* module/system/base/target.scm (cpu-word-size): Consider MIPS64 to
default to n32 or o32.
* test-suite/tests/asm-to-bytecode.test ("cross-compilation")
["mips64el-unknown-linux-gnu"]: New test.
* module/system/base/compile.scm (call-with-output-file/atomic): Call
close-port before deleting the temporary file name, otherwise deletion
fails on MS-Windows (cannot delete a file that is still open).
* module/system/base/language.scm (<language>): Add 'for-humans?'
field, and export new 'language-for-humans?' predicate.
* doc/ref/compiler.texi (Compiler Tower): Document.
* module/language/assembly/spec.scm, module/language/bytecode/spec.scm,
module/language/glil/spec.scm, module/language/objcode/spec.scm,
module/language/tree-il/spec.scm, module/language/value/spec.scm:
Mark these languages as 'for-humans?'.
* module/ice-9/boot-9.scm (current-language): New parameter.
* module/system/base/language.scm (*current-language*): Pull fluid from
parameter.
(current-language): Now a re-exported parameter.
* doc/ref/compiler.texi: Update reference from *current-language* fluid
to current-language parameter.
* module/system/base/compile.scm (compile-and-load):
* module/ice-9/top-repl.scm (top-repl): Default to the current language,
not to Scheme.
* module/ice-9/eval-string.scm:
* module/system/base/language.scm:
* module/system/repl/command.scm:
* module/system/repl/repl.scm: Update to use current-language parameter
and parameterize.
* libguile/vm-i-scheme.c (VM_VALIDATE_STRUCT): Fix the error message if
the value was not a struct.
* module/system/base/compile.scm (find-language-joint): Default to
joining at the target language.
(default-language-joiner): Allow sequences of one compiled expression
to pass through. Otherwise error as before.
(read-and-parse): New helper; actually parses.
(read-and-compile): Use read-and-parse, and fall back to
default-language-joiner.
Thanks to Nala Ginrut for the report.
* module/system/base/compile.scm (read-and-compile): Close the input
port after we read all of its data. Perhaps this cleans up some NFS
ghosts that David Pirotte was seeing.
* module/system/base/ck.scm: New module.
* module/srfi/srfi-9.scm: Import (system base ck).
(getter-type, getter-index, getter-copier): Convert incoming argument
convention to CK form.
(define-tagged-inlinable): Convert return value convention for key
lookup to CK form.
* module/srfi/srfi-9/gnu.scm: Import (system base ck).
Rename '%set-fields-unknown-getter' to 'unknown-getter'.
(c-list, c-same-type-check): New macros.
(%set-fields): Using the CK abstract machine, arrange to check (at
macro expansion time) that all of the getters in head position
correspond to the same record type.
* test-suite/tests/srfi-9.test: Add test.
* module/system/base/compile.scm (ensure-directory): Rename to...
(ensure-directory): ... this. Update callers. When ERRNO is EEXIST,
assume DIR is a writable directory instead of calling `stat' and
`access?' again. Fixes UID/EUID mismatches for setuid binaries.
Reported by rixed@happyleptic.org at
<http://lists.gnu.org/archive/html/guile-user/2012-06/msg00023.html>.
* module/system/base/pmatch.scm (pmatch): Always wrap with 'let', even
if the expression appears atomic, because in the presence of
'identifier-syntax', we cannot know what an atomic expression will
later expand to. Also use '#:export-syntax' instead of '#:export'
to export 'pmatch'.
* module/ice-9/boot-9.scm (cond, case): Reimplement using syntax-case,
with improved error messages and support for '=>' within 'case' as
mandated by the R7RS. Add warnings for duplicate case datums and
case datums that cannot be meaningfully compared using 'eqv?'.
* module/system/base/message.scm (%warning-types): Add 'bad-case-datum'
and 'duplicate-case-datum' warning types.
* test-suite/tests/syntax.test (cond, case): Update tests to reflect
improved error reporting. Add tests for '=>' within 'case'.
