Wingo [Thu, 5 May 2011 08:09:48 +0000 (10:09 +0200)]
silly "optimization" in (language assembly)

* module/language/assembly.scm (byte-length): Silly, minor tweak: put
  the fixed-length instruction case first.  Seems to shave some 10% off
  the time compiling psyntax.scm (when the whole rest of the system is
  compiled, of course).

13 years agoadd gcprof
Andy Wingo [Thu, 5 May 2011 08:08:29 +0000 (10:08 +0200)]
add gcprof

* module/statprof.scm (gcprof): New variant of statprof; instead of
  being driven by setitimer, this one is driven by the after-gc-hook.

13 years agominor statprof tweaks
Andy Wingo [Thu, 5 May 2011 07:59:59 +0000 (09:59 +0200)]
minor statprof tweaks

* module/statprof.scm (statprof-reset): Make full-stacks? into an
  optional arg instead of doing the rest arg dance.
  (statprof-display): Format gc-time-taken appropriately.

13 years agoAutomatically generate `peek-char' decoding error tests.
Ludovic Courtès [Wed, 4 May 2011 21:52:53 +0000 (23:52 +0200)]
Automatically generate `peek-char' decoding error tests.

* test-suite/tests/ports.test ("string ports")[make-peek+read-checks]:
  New macro.
  [test-decoding-error]: Change to take a list of expected characters or
  conditions.  Use `make-peek+read-checks' to generate a `peek-char'
  test.
  [(#xc2 #x41 #x42), (#xe0 #xa0 #x41 #x42)]: New tests.

13 years agomeasure time spent in gc
Andy Wingo [Wed, 4 May 2011 16:48:02 +0000 (18:48 +0200)]
measure time spent in gc

* libguile/gc.c (scm_gc_stats): Set the gc-time-taken entry to our
  recorded value.
  (start_gc_timer, accumulate_gc_timer, scm_init_gc): Arrange to record
  a conservative estimate of time spent in GC.

13 years agoscm_c_get_internal_run_time is more precise
Andy Wingo [Wed, 4 May 2011 18:15:23 +0000 (20:15 +0200)]
scm_c_get_internal_run_time is more precise

* libguile/stime.h (SCM_TIME_UNITS_PER_SECOND): Redefine to point to a C
  variable instead of being a pure preprocessor thing.  This has the
  possibility to break existing compiled C extensions' interpretation of
  the internal-time-units-per-second, but hopefully there's no too much
  of that code out there, and in the worst case they can just
  recompile.  Scheme code will get it right without the need to
  recompile.

* libguile/stime.c (TIME_UNITS_PER_SECOND): New local define, and
  increase to nanosecond resolution if we are on a system in which this
  is useful and practical.
  (time_from_seconds_and_nanoseconds): New helper.
  (get_internal_real_time, get_internal_run_time): New global vars:
  function pointers.
  (get_internal_real_time_posix_timer):
  (get_internal_run_time_posix_timer):
  (get_internal_real_time_gettimeofday):
  (get_internal_run_time_times):
  (get_internal_real_time_fallback): Various implementations.
  (scm_get_internal_real_time): Return the get_internal_real_time()
  result.
  (scm_c_get_internal_run_time): Likewise.
  (scm_gettimeofday): No need for a critical section, and remove
  obsolete ftime block.
  (scm_init_stime): Init all of the new time bases, and decide on
  implementations of real time and run time accessors.

13 years agobuild support for detecting clock_gettime, with -lrt if needed
Andy Wingo [Wed, 4 May 2011 18:16:55 +0000 (20:16 +0200)]
build support for detecting clock_gettime, with -lrt if needed

* acinclude.m4 (gl_CLOCK_TIME):
* configure.ac: Locally include gl_CLOCK_TIME.  To be fixed properly
  when gnulib updates their license to reflect the actual BSD state of
  things.
* libguile/Makefile.am (libguile_@GUILE_EFFECTIVE_VERSION@_la_LDFLAGS):
  Add -lrt for clock_gettime, if needed.
* meta/guile-2.0-uninstalled.pc.in:
* meta/guile-2.0.pc.in: Likewise, in Libs.private.