Patches for solaris and friends are reinstated (thanks jperkin@)
Also thanks to he@ for testing on powerpc and arm
Highlights for 4.10.0:
* A new best-fit allocator for the major heap which reducing both GC cost
and memory usage.
* Immutable strings are now enforced at configuration time
* User-defined indexing operators for multidimensional arrays
* Miscellaneous improvements to the manual
* A more precise exhaustiveness check for GADTs
* Many bug fixes
Highlights for 4.11.0:
* Statmemprof: a new statistical memory profiler
* A new instrumented runtime that logs runtime statistics in a standard
format
* A native backend for the RISC-V architecture
* Improved backtraces that refer to function names
* Suppport for recursive and yet unboxed types
* A quoted extension syntax for ppxs.
* Many quality of life improvements
* Many bug fixes.
4.11.1 is a bugfix release. Full details in the CHANGELOG included with
the distribution.
On NetBSD 8.x, there will be issues with shared library
support, due to missing support for ADDR16_LO, ADDR16_HI
and ADDR16_HA relocations, available in newer code.
There are however other issues, text relocations, and
apparent missing REL24 relocation support(?!?).
The testyield.ml test appears to not terminate, only two
of the 4(?) threads ever get any CPU time, so has to be killed
to complete the selftests. Test result:
2438 tests passed
36 tests skipped
12 tests failed
108 tests not started (parent test skipped or failed)
8 unexpected errors
2602 tests considered
Highlights of this release are:
* new optimisations, in particular for affine functions in matches;
* the graphics library was moved out of the main distribution;
* the vmthread library was removed;
* support for compiler plugins was removed;
* many bug fixes.
For more details see the Changes file in the distribution.
- Claimed shared library support. same as other ELF platforms, but
since it's not GNU ld, no -Wl,-E available.
- Change the PLIST vars based on what is actually built
- consider x86_64-*-solaris to be x86_64 and solaris. the arch=none
option doesn't build.
- Search for socket stuff in -lnsl -lsocket.
- for a POSIX prototype sigwait, we need to define _XOPEN_SOURCE to something
4.08.1 is a bugfix release, fixing compilation failures in presence of the
-pack option, and dynlinking failures.
Highlights in 4.08.0 are:
* Binding operators (let*, let+, and*, etc). They can be used to
streamline monadic code.
* open now applies to arbitrary module expression in structures and to
applicative paths in signatures.
* A new notion of (user-defined) "alerts" generalizes the deprecated
warning.
* New modules in the standard library: Fun, Bool, Int, Option, Result.
* A significant number of new functions in Float, including FMA support,
and a new Float.Array submodule.
* Source highlighting for errors and warnings in batch mode.
* Many error messages were improved.
* Improved AFL instrumentation for objects and lazy values.
A few patches to the configure script to recognise the combination of
NetBSD and aarch64, and a few changes to the Makefile to make sure we've
got the right combination of options for the PLIST. It now compiles on my
Pinebook.
This breaks quite a few dependencies, more commits to follow.
There are many bugfixes and improvements (see the Changes file), I will
summarise the changes that may cause breakage:
* all standard library modules are now packed into the Stdlib module;
* Unix.isatty now works under the native Windows ports
* strings and bytes are distinguishable in bytecode
* Module aliases are no longer removed in 'module type of' and
'with module'
* soundness bugfix with non-generalised type variables and local modules
* Ident.t is now abstract and immutable
* only allow directives with filename and at the begining of the line
Changes are many and can be found at https://ocaml.org/releases/4.06.html
One change that breaks a lot of packages is that the safe-string compiler
option is now enabled by default, which means that all modules need to use
immutable strings (or explicitly use the bytes type).
While MACHINE_ARCH can be earmv6hf on NetBSD, configure has to match
what config.guess outputs, which is armv6. For now, leave the old
earmv6/7 tokens, because this code inexplicably succeeded on earmv7hf.
With this, ocaml builds and builds a working unison, on earmv6hf
(RPI3).
Package changes: PLIST cleanup, and added some options for native
compilation.
Changes from ocaml 4.04.2 include (MPR and GPR changed to M and G to not
trigger our CVS hooks):
(Changes that can break existing programs are marked with a "*")
### Language features:
### Code generation and optimizations:
- M#7201, G#954: Correct wrong optimisation of "0 / <expr>"
and "0 mod <expr>" in the case when <expr> was a non-constant
evaluating to zero
(Mark Shinwell, review by Gabriel Scherer, Leo White and Xavier Leroy)
- M#7357, G#832: Improve compilation time for toplevel
include(struct ... end : sig ... end)
(Alain Frisch, report by Hongbo Zhang, review by Jacques Garrigue)
- M#7533, G#1173: Correctly perform side effects for certain
cases of "/" and "mod"
(Mark Shinwell, report by Jan Mitgaard)
- G#504: Instrumentation support for fuzzing with afl-fuzz.
(Stephen Dolan, review by Alain Frisch, Pierre Chambart, Mark
Shinwell, Gabriel Scherer and Damien Doligez)
- G#863, G#1068, G#1069: Optimise matches with constant
results to lookup tables.
(Stephen Dolan, review by Gabriel Scherer, Pierre Chambart,
Mark Shinwell, and bug report by Gabriel Scherer)
- G#1150: Fix typo in arm64 assembler directives
(KC Sivaramakrishnan)
### Runtime system:
- M#385, G#953: Add caml_startup_exn
(Mark Shinwell)
- M#7423, G#946: expose new exception-raising functions
`void caml_{failwith,invalid_argument}_value(value msg)`
in addition to
`void caml_{failwith,invalid_argument}(char const *msg)`.