* test-suite/tests/tree-il.test (partial evaluation): Update tests to
reflect changes in how 'case' is expanded.
* doc/ref/api-control.texi (Conditionals): Document '=>' within 'case'.
* module/language/tree-il/analyze.scm
(format-analysis)[check-simple-format-args]: New procedure. Use it.
Add support for applications of <module-ref>.
* module/system/base/message.scm (%warning-types): Handle the `format
simple-format' warning.
* module/language/scheme/spec.scm (scheme)[make-default-environment]:
Use `simple-format' as the default `format'.
* test-suite/tests/tree-il.test ("warnings")["format"]: Explicitly use
(@ (ice-9 format) format) where needed.
("simple-format"): New test prefix.
* module/system/base/target.scm (%target-endianness, %target-word-size):
New fluids.
(%native-word-size): New variable.
(with-target): Set these fluids.
(cpu-endianness, cpu-word-size, triplet-cpu, triplet-vendor,
triplet-os): New procedures.
(target-cpu, target-vendor, target-os): Use them.
(target-endianness, target-word-size): Refer to the corresponding
fluid.
* libguile/objcodes.c (target_endianness_var, target_word_size_var): New
global variables.
(NATIVE_ENDIANNESS): New macro.
(target_endianness, target_word_size, to_native_order): New functions.
(make_objcode_from_file): Use `scm_bytecode_to_native_objcode' instead
of `scm_bytecode_to_objcode'.
(bytecode_to_objcode): New function, based on `scm_bytecode_to_objcode',
with the addition of an `endianness' and `word_size' parameters.
(scm_bytecode_to_objcode): Use it.
(scm_bytecode_to_native_objcode): New function.
(scm_write_objcode): Use `target_word_size' and `target_endianness'.
Convert OBJCODE's len and meta-len to native byte order.
(scm_init_objcodes): Initialize `target_endianness_var' and
`target_word_size_var'.
* libguile/objcodes.h (scm_bytecode_to_native_objcode): New declaration.
* libguile/vm.c (really_make_boot_program): Use
`scm_bytecode_to_native_objcode' instead of `scm_bytecode_to_objcode'.
* test-suite/tests/asm-to-bytecode.test (%objcode-cookie-size): New
variable.
(test-target): New procedure.
("cross-compilation"): Add `test-target' calls and the "unknown
target" test.
* module/system/base/target.scm (validate-target): Accept any tuple with
at least 3 parts.
* test-suite/tests/asm-to-bytecode.test (test-triplet): New procedure.
("cross-compilation"): New test prefix.
* libguile/load.c (canonical_to_suffix, scm_primitive_load_path):
* module/ice-9/boot-9.scm (load-in-vicinity):
* module/system/base/compile.scm (compiled-file-name): If the canonical
path of a file is a DOS-style path with a drive letter, turn it into a
path suffix it by removing the colon and prefixing a "/".
Inspired by a patch from Jan Nieuwenhuizen.
* module/system/base/language.scm (current-language): Default to 'scheme
here, rather than relying on *current-language* always being bound.
Thanks to 亚光唐 <texnician@gmail.com> for the report.
* 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.
This is so that compiling the same code on environments with different
locale settings yields the same result.
* module/system/base/compile.scm (compile-file): When ENC if #f, default
to "UTF-8" instead of `(fluid-ref %default-port-encoding)'.
* doc/ref/api-evaluation.texi (Compilation): Document the default output
file name and default source file encoding for `compile-file' and
`guile-tools compile'.
* module/system/base/lalr.scm (source-location->source-properties): New
public function, to produce source properties that can be given to the
compiler.
* module/language/tree-il/analyze.scm (&syntax-error): New variable.
(format-string-argument-count): Throw to &SYNTAX-ERROR when a syntax
error in a format string is encountered.
(format-analysis): Catch &SYNTAX-ERROR and convert as a warning of the
appropriate type.
