Process.Process: Support using Stdio.Fd for "cwd" and "chroot" on OSes with fchdir(2). Fixes #10118.
Merge branch 'patches/checker' into 8.0 * patches/checker: Configure: Make --with-checker work again.
Merge branch 'patches/checker' * patches/checker: Configure: Make --with-checker work again.
Configure: Make --with-checker work again.
Configure: Default to --with-poll except for on specific OSes. Default to --with-poll except for on BSDs, Darwin, GNU, Next and SunOS. Previously the default was --without-poll except for on AIX, HP-UX, Linux, OpenUNIX, OSF/1, Solaris, and UnixWare.
Configure [Darwin]: Drop deprecated linker flag
Build [clang]: Fix warning about unknown "noclone" attribute
Build: Use offsetof() and alignof() if they exist.
Configure: Use $EGREP in preference to plain egrep. Fixes warnings from some versions of egrep that recommend using grep -E.
Configure [Haiku/BeOS]: Fix compiler arguments for dynamic linking and loading.
Configure [Haiku/BeOS]: socket() hides in -lnetwork.
Configure [Haiku/BeOS]: Avoid interactive window during configure. Haiku throws up an interactive window on core-dump...
Build: Add version tag to pike binary on AmigaOS
Configure [Solaris]: Fix check for dp_setp.
Build: Explicit #include of <sys/errno.h> is not needed anywhere. Some libc's (eg musl) complain.
Configure [Solaris]: Add missing checking message.
Configure [Solaris]: Check if struct dvpoll has the field dp_setp. In recent versions of Solaris 11 (eg OpenIndiana) the struct dvpoll has grown with an extra field. Make sure to initialize it properly.
Build: Prefer isnan() to _isnan() if both exist.
Configure [NT]: Inhibit one more warning.
Configure [NT]: Inhibit some common warnings.
Configure: Reverse the stack direction test logic. On compilation failure, the test now defaults to the common default stack direction of down.
Configure: Abstract access to cross compilation indicator The name of the shell variable changed between autoconf versions, so make sure that we can handle both names without having to duplicate the logic for it everywhere.
configure: Set appropriate default master path on AmigaOS
Compiler: Add optional validation that %rsp is properly aligned. Add new generic configure option `--with-experimental` intended to be used to enable various experimental and/or debug code. Exactly what `--with-experimental` enables is NOT intended to be stable, and may be changed at any time. It may also have no effect at all.
Add --with-mc-stack-frames configure option. (Currently X86-64 only.) This will enable frame pointers in machine code, thereby allowing e.g. Linux perf to unwind the stack and get proper stack traces including Pike functions.
Configure: Check for <mach/port.h> and <mach/message.h>. This will hopefully make Hurd happier.
Runtime: Removed the old LFUN lookup code.
configure: Move gmp check to after searching for library paths This prevents the flags provided by pkg-config from upsetting the very delicate ordering interdependency between library path and include path.
Builtin: Use 64-bit version of time_t in current glibc This functionality exists in glibc releases starting with 2.34.
Add more header files to avoid errors due to implicitly declared functions.
Configure: More visibility fixes.
Configure: Do not use -fvisibility=hidden if it causes warnings.
Revert "Configure: Check if __attribute__((visibility("default"))) is supported." This test did not work as intended. Taking a different approach. This reverts commit 5028ef9756bacac464e3263f21450de406425a36.
Configure: Check if __attribute__((visibility("default"))) is supported.
Portability [Solaris 8]: Add implementation of log2(3M).
EFUNs: Use mallinfo2(3C) if available. Fixes warnings on versions of Linux where mallinfo(3C) has been deprecated.
Merge branch 'patches/memory-usage-mallinfo2' * patches/memory-usage-mallinfo2: EFUNs: Use mallinfo2(3C) if available.