The previous functions would not free their message argument, so
were inconvient for dynamically-allocated messages; the messages
passed to the new functions are handled by the garbage collector.
(Gabriel Scherer, review by Mark Shinwell, request by Immanuel Litzroth)
- M#7557, G#1213: More security for getenv
(Damien Doligez, reports by Seth Arnold and Eric Milliken, review by
Xavier Leroy, David Allsopp, Stephen Dolan, Hannes Mehnert)
- G#795: remove 256-character limitation on Sys.executable_name
(Xavier Leroy)
- G#891: Use -fno-builtin-memcmp when building runtime with gcc.
(Leo White)
### Type system:
- M#6608, G#901: unify record types when overriding all fields
(Tadeu Zagallo and Gabriel Scherer, report by Jeremy Yallop,
review by David Allsopp, Jacques Garrigue)
* M#7414, G#929: Soundness bug with non-generalized type variables and
functors.
(compatibility: some code using module-global mutable state will
fail at compile-time and is fixed by adding extra annotations;
see the Mantis and Github discussions.)
(Jacques Garrigue, report by Leo White)
### Compiler user-interface and warnings:
- M#7050, G#748 G#843 G#864: new `-args/-args0 <file>` parameters to
provide extra command-line arguments in a file -- see documentation.
User programs may implement similar options using the new `Expand`
constructor of the `Arg` module.
(Bernhard Schommer, review by J?r?mie Dimino, Gabriel Scherer
and Damien Doligez, discussion with Alain Frisch and Xavier Leroy,
feature request from the Coq team)
- M#7137, G#960: "-open" command line flag now accepts
a module path (not a module name)
(Arseniy Alekseyev and Leo White)
- M#7172, G#970: add extra (ocamlc -config) options
int_size, word_size, ext_exe
(Gabriel Scherer, request by Daniel B?nzli)
- M#7315, G#736: refine some error locations
(Gabriel Scherer and Alain Frisch, report by Matej Ko??k)
- M#7473, G#1025: perform proper globbing for command-line arguments on
Windows
(Jonathan Protzenko)
- M#7479: make sure "ocamlc -pack" is only given .cmo and .cmi files,
and that "ocamlopt -pack" is only given .cmx and .cmi files.
(Xavier Leroy)
- G#796: allow compiler plugins to declare their own arguments.
(Fabrice Le Fessant)
- G#829: better error when opening a module aliased to a functor
(Alain Frisch)
- G#911: ocamlc/ocamlopt do not pass warnings-related options to C
compiler when called to compile third-party C source files
(S?bastien Hinderer, review by Adrien Nader and David Allsopp)
- G#915: fix -dsource (pprintast.ml) bugs
(Runhang Li, review by Alain Frisch)
* G#933: ocamlopt -p now reports an error on platforms that do not
support profiling with gprof; dummy profiling libraries are no longer
installed on such platforms.
This can be tested with ocamlopt -config
(S?bastien Hinderer)
- G#1009: "ocamlc -c -linkall" and "ocamlopt -c -linkall" can now be used
to set the "always link" flag on individual compilation units. This
controls linking with finer granularity than "-a -linkall", which sets
the "always link" flag on all units of the given library.
(Xavier Leroy)
- G#1015: add option "-plugin PLUGIN" to ocamldep too. Use compilerlibs
to build ocamldep. Add option "-depend" to ocamlc/ocamlopt to behave
as ocamldep. Remove any use of ocamldep to build the distribution.
(Fabrice Le Fessant)
- G#1027: various improvements to -dtimings, mostly including time
spent in subprocesses like preprocessors
(Valentin Gatien-Baron, review by Gabriel Scherer)
- G#1098: the compiler now takes the boolean "OCAML_COLOR" environment
variable into account if "-color" is not provided. This allows users
to override the default behaviour without modifying invocations of ocaml
manually.
(Hannes Mehnert, Guillaume Bury,
review by Daniel B?nzli, Gabriel Scherer, Damien Doligez)
### Standard library:
- M#6975, G#902: Truncate function added to stdlib Buffer module
(Dhruv Makwana, review by Alain Frisch and Gabriel Scherer)
- M#7279, G#710: `Weak.get_copy` `Ephemeron.*_copy` doesn't copy
custom blocks anymore
(Fran?ois Bobot, Alain Frisch, bug reported by Martin R. Neuh?u?er,
review by Thomas Braibant and Damien Doligez)
* M#7500, G#1081: Remove Uchar.dump
(Daniel B?nzli)
- G#760: Add a functions List.compare_lengths and
List.compare_length_with to avoid full list length computations
(Fabrice Le Fessant, review by Leo White, Josh Berdine and Gabriel Scherer)
- G#778: Arg: added option Expand that allows to expand a string
argument to a string array of new arguments
(Bernhard Schommer, review by Gabriel Scherer and J?r?mie Dimino)
- G#849: Expose a Spacetime.enabled value
(Leo White)
- G#885: Option-returning variants of stdlib functions
(Alain Frisch, review by David Allsopp and Bart Jacobs)
- G#869: Add find_first, find_first_opt, find_last, find_last_opt to
maps and sets. Find the first or last binding or element
satisfying a monotonic predicate.
(Gabriel de Perthuis, with contributions from Alain Frisch, review by
Hezekiah M. Carty and Simon Cruanes, initial report by Gerd Stolpmann)
- G#875: Add missing functions to ArrayLabels, BytesLabels,
ListLabels, MoreLabels, StringLabels so they are compatible with
non-labeled counterparts. Also add missing @@ocaml.deprecated attributes
in StringLabels and BytesLabels.