* module/system/base/message.scm (%warning-types)[format]: Handle
`syntax-error' warnings.
* test-suite/tests/tree-il.test
("warnings")["conditionals"]("unterminated", "unexpected ~;",
"unexpected ~]"): New tests.
["unterminated ~{...~}"]: New test.
* module/language/tree-il/analyze.scm (format-analysis): Add new
sub-warnings: `wrong-port', `wrong-format-string',
`non-literal-format-string', and `wrong-num-args'.
* module/system/base/message.scm (%warning-types)[format]: Handle
them.
* test-suite/tests/tree-il.test ("warnings")["wrong port arg",
"wrong format string", "non-literal format string",
"wrong number of args"]: New tests.
* module/language/tree-il/analyze.scm (format-string-argument-count):
Return two values, the minimum and maximum number of arguments.
Add support for most of `format' escapes, including conditionals.
(format-analysis): Adjust accordingly.
* module/system/base/message.scm (%warning-types)[format]: Take two
arguments, MIN and MAX, instead of EXPECTED. Display warning
accordingly.
* test-suite/tests/tree-il.test ("warnings")["format"]("~%, ~~, ~&, ~t,
~_, and ~\\n", "~{...~}", "~{...~}, too many args", "~@{...~}",
"~@{...~}, too few args", "~(...~)", "~v", "~v:@y", "~*", "~?",
"complex 1", "complex 2", "complex 3"): New tests.
("conditionals"): New test prefix.
* am/guilec (install-data-hook): Remove.
(guile_install_go_files): New variable.
($(guile_install_go_files)): New dependency.
* libguile/load.c (compiled_is_fresh): Check for ordering of STAT_SOURCE
and STAT_COMPILED, not equality.
* module/ice-9/boot-9.scm (load): Ditto.
* module/system/base/compile.scm (call-with-output-file/atomic): Don't
set the timestamp of TEMPLATE.
* module/system/base/compile.scm:
* module/system/base/language.scm (*current-language*, current-language):
Move this fluid and thunk down to (system base language).
* module/system/base/language.scm (<language>): Remove the `version'
field from languages. It just wasn't useful.
* module/language/assembly/spec.scm:
* module/language/brainfuck/spec.scm:
* module/language/bytecode/spec.scm:
* module/language/ecmascript/spec.scm:
* module/language/elisp/spec.scm:
* module/language/glil/spec.scm:
* module/language/objcode/spec.scm:
* module/language/scheme/spec.scm:
* module/language/tree-il/spec.scm:
* module/language/value/spec.scm: Remove #:version from all language
definitions. Shorten some language names (e.g. "Guile Scheme" ->
"Scheme").
* module/system/base/compile.scm (compile-file, compile-and-load): Add a
keyword arg #:canonicalization, which defaults to 'relative. In this
way, one might compile "../module/ice-9/boot-9.scm", but the path that
gets residualized into the .go is "ice-9/boot-9.scm".
* module/system/base/compile.scm (compiled-file-name): Add a comment.
* module/ice-9/boot-9.scm (load): Avoid loading up (system base compile)
just to compute an autocompiled file name. Fixes the issue whereby
guile-tools snarf-check-and-output-texi was inadvertantly loading up
srfi-1, and thereby a stale library, just to see if guile-tools itself
had a compiled version.
Not sure what the right unit test is here, other than vigilance.
* libguile/posix.h:
* libguile/posix.c (scm_utime): Add optional nanosecond arguments. This
is an incompatible change on the C level, but it's unlikely people are
using this POSIX wrapper function, because they would just use the
POSIX function directly. Hopefully, anyway.
* module/system/base/compile.scm (call-with-output-file/atomic):
Propagate source timestamps to targets with nanosecond precision, if
available. Fixes build on systems with ext4 filesystems.
* libguile/filesys.c (scm_stat2scm):
* module/ice-9/posix.scm (stat:atimensec, stat:mtimensec)
(stat:ctimensec): Add three new elements to Scheme stat structures,
for nanosecond-level timestamps.