Merge branch 'tobij/fix-find-stack-direction-inlining' * tobij/fix-find-stack-direction-inlining: Build [Configure]: Prevent inling of find_stack_direction() Build [Configure]: Remove $Id$
Merge branch 'tobij/fix-find-stack-direction-inlining' into 8.0 * tobij/fix-find-stack-direction-inlining: Build [Configure]: Prevent inling of find_stack_direction() Build [Configure]: Remove $Id$
Build [Configure]: Prevent inling of find_stack_direction() Clang 11 has either learned to inline find_stack_direction(), or it may have inlined find_stack_direction() before but changed allocation order for our "objects with automatic storage duration" that we use to check the stack direction. Calling find_stack_direction() through a volatile function pointer should reliably prevent it from getting inlined and thus flush out the real stack direction. Thanks to Cezary Cerekwicki <ccerekwicki@opera.com> for the report.
Build [Configure]: Remove $Id$
block_allocator: Remove unused alignment functionality
configure: add a more targeted path inclusion
configure: add --with-exclude-site-prefixes option which excludes a path from being used for bin/lib/incudes
configure: add some additional software locations (pkgsrc)
I/O [macOS]: Include <util.h> for openpty
Build [Configure]: Blacklist chroot on macOS The prototype is not available with _POSIX_C_SOURCE.
Backport configure fixes from 8.1
Build: Look for libs in /opt/local. Useful with MacPorts libs.
Build [Configure]: Include prototypes for OOB test
Build [Configure]: Add prototype to dlopen test
Build [Configure]: Fix likely typo in const check If anyone with "IBM C" suffers a regression because of this, please let me know. :-)
Build [Configure]: Add /opt/homebrew to locations to search This is the new default location for macOS homebrew.
Fix machine code on macOS 11 macOS 11 adds two JIT hardening features: * Memory can only be mmapped as both PROT_WRITE and PROT_EXEC if MAP_JIT is also specified. * Memory mapped with MAP_JIT is never writeable and executable at the same time. It is necessary to call a function before writing, and then again after writing to make it executable again.
Build [Configure]: Blacklist getpagesize on macOS getpagesize() was removed from POSIX.1-2001. macOS therefore removes the prototype from <unistd.h> if _POSIX_C_SOURCE is 200112 or higher. However, the function still exists for backwards compat, which leads to the configure check finding it (since it fakes a prototype). Prevent the configure check from running and finding the prototype-less function so that we don't get a compilation error trying to call it. (The correct way of getting the page size in POSIX is sysconf(_SC_PAGESIZE).)
Build [Configure]: mkdir test needs <sys/stat.h> on MacOS 11
Build [Configure]: Include <stdlib.h> in tests using exit()
Configure: Remove /opt/local from default search path. /opt/local on MacOS X can contain libraries that conflict with and are incompatible with the system libraries (eg libiconv.dylib). This has a tendency to break stuff. If you need /opt/local, please use --with-site-prefixes. Fixes linking with MariaDB and MySQL.
build: add common pkgsrc locations
Build [Configure]: Use the new 2-level lfun lookup table by default.
Merge branch 'grubba/pty-handling' into 8.0 * grubba/pty-handling: Build [Linux]: openpty(3BSD) may be hidden in libutil.
Merge branch 'grubba/pty-handling' * grubba/pty-handling: Build [Linux]: openpty(3BSD) may be hidden in libutil.
Build [Linux]: openpty(3BSD) may be hidden in libutil.
Merge branch 'grubba/pty-handling' into 8.0 * grubba/pty-handling: (23 commits) I/O [NT]: Fixed typo in FD_DEBUG code. I/O [NT]: Updated note. I/O [NT]: Added some notes about pty handling. I/O [NT]: More pty handling fixes. I/O [NT]: Keep track of processes using ConPTY slaves. I/O [NT]: Do not attempt to read too much data from ptys. I/O [NT]: Changed calling convention for close_pty(). I/O [NT]: debug_fd_openpty() now returns ENOTSUPP when no ConPTY. I/O [NT]: Fixed some issues with debug_fd_openpty(). Process [NT]: Fix process creation. I/O [NT]: More debug_fd_openpty() fixes: I/O [NT]: Fixed multiple isses with debug_fd_openpty(). I/O [NT]: Adjusted preprocessor conditionals. I/O [NT]: Fixed typo. I/O [NT]: Fixed some issues with debug_fd_openpty(). Process [NT]: Fixed some typos in recent commits. Process.create_process [NT]: First go at attempting to support ptys. I/O [NT]: Make Pike_NT_* available to the rest of Pike. Process[NT]: Switch to using STARTUPINFOEXW. I/O [NT]: Added debug_fd_openpty(). ...