(Roma Sokolov, review by Gabriel Scherer, Jacques Garrigue,
Gabriel Radanne, Alain Frisch)
- G#999: Arg, do not repeat the usage message thrice when reporting an error
(this was a regression in 4.03)
(Florian Angeletti, review by Gabriel Scherer)
- G#1042: Fix escaping of command-line arguments in
Unix.create_process{,_env} under Windows. Arguments with tabs should now
be received verbatim by the child process.
(Nicolas Ojeda Bar, Andreas Hauptmann review by Xavier Leroy)
### Debugging and profiling:
- M#7258: ocamldebug's "install_printer" command had problems with
module aliases
(Xavier Leroy)
- G#378: Add [Printexc.raise_with_backtrace] to raise an exception using
an explicit backtrace
(Fran?ois Bobot, review by Gabriel Scherer, Xavier Leroy, Damien Doligez,
Fr?d?ric Bour)
### Manual and documentation:
- M#6597, G#1030: add forward references to language extensions
that extend non-terminal symbols in the language reference section.
(Florian Angeletti, review by Gabriel Scherer)
- M#7497, G#1095: manual, enable numbering for table of contents
(Florian Angeletti, request by Daniel B?nzli)
- M#7539, G#1181: manual, update dead links in ocamldoc chapter
(Florian Angeletti)
- G#633: manpage and manual documentation for the `-opaque` option
(Konstantin Romanov, Gabriel Scherer, review by Mark Shinwell)
- G#751, G#925: add a HACKING.adoc file to contain various
tips and tricks for people hacking on the repository. See also
CONTRIBUTING.md for advice on sending contributions upstream.
(Gabriel Scherer and Gabriel Radanne, review by David Allsopp,
inspired by John Whitington)
- G#916: new tool lintapidiff, use it to update the manual with
@since annotations for API changes introduced between 4.00-4.05.
(Edwin T?r?k, review by Gabriel Scherer, discussion with Alain Frisch,
David Allsopp, S?bastien Hinderer, Damien Doligez and Xavier Leroy)
- G#939: activate the caml_example environment in the language
extensions section of the manual. Convert some existing code
examples to this format.
(Florian Angeletti)
- G#1082: clarify that the use of quoted string for preprocessed
foreign quotations still requires the use of an extension node
[%foo ...] to mark non-standard interpretation.
(Gabriel Scherer, request by Matthew Wahab in G#1066,
review by Florian Angeletti)
### Other libraries:
- M#7158: Event.sync, Mutex.create, Condition.create cause too many GCs.
The fix is to no longer consider mutexes and condition variables
as rare kernel resources.
(Xavier Leroy)
- M#7264: document the different behaviors of Unix.lockf under POSIX
and under Win32.
(Xavier Leroy, report by David Allsopp)
- M#7339, G#787: Support the '0 dimension' case for bigarrays
(see Bigarray documentation)
(Laurent Mazare,
review by Gabriel Scherer, Alain Frisch and Hezekiah M. Carty)
* M#7342, G#797: fix Unix.read on pipes with no data left on Windows
it previously raised an EPIPE error, it now returns 0 like other OSes
(Jonathan Protzenko, review by Andreas Hauptmann and Damien Doligez)
- G#650: in the Unix library, add `?cloexec:bool` optional arguments to
functions that create file descriptors (`dup`, `dup2`, `pipe`, `socket`,
`socketpair`, `accept`). Implement these optional arguments in the
most atomic manner provided by the operating system to set (or clear)
the close-on-exec flag at the same time the file descriptor is created,
reducing the risk of race conditions with `exec` or `create_process`
calls running in other threads, and improving security. Also: add a
`O_KEEPEXEC` flag for `openfile` by symmetry with `O_CLOEXEC`.
(Xavier Leroy, review by Mark Shinwell, David Allsopp and Alain Frisch,
request by Romain Beauxis)
- G#996: correctly update caml_top_of_stack in systhreads
(Fabrice Le Fessant)
- G#997, G#1077: Deprecate Bigarray.*.map_file and add Unix.map_file as a
first step towards moving Bigarray to the stdlib
(J?r?mie Dimino and Xavier Leroy)
### Toplevel:
- M#7060, G#1035: Print exceptions in installed custom printers
(Tadeu Zagallo, review by David Allsopp)
### Tools:
- M#5163: ocamlobjinfo, dump globals defined by bytecode executables
(St?phane Glondu)
- M#7333: ocamldoc, use the first sentence of text file as
a short description in overviews.
(Florian Angeletti)
- G#848: ocamldoc, escape link targets in HTML output
(Etienne Millon, review by Gabriel Scherer, Florian Angeletti and
Daniel B?nzli)
- G#986: ocamldoc, use relative paths in error message
to solve ocamlbuild+doc usability issue (ocaml/ocamlbuild#79)
(Gabriel Scherer, review by Florian Angeletti, discussion with Daniel B?nzli)
- G#1017: ocamldoc, add an option to detect code fragments that could be
transformed into a cross-reference to a known element.
(Florian Angeletti, review and suggestion by David Allsopp)
- clarify ocamldoc text parsing error messages
(Gabriel Scherer)
### Compiler distribution build system:
- M#7377: remove -std=gnu99 for newer gcc versions
(Damien Doligez, report by ygrek)
- M#7452, G#1228: tweak GCC options to try to avoid the
Skylake/Kaby lake bug
(Damien Doligez, review by David Allsopp, Xavier Leroy and Mark Shinwell)
- G#693: fail on unexpected errors or warnings within caml_example
environment.
(Florian Angeletti)
- G#803: new ocamllex-based tool to extract bytecode compiler
opcode information from C headers.