* configure.ac: Add checks for utimensat, and for nanosecond fields in
struct stat. We should switch to using Gnulib things for these,
though.
* doc/ref/posix.texi (File System): Add documentation for utime's
additional arguments, and nanosecond stat timestamp accessors.
* module/language/tree-il/analyze.scm (<reference-dag>): New record
type.
(dag-reachable-nodes, dag-reachable-nodes*, unused-variable-analysis):
New variables.
(unbound-variable-analysis): Slightly simplify the `up' procedure.
* module/language/tree-il/compile-glil.scm (%warning-passes): Add
`unused-toplevel'.
* module/system/base/message.scm (%warning-types): Likewise.
* test-suite/tests/tree-il.test (%opts-w-unused-toplevel): New variable.
("warnings")["unused-toplevel"]: New test prefix.
* module/language/tree-il/analyze.scm (<arity-info>): New record type.
(validate-arity, arity-analysis): New variables.
* module/language/tree-il/compile-glil.scm (%warning-passes): Add
`arity-mismatch'.
* module/system/base/message.scm (%warning-types): Likewise.
* test-suite/tests/tree-il.test (read-and-compile): Remove, as it's now
public.
(%opts-w-arity): New.
("warnings")["arity mismatch"]: New test prefix.
* test-suite/Makefile.am:
* test-suite/tests/brainfuck.test: Add a brainfuck test.
* module/system/base/compile.scm: Also export read-and-compile.
* module/language/tree-il/spec.scm (join): Fix the joiner in the
0-expression case.
* module/language/tree-il/primitives.scm (+): Recognize (+ x -1) as 1-.
* module/language/brainfuck/parse.scm (read-brainfuck): Return EOF if we
actually received EOF, and there were no expressions read.
* module/language/brainfuck/compile-tree-il.scm (compile-body): Fix the
compiler for the new format of "lambda" in tree-il.
* module/ice-9/boot-9.scm (make-fresh-user-module): New public function,
makes an anonymous beautified module.
* module/language/objcode/spec.scm: We used to have some things in here
that allowed lexical variable names and values to be a part of the
environment, but no more. Now an environment is just a module. If you
want to "inject" free variables into code, just use lambda.
* module/language/scheme/compile-tree-il.scm (compile-tree-il): Same
here. Also, rely on the fact that an environment *will* be a module --
because (system base compile) guarantees that for us.
* module/language/scheme/spec.scm (scheme): In the reader, rely on the
environment being a module. Define a #:make-default-environment
handler, which returns a beautified module, augmented with a fresh
definition for current-reader, so that side effects to current-reader
are restricted to the compilation unit.
* module/language/tree-il/analyze.scm
(report-possibly-unbound-variables):
* module/language/tree-il/compile-glil.scm (compile-glil):
* module/language/tree-il/optimize.scm (optimize!): The environment will
be a module.
* module/system/base/language.scm (<language>): New field,
`make-default-environment'. Defaults to `make-fresh-user-module'.
(default-environment): New accessor, returns a default environment for
a language.
* module/system/repl/common.scm (repl-compile): Always compile relative
to the current module, because a module is always acceptable as an
environment.
* module/system/base/compile.scm (compile-file, compile-and-load): Both
of these have a new keyword argument, #:env. For `compile-file', it
defaults to the default environment of the source language, and for
`compile-and-load', to the current module.
(read-and-compile): If there are no expressions read, pass the joiner
its default environment (via `default-environment joint').
* module/system/base/compile.scm (current-compilation-environment):
Remove, as the only thing that needed it (language readers) now get
the environment as an argument.
(read-and-compile, compile): Rework for no *compilation-environment*,
and default the environment using the define* mechanism.
* module/language/tree-il/analyze.scm (env-module): Hack around the lack
of a current compilation module. Will fix this in the next commit so
that the environment is always valid.