Merge branch 'grubba/pty-handling' * grubba/pty-handling: I/O [NT]: Fixed some issues with debug_fd_openpty(). Process [NT]: Fixed some typos in recent commits.
I/O [NT]: Fixed some issues with debug_fd_openpty(). Setting the window size should now work.
Merge branch 'grubba/pty-handling' * grubba/pty-handling: Process.create_process [NT]: First go at attempting to support ptys. I/O [NT]: Make Pike_NT_* available to the rest of Pike. Process[NT]: Switch to using STARTUPINFOEXW. I/O [NT]: Added debug_fd_openpty(). I/O [NT]: Reading from the console on NT may fail with ENOMEM. I/O [NT]: Restructured debug_fd_read() somewhat. I/O: Added fd_*-wrapper for openpty().
I/O: Added fd_*-wrapper for openpty(). Preparation for supporting ptys on NT.
Merge branch 'grubba/pty-handling' into 8.0 * grubba/pty-handling: Process [NT]: Fixed typo in previous commit. Process [NT]: Support UTF-8 in create_process(). I/O [NT]: Added low_dwim_utf8_to_utf16(). I/O [NT]: Fixed some typos. I/O [NT]: Added dynamic lookup of symbols needed for pty handling. I/O [NT]: Added fallbacks for some stuff needed for pty handling. Build [Stdio]: Make systems without termios.h (ie NT) happy again. Stdio: Make Coverity happy. [CID 1457890] Stdio: Fixed some inverted tests in my_openpty(). Stdio: Removed extraneous and broken #include of <pty.h>. Stdio: Added support for PROP_TTY to pipe(). Stdio: Some pty-handling cleanup.
I/O [NT]: Added fallbacks for some stuff needed for pty handling.
Merge branch 'grubba/fdlib' into grubba/pty-handling-new * grubba/fdlib: (55 commits) I/O [NT]: Use read-write locking for fd_busy and fd_to_{socket,handle()}. I/O [NT]: Fixed typo. I/O [NT]: Avoid blocking on read from pipes with data. Added some paranoia System [NT]: Normalize the drive letter explicitly in normalize_path(). System [NT]: More normalize_path() fixes. Configure [Solaris]: Fixed quoting level issue. System [NT]: Fixed trailing slashes from normalize_path(). I/O [Solaris]: Fix getcwd() on Solaris. I/O [NT]: Fixed a few warnings. I/O [NT]: Moved bulk of normalize_path() to fdlib. System [NT]: Update normalize_path() to use UTF8. I/O [NT]: Get rid of DO_NOT_WARN(). I/O [NT]: Added some more FD_DEBUG. I/O [NT]: Fixed typo in FD_DEBUG code. Build [NT]: Always use or own implementation of _dosmaperr(). Build [NT]: Improved support for MSVC 2010 (aka MSC 1600). I/O [NT]: Fixed deadlock. I/O [NT]: Updated FD_DEBUG diagnostics. I/O [NT]: Survive invalid UTF-16 surrogate pairs. ...
Merge branch 'grubba/pty-handling' * grubba/pty-handling: I/O [NT]: Added dynamic lookup of symbols needed for pty handling. I/O [NT]: Added fallbacks for some stuff needed for pty handling. Build [Stdio]: Make systems without termios.h (ie NT) happy again.