(Nicolas Ojeda Bar)
- G#827: install missing mli and cmti files, new make target
install-compiler-sources for installation of compiler-libs ml files
(Hendrik Tews)
- G#887: allow -with-frame-pointers if clang is used as compiler on Linux
(Bernhard Schommer)
- G#898: fix locale-dependence of primitive list order,
detected through reproducible-builds.org.
(Hannes Mehnert, review by Gabriel Scherer and Ximin Luo)
- G#907: Remove unused variable from the build system
(S?bastien Hinderer, review by whitequark, Gabriel Scherer, Adrien Nader)
- G#911: Clarify the use of C compiler related variables in the build system.
(S?bastien Hinderer, review by Adrien Nader, Alain Frisch, David Allsopp)
- G#919: use clang as preprocessor assembler if clang is used as compiler
(Bernhard Schommer)
- G#927: improve the detection of hashbang support in the configure script
(Arma?l Gu?neau)
- G#932: install ocaml{c,lex}->ocaml{c,lex}.byte symlink correctly
when the opt target is built but opt.opt target is not.
(whitequark, review by Gabriel Scherer)
- G#935: allow build in Android's termux
(ygrek, review by Gabriel Scherer)
- G#984: Fix compilation of compiler distribution when Spacetime
enabled
(Mark Shinwell)
- G#991: On Windows, fix installation when native compiler is not
built
(S?bastien Hinderer, review by David Allsopp)
- G#1033: merge Unix and Windows build systems in the root directory
(S?bastien Hinderer, review by Damien Doligez and Adrien Nader)
- G#1047: Make .depend files generated for C sources more portable
(S?bastien Hinderer, review by Xavier Leroy and David Allsopp)
- G#1076: Simplify ocamlyacc's build system
(S?bastien Hinderer, review by David Allsopp)
### Compiler distribution build system: Makefile factorization
The compiler distribution build system (the set of Makefiles used to
build the compiler distribution) traditionally had separate Makefiles
for Unix and Windows, which lead to some amount of duplication and
subtle differences and technical debt in general -- for people working
on the compiler distribution, but also cross-compilation or porting to
new systems. During the 4.05 development period, S?bastien Hinderer
worked on harmonizing the build rules and merging the two build
systems.
* Some changes were made to the config/Makefile file which
is exported as $(ocamlc -where)/Makefile.config, and on
which some advanced users might rely. The changes are
as follows:
- a BYTERUN variable was added that points to the installed ocamlrun
- the PARTIALLD variable was removed (PACKLD is more complete)
- the always-empty DLLCCCOMPOPTS was removed
- the SHARED variable was removed; its value is "shared" or "noshared",
which duplicates the existing and more convenient
SUPPORTS_SHARED_LIBRARIES variable whose value is "true" or "false".
Note that Makefile.config may change further in the future and relying
on it is a bit fragile. We plan to make `ocamlc -config` easier to use
for scripting purposes, and have a stable interface there. If you rely
on Makefile.config, you may want to get in touch with S?bastien Hinderer
or participate to M#7116 (Allow easy retrieval of Makefile.config's values)
or M#7172 (More information in ocamlc -config).
The complete list of changes is listed below.
- G#705: update Makefile.nt so that ocamlnat compiles
for non-Cygwin Windows ports.
(S?bastien Hinderer, review by Alain Frisch)
- G#729: Make sure ocamlnat is built with a $(EXE) extension, merge
rules between Unix and Windows Makefiles
(S?bastien Hinderer, review by Alain Frisch)
- G#762: Merge build systems in the yacc/ directory.
(S?bastien Hinderer, review by David Allsopp, Alain Frisch)
- G#764: Merge build systems in the debugger/ directory.
(S?bastien Hinderer, review by Alain Frisch)
- G#785: Merge build systems in otherlibs/systhreads/
(S?bastien Hinderer, review by Alain Frisch, David Allsopp,
testing and regression fix by J?r?mie Dimino)
- G#788: Merge build systems in subdirectories of otherlibs/.
(S?bastien Hinderer, review by Alain Frisch)
- G#808, G#906: Merge Unix and Windows build systems
in the ocamldoc/ directory
(S?bastien Hinderer, review by Alain Frisch)
- G#812: Merge build systems in the tools/ subdirectory
(S?bastien Hinderer, review by Alain Frisch)
- G#866: Merge build systems in the stdlib/ directory
(S?bastien Hinderer, review by David Allsopp and Adrien Nader)
- G#941: Merge Unix and Windows build systems in the asmrun/ directory
(S?bastien Hinderer, review by Mark Shinwell, Adrien Nader,
Xavier Leroy, David Allsopp, Damien Doligez)
- G#981: Merge build systems in the byterun/ directory
(S?bastien Hinderer, review by Adrien Nader)
- G#1033, G#1048: Merge build systems in the root directory
(S?bastien Hinderer, review by Adrien Nader and Damien Doligez,
testing and regression fix by Andreas Hauptmann)
### Internal/compiler-libs changes:
- G#673: distinguish initialization of block fields from mutation in lambda.
(Fr?d?ric Bour, review by Xavier Leroy, Stephen Dolan and Mark Shinwell)
- G#744, G#781: fix duplicate self-reference in imported cmi_crcs
list in .cmti files + avoid rebuilding cmi_info record when creating
.cmti files
(Alain Frisch, report by Daniel B?nzli, review by J?r?mie Dimino)
- G#881: change `Outcometree.out_variant` to be more general.
`Ovar_name of out_ident * out_type list` becomes `Ovar_type of out_type`.
