Note especially that the variable 'i' has two different uses in this
function, and they get confused.
* libguile/read.c (scm_i_scan_for_encoding): cleanup
* libguile/ports.c (scm_getc): Provide `u32_conv_from_encoding' with the
RESULT_BUF stack-allocated buffer to avoid heap allocation.
(find_valid_encoding): Likewise.
(scm_ungetc): Ditto with `u32_conv_to_encoding'.
* module/system/repl/error-handling.scm (call-with-error-handling):
Previous post-error changed to "report"; now "catch", the default,
doesn't re-print the error.
* libguile/vm-engine.c: Add func_name local, for error reporting.
(vm_error_apply_to_non_list): New error case.
(vm_error_wrong_type_arg): Remove this generic error case.
(vm_error_wrong_type_apply): Remove FUNC_NAME -- no sense in seeing
"vm-debug-engine" in the error report.
(vm_error_not_a_pair, vm_error_not_a_bytevector)
(vm_error_not_a_struct, vm_error_not_a_thunk): Use func_name instead
of FUNC_NAME, so we can indicate what caused the error.
* libguile/vm-i-scheme.c (VM_VALIDATE_CONS, car, cdr, set-car!)
(set-cdr!): Indicate provenance of errors.
(VM_VALIDATE_STRUCT, struct-vtable):
(VM_VALIDATE_BYTEVECTOR, BV_FIXABLE_INT_REF, BV_INT_REF)
(BV_FLOAT_REF, BV_FIXABLE_INT_SET, BV_INT_SET, BV_FLOAT_SET): Same.
* libguile/vm-i-system.c (apply, tail-apply): Use
vm_error_apply_to_non_list.
* libguile/backtrace.h:
* libguile/backtrace.c (scm_display_error): Change "stack" arg to
"frame". Still accept stacks for backward compatibility.
(display_header, display_error_body): Show the source of the error, if
possible.
Guile implements this library in terms of SRFI-69, which is a bit vague on
the arity of hash functions, whereas `(rnrs hashtables)' explicitly
specifies unary ones.
* module/rnrs/hashtables.scm (wrap-hash-function): Assume SRFI-69 will pass
the table capacity as the second argument; return the result of proc
modulo the capacity.
* test-suite/tests/r6rs-hashtables.test (make-hashtable): New test case for
hash functions that return large values.
* module/ice-9/buffered-input.scm (make-buffered-input-port): Instead of
always returning EOF once EOF is seen once, call the producer again.
Allows Ctrl-D to cause one EOF at the REPL, returning once from a
recursive edit, but input continues normally. Thanks to Andrew
Bagdanov for a clue that led to the fix.
* module/Makefile.am (ice-9/psyntax-pp.scm.gen): No longer automatically
regenerate psyntax-pp.scm, as the gensym dance is getting too
complicated. Instead when you change psyntax.scm now you should make
ice-9/psyntax-pp.scm.gen.
* module/system/repl/repl.scm (run-repl): Run the thunk in a stack in a
prompt, similar to the default prompt. Gives proper backtraces.
* module/system/repl/error-handling.scm (call-with-error-handling):
Narrow one more outer frame, for the %start-stack thunk invocation.
* module/ice-9/boot-9.scm (%start-stack): Reindent.
* doc/ref/compiler.texi: Update for new ,pp meta-command.
* doc/ref/scheme-using.texi (Using Guile Interactively): Show value
history in examples.
(Value Historyx): Update docs to mention the repl option and the
programmatic interface.
* module/system/repl/common.scm: Use (ice-9 history). Turns on value
history by default.
(repl-default-options): Expand the format of options to include an
optional value transformer, run when setting a value. Add prompt and
value-history options.
(repl-prepare-eval-thunk): Use repl-option-ref.
(repl-option-ref): Error if the option is unknown.
(repl-option-set!, repl-default-option-set!): Error if the option is
unknown. Pass the val through the transformer procedure.
(repl-default-prompt-set!): Just use repl-default-option-set!.
* module/system/repl/command.scm (option): Update for the new options
format.
* module/system/repl/common.scm (repl-default-options): Heap-allocate
the repl-default-options, to avoid mutating a literal.
(make-repl): Copy repl-default-options.
* libguile/init.c (scm_i_init_guile): Move the call to
scm_init_threads_default_dynamic_state after the call to
scm_load_startup_files, so that the default dynamic state is in the
(guile-user) module, not (guile).
* module/system/repl/repl.scm (abort-on-error): New helper.
(run-repl): Don't enter the debugger during parsing or compilation of
a repl expression. If you want to debug compilation, run compilation
from the repl, not as part of the repl.
* module/system/repl/debug.scm (print-locals): Run the before-print-hook
on the values, so we can hook into (ice-9 history) if available. Don't
bother printing binding indices. Give a little per-line-prefix.
* module/system/repl/command.scm (read-command): Remove a pk.
* module/system/repl/repl.scm (run-repl): Export. Use % and abort to
implement the prompt.