Merge branch 'grubba/fdlib' into grubba/pty-handling * grubba/fdlib: (55 commits) I/O [NT]: Use read-write locking for fd_busy and fd_to_{socket,handle()}. I/O [NT]: Fixed typo. I/O [NT]: Avoid blocking on read from pipes with data. Added some paranoia System [NT]: Normalize the drive letter explicitly in normalize_path(). System [NT]: More normalize_path() fixes. Configure [Solaris]: Fixed quoting level issue. System [NT]: Fixed trailing slashes from normalize_path(). I/O [Solaris]: Fix getcwd() on Solaris. I/O [NT]: Fixed a few warnings. I/O [NT]: Moved bulk of normalize_path() to fdlib. System [NT]: Update normalize_path() to use UTF8. I/O [NT]: Get rid of DO_NOT_WARN(). I/O [NT]: Added some more FD_DEBUG. I/O [NT]: Fixed typo in FD_DEBUG code. Build [NT]: Always use or own implementation of _dosmaperr(). Build [NT]: Improved support for MSVC 2010 (aka MSC 1600). I/O [NT]: Fixed deadlock. I/O [NT]: Updated FD_DEBUG diagnostics. I/O [NT]: Survive invalid UTF-16 surrogate pairs. ...
Build: Fix AmigaOS build
Build: Added configure option --with-new-lfun-lookup.
Configure [NetBSD]: Search for paxctl. Attempt to use paxctl to alter the PaX flags for the pike binary to allow for generation of machine code. Note that paxctl on NetBSD and Linux are incompatible.
configure: Fix typo in byte order test for cross compilation
Merge remote-tracking branch 'origin/master' into new_utf8
configure: Detect byteorder when cross compiling Also, fail if the byteorder can not be detected, since currently Pike needs to know the byteorder for compilation to be possible (due to siphash24).
Build: Unbreak --without-dynamic-modules When module.c was renamed to pike_modules.c, apparently this reference was not updated.
Builtin: Fixed mktime_zone() when localtime is in dst. Adds variant using timezone/altzone (eg Solaris). Fixes the fallback implementation (using gmtime(3F)) to not drift by an hour when the local time zone has tm_isdst == 1.
Merge commit '722771973bd' into patches/lyslyskom22891031 * commit '722771973bd': (6177 commits) Verify that callablep responses are aligned with reality. ...
Merge commit '2470270f500c728d10b8895314d8d8b07016e37b' into grubba/typechecker-automap * commit '2470270f500c728d10b8895314d8d8b07016e37b': (18681 commits) Removed the old typechecker. ...
EFUNs: Attempt to use thread-safe implementations of crypt(3C).
Merge branch 'patches/pike144' into 8.0 * patches/pike144: Core: Do not use the obsolete (~20 years) ioctl /proc API.
Merge branch 'patches/pike144' * patches/pike144: Core: Do not use the obsolete (~20 years) ioctl /proc API.
Core: Do not use the obsolete (~20 years) ioctl /proc API. The ioctl-based /proc API was obsoleted in Solaris 2.6 (July 1997), but to be supported "until the next major release". In Solaris 11.4 (August 2018) the support has now finally been removed. Most of the removed code was from 1995, and was thus older than Pike...
Merge remote-tracking branch 'origin/8.1' into gobject-introspection
RISCV: Minimal machine code backend
Configure: Added some quotes to make editors happy.
Builtin: Use 64-bit version of time_t in future glibc This functionality does not exist in mainline glibc yet, but the API design can be found on the glibc wiki page https://sourceware.org/glibc/wiki/Y2038ProofnessDesign and a draft implementation exists on a branch.
configure: fixed check for off64_t
Merge branch 'grubba/fdlib' into 8.0 * grubba/fdlib: Configure [Solaris]: Fixed quoting level issue. System [NT]: Fixed trailing slashes from normalize_path().
Configure [Solaris]: Fixed quoting level issue. Configure should now detect working == some correctly on Solaris.