(Valentin Gatien-Baron, review by Leo White)
- G#908: refactor PIC-handling in the s390x backend
(Gabriel Scherer, review by Xavier Leroy and Mark Shinwell)
### Bug fixes
- M#5115: protect all byterun/fail.c functions against
uninitialized caml_global_data (only changes the bytecode behavior)
(Gabriel Scherer, review by Xavier Leroy)
- M#6136, G#967: Fix Closure so that overapplication evaluation order
matches the bytecode compiler and Flambda.
(Mark Shinwell, report by Jeremy Yallop, review by Fr?d?ric Bour)
- M#6550, G#1094: Allow creation of empty .cmxa files on macOS
(Mark Shinwell)
- M#6594, G#955: Remove "Istore_symbol" specific operation on x86-64.
This is more robust and in particular avoids assembly failures on Win64.
(Mark Shinwell, review by Xavier Leroy, testing by David Allsopp and
Olivier Andrieu)
- M#6903: Unix.execvpe doesn't change environment on Cygwin
(Xavier Leroy, report by Adrien Nader)
- M#6987: Strange error message probably caused by
universal variable escape (with polymorphic variants)
(Jacques Garrigue, report by Mikhail Mandrykin and Leo White)
- M#7216, G#949: don't require double parens in Functor((val x))
(Jacques Garrigue, review by Valentin Gatien-Baron)
- M#7331: ocamldoc, avoid infinite loop in presence of self alias,
i.e. module rec M:sig end = M
(Florian Angeletti, review Gabriel Scherer)
- M#7346, G#966: Fix evaluation order problem whereby expressions could
be incorrectly re-ordered when compiling with Flambda. This also fixes one
example of evaluation order in the native code compiler not matching the
bytecode compiler (even when not using Flambda)
(Mark Shinwell, Leo White, code review by Pierre Chambart)
- M#7348: Private row variables can escape their scope
(Jacques Garrigue, report by Leo White)
- M#7407: Two not-quite-standard C idioms rejected by SUNWSPro compilers
(Xavier Leroy)
- M#7421: Soundness bug with GADTs and lazy
(Jacques Garrigue, report by Leo White)
- M#7424: Typechecker diverges on unboxed type declaration
(Jacques Garrigue, report by Stephen Dolan)
- M#7426, G#965: Fix fatal error during object compilation (also
introduces new [Pfield_computed] and [Psetfield_computed] primitives)
(Mark Shinwell, report by Ulrich Singer)
- M#7427, G#959: Don't delete let bodies in Cmmgen
(Mark Shinwell, report by Valentin Gatien-Baron)
- M#7432: Linking modules compiled with -labels and -nolabels is not safe
(Jacques Garrigue, report by Jeremy Yallop)
- M#7437: typing assert failure with nonrec priv
(Jacques Garrigue, report by Anil Madhavapeddy)
- M#7438: warning +34 exposes #row with private types
(Alain Frisch, report by Anil Madhavapeddy)
- M#7443, G#990: spurious unused open warning with local open in patterns
(Florian Angeletti, report by Gabriel Scherer)
- M#7456, G#1092: fix slow compilation on source files containing a lot
of similar debugging information location entries
(Mark Shinwell)
- M#7504: fix warning 8 with unconstrained records
(Florian Angeletti, report by John Whitington)
- M#7511, G#1133: Unboxed type with unboxed argument should not be accepted
(Damien Doligez, review by Jeremy Yallop and Leo White)
- G#805, G#815, G#833: check for integer overflow in String.concat
(Jeremy Yallop,
review by Damien Doligez, Alain Frisch, Daniel B?nzli, Fabrice Le Fessant)
- G#881: short-paths did not apply to some polymorphic variants
(Valentin Gatien-Baron, review by Leo White)
- G#886: Fix Ctype.moregeneral's handling of row_name
(Leo White, review by Jacques Garrigue)
- G#934: check for integer overflow in Bytes.extend
(Jeremy Yallop, review by Gabriel Scherer)
- G#956: Keep possibly-effectful expressions when optimizing multiplication
by zero.
(Jeremy Yallop, review by Nicol?s Ojeda B?r, Xavier Leroy and Mark Shinwell)
- G#977: Catch Out_of_range in ocamldebug's "list" command
(Yunxing Dai)
- G#983: Avoid removing effectful expressions in Closure, and
eliminate more non-effectful ones
(Alain Frisch, review by Mark Shinwell and Gabriel Scherer)
- G#987: alloc_sockaddr: don't assume a null terminator. It is not inserted
on macOS by system calls that fill in a struct sockaddr (e.g. getsockname).
(Anton Bachin)
- G#998: Do not delete unused closures in un_anf.ml.
(Leo White, review by Mark Shinwell and Pierre Chambart)
- G#1019: Fix fatal error in Flambda mode "[functions] does not map set of
closures ID"
(Pierre Chambart, code review by Mark Shinwell and Leo White)
- G#1075: Ensure that zero-sized float arrays have zero tags.
(Mark Shinwell, Leo White, review by Xavier Leroy)
* G#1088: Gc.minor_words now returns accurate numbers.
(compatibility: the .mli declaration of `Gc.minor_words`
and `Gc.get_minor_free` changed, which may break libraries
re-exporting these values.)
(Stephen Dolan, review by Pierre Chambart and Xavier Leroy)
Pkgsrc changes:
* Add required macros for accessing ucontext for NetBSD/powerpc
* Omit files from PLIST.opt which are in PLIST.natdynlink.
So far macppc doesn't do "natdynlink", tests are failing.
* Adapt Makefile to features enabled on NetBSD/powerpc.
* Bump PKGREVISION.
The self-test results are nearly the same as on NetBSD/i386, with one
test failing, difference is one unexpected failure related to native
dynlink which I thought I had not enabled, and which obviously needs
more attention since it tries to reference a Linux linker script.