* module/system/repl/debug.scm: New file, defines a data type to hold
state for a debugger stack, and some helper procedures to print the
stack or print a frame. Most pieces are from (system vm debug).
* module/system/repl/error-handling.scm: New file, implements
call-with-error-handling and with-error-handling, and instead of going
into a debugger, we go into a recursive repl that happens to have
debugging information. Will be removing the old debugger from (system
vm debug) shortly.
* module/Makefile.am (SYSTEM_SOURCES): Add error-handling and debug scm
files.
* module/system/repl/repl.scm (prompting-meta-read): Better error
handling -- we don't want to go into a debugger when reading a
command.
(start-repl): Add #:debug keyword argument, and just dispatch to
run-repl.
(run-repl): New function, with the guts of the old start-repl. Added a
prompt, to which a throw to 'quit will abort.
* module/system/repl/common.scm (repl-prepare-eval-thunk): New
helper. In the future we will use this to not enter the debugger on
errors that happen at compile time.
(repl-eval): Use repl-prepare-eval-thunk.
(repl-print): Run the before-print-hook when printing a value.
* module/system/repl/command.scm (*command-table*): Move `option' to the
`system' group. Move `trace' to the `profile' group. Add `debug' and
`inspect' groups.
(command-abbrevs): Rename from command-abbrev, and allow multiple
abbreviations.
(display-group): Fix the case where abbrev? was #f.
(display-summary): Fix alignment of the command and
abbreviations. Allow multiple abbreviations.
(read-command): Rename from read-datum, and have better error
handling.
(meta-command): Better error handling.
(define-meta-command): Better error handling.
(help, show, import, compile, disassemble, time, profile, trace): Fix
docstrings and error messages.
(define-stack-command): New helper, for commands that operate on a
saved stack.
(backtrace, up, down, frame, procedure, locals): New debugger
commands, in the REPL now.
(inspect, pretty-print): New "inspect" commands.
* acinclude.m4 (GUILE_READLINE): Check for rl_catch_signals and
rl_catch_sigwinch.
* guile-readline/readline.c (scm_init_readline): If we can, turn off
readline's signal handling, because we can do our own.
(scm_readline): Use dynwinds to handle resetting readline's state on
nonlocal exit, not catches.
(unwind_readline): Rename from handle_error.
* module/ice-9/psyntax.scm (lookup): Reflow comment.
(chi-top, syntax): Add comments about mod for lookup.
(set!): Lookup the identifier in the module attached to its syntax
object. In the (set! MACRO foo) case, after expanding the macro, chi
the resulting expression with the empty wrap, as syntax-type
does. Seems to fix the case where the expansion references
lexically-bound variables.
* module/ice-9/psyntax-pp.scm: Regenerated.
* test-suite/tests/syncase.test: Add a bunch of tests.
* libguile/strings.c (STRINGBUF_CONTENTS): New macro.
(STRINGBUF_CHARS, STRINGBUF_WIDE_CHARS): Use it.
(scm_i_string_data): New function.
* libguile/strings.h (scm_i_string_data): New declaration.
The "open-output-pipe":"no duplicate" test has been hanging, on and
off, and not completely reliably, for a few years. It's now doing so
fairly reliably for me, and investigation shows that
- the child shell process is in a tight loop (99% CPU)
- the parent Guile process is stuck calling waitpid().
The problem is that the child hasn't got the SIGPIPE that the test
intends, and so is continuing to echo "closed" forever; and Guile is
waiting for it to terminate, forever.
I haven't fully debugged the SIGPIPE problem, but it sounds very like
what Chet Ramey describes here:
http://old.nabble.com/Re%3A-SIGPIPE-not-properly-reset-with-%27trap---PIPE%27-p20985595.html.
(And my version of bash is 3.2.39.)
So, a fix should be to use something other than shell to implement the
child; and it appears that this works.
* check-guile.in (TEST_SUITE_DIR): Export.
* test-suite/tests/popen-child.scm: New script file.
* test-suite/tests/popen.test ("open-output-pipe", "no duplicate"):
Use Guile for the child process, instead of shell.
* module/ice-9/boot-9.scm (*repl-stack*): Instead of repl-level, have a
stack.
(batch-mode?): Change to poke the stack.
* module/ice-9/deprecated.scm (set-batch-mode?!): Update deprecation
method.
* module/system/repl/common.scm (repl-prompt): Update to poke
*repl-stack* to get the level.
* module/system/repl/repl.scm (start-repl): Bind *repl-stack*
appropriately.
* module/system/repl/common.scm (<repl>): Add inport and outport fields
and accessors.
(make-repl): Add optional "debug" argument. Bind inport and outport to
the current inport and output ports at the time of repl creation.
(repl-read): Read from the repl inport.
(repl-print): Write to the repl outport.
* module/system/repl/command.scm (read-datum, read-line, meta-command):
Respect repl-inport, and bind the outport of meta-commands to the repl
outport.