Merge branch 'grubba/fdlib' into 8.0 * grubba/fdlib: I/O [Solaris]: Fix getcwd() on Solaris.
I/O [Solaris]: Fix getcwd() on Solaris. On Solaris getcwd(NULL, x) only works if x is > 0. Fixes building issues on Solaris.
Merge branch 'grubba/fdlib' into 8.0 * grubba/fdlib: I/O [NT]: Added some more FD_DEBUG. I/O [NT]: Fixed typo in FD_DEBUG code. Build [NT]: Always use or own implementation of _dosmaperr(). Build [NT]: Improved support for MSVC 2010 (aka MSC 1600).
Merge branch 'grubba/fdlib' into 8.1 * grubba/fdlib: I/O [NT]: Added some more FD_DEBUG. I/O [NT]: Fixed typo in FD_DEBUG code. Build [NT]: Always use or own implementation of _dosmaperr(). Build [NT]: Improved support for MSVC 2010 (aka MSC 1600).
Build [NT]: Always use or own implementation of _dosmaperr(). Also consolidates the two implementations.
Merge branch 'grubba/fdlib' into 8.0 * grubba/fdlib: (36 commits) I/O [NT]: Updated FD_DEBUG diagnostics. I/O [NT]: Moved FD_SETSIZE definition to global.h. I/O [NT]: Use new low-level functions from Process. I/O [NT]: Added some assertions. I/O [NT]: Added serialization of access to fd handles. I/O [NT]: Use the new low-level functions. I/O [NT]: Added some low-level fd handling functions. Build [NT]: Fix typo of EACCES. I/O [NT]: Document some internal data structures. Stdio: Cleanup File()->truncate(). Configure: Moved tests for truncate64() and ftruncate64(). Stdio [NT]: Fixed warning on NT. Stdio [NT]: Added detection constant __HAVE_UTF8_FS__. I/O [NT]: Added note about the intent of fdlib. Stdio [NT]: Update get_dir() to accept/return UTF8. Stdio [NT]: Cleanup filesystem_stat(). Stdio [NT]: Clean up getcwd(). I/O [NT]: Added fd_get_current_dir_name(). Configure: Check for getwd() and get_current_dir_name(). Stdio [NT]: Clean up cd(). ...
Merge branch 'grubba/fdlib' into 8.1 * grubba/fdlib: Stdio: Cleanup File()->truncate(). Configure: Moved tests for truncate64() and ftruncate64().
Configure: Moved tests for truncate64() and ftruncate64().
Merge branch 'grubba/fdlib' into 8.1 * grubba/fdlib: (25 commits) Stdio [NT]: Fixed warning on NT. Stdio [NT]: Added detection constant __HAVE_UTF8_FS__. I/O [NT]: Added note about the intent of fdlib. Stdio [NT]: Update get_dir() to accept/return UTF8. Stdio [NT]: Cleanup filesystem_stat(). Stdio [NT]: Clean up getcwd(). I/O [NT]: Added fd_get_current_dir_name(). Configure: Check for getwd() and get_current_dir_name(). Stdio [NT]: Clean up cd(). I/O [NT]: Added fd_chdir(). Stdio [NT]: Clean up mv(). I/O [NT]: Added fd_rename(). I/O [NT]: Added facility for dynamic lookup of library symbols. Stdio [NT]: Clean up mkdir(). I/O [NT]: Added fd_mkdir(). Configure: Moved API test of mkdir(). Stdio [NT]: Clean up rm(). I/O [NT]: Added fd_unlink(). I/O [NT]: Added fd_rmdir(). Stdio [NT]: Clean up file_truncate(). ...
Configure: Check for getwd() and get_current_dir_name().
Configure: Moved API test of mkdir().
Configure [Solaris]: Improved compat with old system header files. Old Solaris header files don't like __STDC__ being set to 1 (or higher), at which point they start using unions for 64-bit integers. This causes other headerfiles to generate broken macros (cf <sys/types.h> longlong_t and <sys/resource.h> RLIM_INFINITY).