...
Summary:
637 tests passed
13 tests skipped
1 tests failed
1 unexpected errors
652 tests considered
List of failed tests:
tests/tool-ocamldoc-man/Inline_records.mli
List of unexpected errors:
tests/lib-dynlink-native
...
OCaml 4.04.2 (23 Jun 2017):
---------------------------
### Security fix:
- Local privilege escalation issue with ocaml binaries.
(Damien Doligez, report by Eric Milliken, review by Xavier Leroy)
OCaml 4.04.1 (14 Apr 2017):
---------------------------
### Code generation and optimizations:
- Consider arrays of length zero as constants
when using Flambda.
(Pierre Chambart, review by Mark Shinwell and Leo White)
### Standard library:
- fix a bug in Set.map as introduced in 4.04.0
(Gabriel Scherer, report by Thomas Leonard)
### Tools:
- ocamldoc, avoid nested <pre> tags in module description.
(Florian Angeletti, report by user 'kosik')
- ocamldoc, wrong Latex output for variant types
with constructors without arguments.
(Florian Angeletti, report by Xavier Leroy)
### Build system:
- New flexlink target in Makefile.nt to bootstrap the
flexlink binary only, rather than the flexlink binary and the FlexDLL C
objects.
(David Allsopp)
### Bug fixes
- Str.regexp raises "Invalid_argument: index out of bounds"
(Damien Doligez, report by John Whitington)
- Fix ocamlmklib with bootstrapped FlexDLL. Bootstrapped
FlexDLL objects are now installed to a subdirectory flexdll of the Standard
Library which allows the compilers to pick them up explicitly and also
ocamlmklib to include them without unnecessarily adding the entire Standard
Library.
(David Allsopp)
- fix incorrect timestamps returned by Unix.stat on Windows
when either TZ is set or system date is in DST.
(David Allsopp, report and initial fix by Nicolás Ojeda Bär, review and
superior implementation suggestion by Xavier Leroy)
- s390x: Fix address of caml_raise_exn in native dynlink modules
(Richard Jones, review by Xavier Leroy)
- ensure 16 byte stack alignment inside caml_allocN on x86-64
for ocaml build with WITH_FRAME_POINTERS defined
(Christoph Cullmann)
- fix slow compilation on source files containing a lot
of similar debugging information location entries
(Mark Shinwell)
- a case of double free in the systhreads library (POSIX implementation)
(Xavier Leroy, report by Chet Murthy)
- catch uncaught exception when unknown files are passed
as argument (regression in 4.04.0)
(Bernhard Schommer, review by Florian Angeletti and Gabriel Scherer,
report by Stephen Dolan)
- Memory cannot be released after calling
Bigarray.Genarray.change_layout.
(Damien Doligez and Xavier Leroy, report by Liang Wang)
- Fix segfault in Unix.create_process on Windows caused by wrong header
configuration.
(David Allsopp)
- add dynlink options to ocamlbytecomp.cmxa to allow ocamlopt.opt
to load plugins. See http://github.com/OCamlPro/ocamlc-plugins for examples.
(Fabrice Le Fessant, review by David Allsopp)
- caml-types.el: Fix missing format argument, so that it can show kind
of call at point correctly.
(Chunhui He)
- Allow Windows CRLF line-endings in ocamlyacc on Unix and Cygwin.
(David Allsopp, review by Damien Doligez and Xavier Leroy)
- Fix segfault in Sys.runtime_parameters when exception backtraces
are enabled.
(Olivier Andrieu)
some patches that have been migrated upstream.
Changes (bug numbers removed so they don't trigger our system) include:
(Changes that can break existing programs are marked with a "*")
### Language features:
- Support GADT equations on non-local abstract types
(Jacques Garrigue)
- Local opening of modules in a pattern.
Syntax: "M.(p)", "M.[p]","M.[| p |]", "M.{p}"
(Florian Angeletti, Jacques Garrigue, review by Alain Frisch)
- local exception declarations "let exception ... in"
(Alain Frisch)
- Allow shortcut for extension on semicolons: ;%foo
(Jeremie Dimino)
- optimized representation for immutable records with a single
field, and concrete types with a single constructor with a single argument.
This is triggered with a [@@unboxed] attribute on the type definition.
Currently mutually recursive datatypes are not well supported, this
limitation should be lifted in the future (see M).
(Damien Doligez)
### Compiler user-interface and warnings:
* interpret all command-line options before compiling any
files, changes (improves) the semantics of repeated -o options or -o
combined with -c see the super-detailed commit message at
da56cf6dfd
(whitequark)
- clarify the wording of Warning 38
(Unused exception or extension constructor)
(Gabriel Scherer)
* add colors when reporting errors generated by ppx rewriters.
Remove the `Location.errorf_prefixed` function which is no longer relevant
(Simon Cruanes, Jérémie Dimino)
- clarify the wording of Warning 8
(Non-exhaustivity warning for pattern matching)
(Florian Angeletti, review and report by Gabriel Scherer)
* Improve support for OCAMLPARAM: (i) do not use objects
files with -a, -pack, -shared; (ii) use "before" objects in the toplevel
(but not "after" objects); (iii) use -I dirs in the toplevel,
(iv) fix bug where -I dirs were ignored when using threads
(Marc Lasson, review by Damien Doligez and Alain Frisch)
- New -plugin option for ocamlc and ocamlopt, to dynamically extend
the compilers at runtime.
(Fabrice Le Fessant)
- Detect unused module declarations
(Alain Frisch)
- Add a settable Env.Persistent_signature.load function so
that cmi files can be loaded from other sources. This can be used to
create self-contained toplevels.