Configure: Detect if off64_t is not a scalar.
Merge commit '75c9d1806f1a69ca21c27a2c2fe1b4a6ea38e77e' into patches/pike63 * commit '75c9d1806f1a69ca21c27a2c2fe1b4a6ea38e77e': (19587 commits) ...
[build] Explicitely check --with-mpi in sub configure.in files The ac_user_opts related variables do not always get propagated in the way that src/modules/configure.in expects. Checking --with-mpi explicitely in sub-configures that use the setting gets around the issue. AC_ARG_WITH lowercases the variable name as well, so src/modules/configure.in now also lowercases ${with_{$MODNAME}} when checking whether to build a module.
Merge branch '8.1' into peter/travis
Build [x86/x86_64]: Suvive SSE with old gcc. Fixes compilation issues with some older versions of gcc.
First go at MPI bindings
Restore part of AmigaOS support.
Rip out final remains of OS2 and AmigaOS.
Remove computed goto interpreter mode By descision of the Pike conference, computed goto mode was removed.
Remove !SHARED_NODES Pike with SHARED_NODES disabled did not work. By descision of the Pike Conference it is removed.
smartlink: NetBSD uses the Darwin convention for rpath.
smartlink: support Darwin-style rpath
Merge branch 'marty/call_frames' into 8.1 This introduces the --with-mc-stack-frames configure option, which will instruct the machine code generator to insert proper stack frames (currently only supported on X86-64). This is useful for profiling, especially in combination with Debug.generate_perf_map() on Linux.
sys/types.h has been part of POSIX since issue 1. It was randomly guarded in configure tests, so lets use it without ifdef.
Build [Stdio]: Prefer <termios.h> to <sys/termios.h>. Use the same header file consistently. Fixes compilation warnings on FreeBSD 10.3.
Remove the GCC patch workaround for Solaris.
Only test for sys/types.h once.
TenDRA/TCC is abandoned since 2007.
We are only using VPATH=@srcdir@ so let's test for VPATH=
Last major release of Irix was in 1998 with support updates until 2006 and declared unsupported 2013. Remove support.
Remove support for AIX <5 (end of support 2004).
Build: Attempt to get both Solaris and OSX to build with c99 support.
Build: -std=c99 needs to be in $CPP too. Potential fix for configuration failures on Solaris.
Build: Move -std=c99 flag to $CC. The default rules for autoconf and make apparently don't differ between flags fro the C and c++ preprocessors.
Revert "Build: -std=c99 should be a C-compiler flag, not a preprocessor flag (the clang C++ preprocessor rejects it)" This reverts commit 480238bb4720f465f266c0a86dfb9fe57b88e9e8. -std=c99 affects the behavior of the C-preprocessor, and must thus be in CPPFLAGS. CPPFLAGS should however NOT be used with a C++ compiler, where CXXPPFLAGS should be used. Fixes configuration failures on Solaris.
Build: -std=c99 should be a C-compiler flag, not a preprocessor flag (the clang C++ preprocessor rejects it)
build: Drop accidental space.
build: NetBSD has bswap(16|32|64) in libc, so we should avoid redefining them.
Compiler: Provide descriptive strings for all tokens. Instead of showing the internal TOK_* symbol in parser errors (like eg 'TOK_LOR'), the parser should now show its corresponding token (like eg '||'). Fixes [bug 2711 (#2711)].
Build: Set and use the CONFIG_HEADERS variable. The main Makefile.in should now validate. Fixes testsuite failure.
Fix reversed test (this was NULL_IS_ZERO before)
A first draft of reduction of object C initializers.
Configure: Define _POSIX_C_SOURCE after configuring C compiler. Fixes configuration issues on eg Solaris due to C- and POSIX- standard mismatches.
Configure: C99 affects the preprocessor too... NB: On Solaris XPG3, XPG4, XPG4v2, or XPG5 as well as POSIX.1-1990, POSIX.2-1992, POSIX.1b, and POSIX.1c are invalid with C99, while XPG6 and POSIX.1-2001 are invalid without C99.
Configure: Added macro PIKE_USE_SYSTEM_EXTENSIONS.
Build [POSIX]: #define _POSIX_C_SOURCE. Eg clock_gettime() and friends are sometimes not available without this.
Build [NT]: Fix building of post_modules with RUNPIKE. BUILD_PIKE is used by DEFAULT_RUNPIKE as is, and the EXEEXT breaks the build as it isn't valid on the UNIX side. Potentially fixes building of eg GTK2 on NT.
Revert "rint/rintf/rintl is part of C99." MSVC lagged in implementing the C99 standard, and apparently first started providing these functions in 2013. This reverts commit de299dab684bfbe9850e56e324bc583662ec0e36.
Build: Support compilers without __builtin_constant_p().
Configure: Probe support for C99-style struct literal expressions. In C99 the expression syntax was extended to allow struct literals in expressions, provided that they are preceeded by a suitable cast. Adds macro CAST_STRUCT_LITERAL(), and updates buffer.h to use it. Potential fix for build failures on NT.
Faster string_to_utf8() This commit splits the length calculation of the resulting string and the actual encoding into two new functions. This makes it possible to e.g. encode utf8 directly into a buffer. The length calculation has been rewritten for different shift sizes. For 8bit strings it uses a popcount loop, which calculates the number of high bits (code points bigger than 0x7f) on machine size chunks. On machines which have popcount instructions this is much faster. With compilers which do not support __builtin_popcount it uses a simple manual popcount. For 16bit and 32bit strings the length calculation uses clz to count the number of bits in the codepoint to calculate the length without branches. The encoding function is split into one version for each shift size. For 32bit strings it avoids branches by using the resulting byte lengths as a jump size. This generates reasonable code, at least in gcc. Benchmark results on my i7: utf8/code.pike#encode_7bit | 1.3 G 1.6 % | 8.3 G 3.4 % | utf8/code.pike#encode_8bit | 651.1 M 1.8 % | 1.1 G 1.2 % | utf8/code.pike#encode_arabic | 498.4 M 0.8 % | 710.3 M 1.2 % | utf8/code.pike#encode_bulgarian | 488.2 M 1.2 % | 688.4 M 2.6 % | utf8/code.pike#encode_estonian | 614.8 M 6.6 % | 969.5 M 1.5 % | utf8/code.pike#encode_hebrew | 496.9 M 1.8 % | 710.1 M 1.0 % | utf8/code.pike#encode_japanese | 704.9 M 4.0 % | 785.4 M 1.6 % | utf8/code.pike#encode_polish | 388.9 M 0.4 % | 710.1 M 1.3 % | utf8/code.pike#encode_thai | 642.8 M 3.3 % | 858.0 M 0.9 % | utf8/code.pike#encode_yiddish | 485.9 M 3.3 % | 692.5 M 3.8 % | I also tested on arm32, the speedups are around 50%.
Configure: Potential fix for dynamic modules on FreeBSD 10.1. The LDSHARED test attempted to link the shared object with crt1.o, which failed due to relocation errors. As we don't want to link with the startup files anyway, make sure to ask gcc to not do it.
New byte buffer This new byte buffer implements a simple dynamic buffer. It aims to * Provide a clean API. * Convince the compiler to generate good code. In particular it tries to avoid unnecessary loads and stores which are otherwise generated due to the C aliasing rules. This is done using the STATIC_ASSUME(X) macro, which declare !X as unreachable. * Have the fast path (when the buffer has enough space) completely inlined to allow the compiler to batch single byte writes to the buffer into larger stores.
Build [C++]: Fixed warning about unsupported option. The option -Wimplicit-function-declaration is not supported with C++ (as it is required behavior).
Build [C++]: Improved support for modules written in C++. Differentiate between CFLAGS and CXXFLAGS.