(Jérémie Dimino)
### Standard library:
- Provide `Sys.backend_type` so that user can write backend-specific
code in some cases (for example, code generator).
(Hongbo Zhang)
- implement Set.map
(Gabriel Scherer)
- Add Obj.reachable_words to compute the
"transitive" heap size of a value
(Alain Frisch, review by Mark Shinwell and Damien Doligez)
- Add a non-allocating function to recover the number of
allocated minor words.
(Pierre Chambart, review by Damien Doligez and Gabriel Scherer)
- String.split_on_char
(Alain Frisch)
- Filename.extension and Filename.remove_extension
(Alain Frisch, request by Edgar Aroutiounian, review by Daniel Bunzli
and Damien Doligez)
### Code generation and optimizations:
- Optimize Hashtbl by using in-place updates of its
internal bucket lists. All operations run in constant stack size
and are usually faster, except Hashtbl.copy which can be much
slower
(Alain Frisch)
* Optimize performance of record update:
no more performance cliff when { foo with t1 = ..; t2 = ...; ... }
hits 6 updated fields
(Olivier Nicole, review by Thomas Braibant and Pierre Chambart)
- Better unboxing strategy
(Alain Frisch, Pierre Chambart)
- Ocamlopt + flambda requires a lot of memory
to compile large array literal expressions
(Pierre Chambart, review by Mark Shinwell)
- Handle specialisation of recursive function that does
not always preserve the arguments
(Pierre Chambart, Mark Shinwell, report by Simon Cruanes)
- Obj.is_block is now an inlined OCaml function instead of a
C external. This should be faster.
(Demi Obenour)
- Optimize immutable float records
(Pierre Chambart, review by Mark Shinwell)
- Do not generate dummy code to force module linking
(Pierre Chambart, reviewed by Jacques Garrigue)
- Do not eliminate boxed int divisions by zero and
avoid checking twice if divisor is zero with flambda.
(Pierre Chambart, report by Jeremy Yallop)
- Optimize some constant string operations when the "-safe-string"
configure time option is enabled.
(Pierre Chambart)
- Load cross module information during a meet
(Pierre Chambart, report by Leo White, review by Mark Shinwell)
- Share a few more equal switch branches
(Pierre Chambart, review by Gabriel Scherer)
- Small improvements to type-based optimizations for array
and lazy
(Alain Frisch, review by Pierre Chambart)
- Prevent warning 59 from triggering on Lazy of constants
(Pierre Chambart, review by Leo White)
- Sort emitted functions according to source location
(Pierre Chambart, review by Mark Shinwell)
- Lack of type normalization lead to missing simple compilation for "lazy x"
(Alain Frisch)
### Runtime system:
- Allows to register finalisation function that are
called only when a value will never be reachable anymore. The
drawbacks compared to the existing one is that the finalisation
function is not called with the value as argument. These finalisers
are registered with `GC.finalise_last`
(François Bobot reviewed by Damien Doligez and Leo White)
- Do not perform compaction if the real overhead is less than expected
(Thomas Braibant)
### Tools:
- toplevel #show, follow chains of module aliases
(Gabriel Scherer, report by Daniel Bünzli, review by Thomas Refis)
- have ocamldep interpret -open arguments in left-to-right order
(Gabriel Scherer, report by Anton Bachin)
- ocamldoc, missing line breaks in type_*.html files
(Florian Angeletti)
- ocamldoc, improved support for inline records
(Florian Angeletti)
- ensure "ocamllex -ml" works with -safe-string
(Hongbo Zhang)
- ocamldoc, add viewport metadata to generated html pages
(Florian Angeletti, request by Daniel Bünzli)
- Make the output of ocamldep more stable
(Alain Frisch)
- empty documentation comments
(Florian Angeletti)
- Add the -no-version option to the toplevel
(Sébastien Hinderer)
- Add a --strict option to ocamlyacc treat conflicts as errors
(this option is now used for the compiler's parser)
(Jeremy Yallop)
- make ocamldoc use -open arguments
(Florian Angeletti)
- ocamldoc, fix order of extensible variant constructors
(Florian Angeletti)
### Debugging and profiling:
- Spacetime, a new memory profiler (Mark Shinwell, Leo White)
### Runtime system:
- Add a new primitive caml_alloc_float_array to allocate an
array of floats
(Thomas Braibant)
### Manual and documentation:
- document the existence of OCAMLPARAM and
ocaml_compiler_internal_params
(Damien Doligez, reports by Wim Lewis and Gabriel Scherer)
- warn users against using WinZip to unpack the source archive
(Damien Doligez, report by Shayne Fletcher)
- clarification to the wording and documentation
of Warning 52 (fragile constant pattern)
(Gabriel Scherer, William, Adrien Nader, Jacques Garrigue)
- Restore 4.02.3 behaviour of Unix.fstat, if the
file descriptor doesn't wrap a regular file (win32unix only)
(Andreas Hauptmann, review by David Allsopp)
- flatten : Avoid confusion
(Damien Doligez, report by user 'tormen')
- Gc.finalise and lazy values
(Jeremy Yallop)
- Document that [Store_field] must not be used to populate
arrays of values declared using [CAMLlocalN] (Mark Shinwell)
### Build system:
- Compiler developers: Adding new primitives to the
standard runtime doesn't require anymore to run `make bootstrap`
(François Bobot)
- Fix compilation using old Microsoft C Compilers not
supporting secure CRT functions (SDK Visual Studio 2005 compiler and
earlier) and standard 64-bit integer literals (Visual Studio .NET
2002 and earlier)
(David Allsopp)
- More sharing between Unix and Windows makefiles
(whitequark, review by Alain Frisch)
* Installed `ocamlc`, `ocamlopt`, and `ocamllex` are
now the native-code versions of the tools, if those versions were
built.
(Demi Obenour)
- "./configure -safe-string" to get a system where
"-unsafe-string" is not allowed, thus giving stronger non-local
guarantees about immutability of strings
(Alain Frisch, review by Hezekiah M. Carty)
### Bug fixes:
* Missed Type-error leads to a segfault upon record access.
(Jacques Garrigue, extra report by Stephen Dolan)
Proper fix required a more restrictive approach to recursive types:
mutually recursive types are seen as abstract types (i.e. non-contractive)
when checking the well-foundedness of the recursion.
* Nominal types and scope escaping.
Revert to strict scope for non-generalizable type variables, cf. Mantis.
Note that this is actually stricter than the behavior before 4.03,
cf. , meaning that you may sometimes need to add type annotations
to explicitly instantiate non-generalizable type variables.
(Jacques Garrigue, following discussion with Jeremy Yallop,
Nicolas Ojeda Bar and Alain Frisch)
- Aliased arguments ignored for equality of module types
(Jacques Garrigue, report by Leo White)
- compiler forcing aliases it shouldn't while reporting type errors
(Jacques Garrigue, report and suggestion by sliquister)
- document that Unix.SOCK_SEQPACKET is not really usable.
- uncaught exception on invalid lexer directive
(Gabriel Scherer, report by KC Sivaramakrishnan using afl-fuzz)
- revert a 4.03 change of behavior on (Unix.sleep 0.),
it now calls (nano)sleep for 0 seconds as in (< 4.03) versions.
(Hannes Mehnert, review by Damien Doligez)
- GADT + subtyping compile time crash
(Jacques Garrigue, report by Nicolas Ojeda Bar)
- Segfault from conjunctive constraints in GADT
(Jacques Garrigue, report by Stephen Dolan)
- Support more than FD_SETSIZE sockets in Windows' emulation
of select
(David Scott, review by Alain Frisch)
* Prevent private inline records from being mutated
(Alain Frisch, report by Pierre Chambart)
- Bug in mcomp_fields leads to segfault
(Jacques Garrigue, report by Leo White)
- Relaxed value restriction broken with principal
(Jacques Garrigue, report by Leo White)
- -strict-sequence turns off Warning 21
(Jacques Garrigue, report by Valentin Gatien-Baron)
- remove access to OCaml heap inside blocking section in win32unix
(David Allsopp, report by Andreas Hauptmann)
- remove access to OCaml heap inside blocking in Unix.sleep on Windows
(David Allsopp)
- -principal causes loop in type checker when compiling
(Jacques Garrigue, report by Anil Madhavapeddy, analysis by Leo White)
- Missing exhaustivity check for extensible variant
(Jacques Garrigue, report by Elarnon *)
- Contractiveness check unsound with constraints
(Jacques Garrigue, report by Leo White)
- GADT constructors can be re-exposed with an incompatible type
(Jacques Garrigue, report by Alain Frisch)
- Unsoundness in GADT exhaustiveness with existential variables
(Jacques Garrigue, report by Stephen Dolan)
* Thread library: fixed [Thread.wait_signal] so that it
converts back the signal number returned by [sigwait] to an
OS-independent number
(Jérémie Dimino)
- (similar to ) ensure that register typing constraints are
respected at N-way join points in the control flow graph
(Mark Shinwell)
- Fix float_of_hex parser to correctly reject some invalid forms
(Bogdan Tătăroiu, review by Thomas Braibant and Alain Frisch)
- Fix maximum weak bucket size
(Nicolas Ojeda Bar, review by François Bobot)
- Allow more module aliases in strengthening (Leo White)
- Fix wrong code generation involving lazy values in Flambda
mode
(Mark Shinwell, review by Pierre Chambart and Alain Frisch)
- Fix infinite loop in flambda due to [@@specialise] annotations
- Building native runtime on Windows could fail when bootstrapping
FlexDLL if there was also a system-installed flexlink
(David Allsopp, report Michael Soegtrop)
- check for integer overflow in String.concat
(Jeremy Yallop,
review by Damien Doligez, Alain Frisch, Daniel Bünzli, Fabrice Le Fessant)
- check for integer overflow in Array.concat
(Jeremy Yallop)
- fix the Buffer.add_substring bounds check to handle overflow
(Jeremy Yallop)
- Fix [@@inline] with default parameters in flambda (Leo White)
- fix build on OpenIndiana
(Sergey Avseyev, review by Damien Doligez)
### Internal/compiler-libs changes:
- Improve, fix, and add test for parsing/pprintast.ml
(Runhang Li, David Sheets, Alain Frisch)
- make driver/pparse.ml functions type-safe
(Gabriel Scherer, Dmitrii Kosarev, review by Jérémie Dimino)
- Improve Texp_record constructor representation, and
propagate updated record type information
(Pierre Chambart, review by Alain Frisch)
- Graphics.close_graph crashes 64-bit Windows ports (re-implementation
of )
(David Allsopp)
- delay registration of docstring after the mapper is applied
(Hugo Heuzard, review by Leo White)
- don't attach (**/**) comments to any particular node
(Thomas Refis, review by Leo White)
distinfo from cvs: trusted to convey hashes
patches from same cvs: untrusted thus require hashes to verify
absurdness: Any intruder that can modify the patch in cvs can also modify
the distinfo file.
Only externally hosted patches require hashes. Hopefully some day
thousands of future commits to distinfo files everywhere can be avoided by
making the framework smarter. </rant>
unneeded patches.
The list of changes is a thousand lines long, so I'm not including it
here: details can be found in the Changes file in the top directory of the
package after extraction.