Build [NT]: Test for C99-style struct initialization. MSVC lagged in implementing the C99 standard, and apparently first started supporting C99-style structure initialization in 2013. Fixes compilation issues on NT (and other platforms with old compilers).
Build: activate ARM32 machine code in Thumb mode After some testing, it seems like interworks just works as expected.
ARM64: Initial commit
Skeleton implementation of AARCH32 (ARM) machine code support To be extended, also into AARCH64 machine code generation Author: Arne Goedeke <el@laramies.com> Author: Tobias S. Josefowitz <tobij@tobij.de>
LONGEST is just INT64, so use that instead.
Removed IA64 configurations.
Build [Configure]: No need to disable SSE2 anymore.
Build [Configure]: Avoid SSE opcodes on 32-bit x86. Potential fix for SIGSEGVs in SSE-code in multiple places for 32-bit ABI on x86_64.
Build [Configure]: Second attempt at fixing --with-abi. The previous patch broke the AC_CHECK_SIZEOF-tests when using a non-default ABI.
Build [Configure]: Strip arguments from REALCC et al. Fixes compiler propagation to sub-configures when eg compiling --with-abi.
Since we now only use rdrand64, test for that instead of rdrand32.
Allow "paranoia mode" code compilation
More readable macros.
Build [Gmp]: Try using pkg-config to find gmp and mpfr. Solaris 11 installs the header files for gmp and mpfr in sub- directories of /usr/include/, but it also installs corresponding pkg-config files. Use pkg-config to adjust the search paths so that the following configure tests can find the header files.
Simplified MAKE_INF and MAKE_NAN. Hopefully not too much.
Revert "infnan isn't used." This grep-dosn't-show-all-hits issue is annoying... This reverts commit 18a4150c9710e43e702c6ede5e36592a2797d091.
infnan isn't used.
isnan is part of C99. Assume it exists unless we know it does not.
Union init is part of C99 init
inline is part of c99 We do have.. excessive amounts of inline, incidentally. :)
Don't warn about declaration after statement.
Let's assume that finite really is _finite from Windows CRT.
rint/rintf/rintl is part of C99.
vsnprintf is part of C99.
snprintf is part of C99.
Added -std=c99 compileflag.
Revert "setjmp and longjmp are standardized C89." This reverts commit 0e9cd75e000d4d87c212721cb3494f609632807c. They are standardized, but we only use them as fallback if sigsetjmp() doesn't exist.
setjmp and longjmp are standardized C89.
No need to check for stddef.h. It's part of C89.
No need to check for time.h. It's part of C89 and we don't use HAVE_TIME_H everywhere.
No need to check for string.h. It's part of C89 and we don't use HAVE_STRING_H everywhere.
Don't use STDC_HEADERS inconsistently.
No need to check for stdlib.h. It's part of C89 and we don't use HAVE_STDLIB_H everywhere.
No need to check for stdio.h. It's part of C89.
No need to check for signal.h. It's part of C89 and we don't use HAVE_SIGNAL_H everywhere.
No need to check for setjmp.h. It's part of C89.
No need to check for locale.h. It's part of C89 and we don't use HAVE_LOCALE_H everywhere.
No need to check for limits.h. It's part of C89 and we don't use HAVE_LIMITS_H everywhere.
No need to check for errno.h anymore.
No need to check for float.h. It's part of C89 and we don't use HAVE_FLOAT_H anywhere.
Always have PORTABLE_BYTECODE enabled. This has been the default for quite some time, and encode/decode of programs is not all that useful without it, really.
Merge branch '8.1' into per/substrings
We don't need to test for __builtin_expect twice.
We don't use URCU anymore.
Removed unused headers.
No need to check for the same headers more than once.
Assume that builtins never worked and fix them by removing one level of quoting.
Tweak OS X 64-bit compilation by not specifying old optimization flags when the defaults are better (+3% on average on pike -x benchmark on MBP Core i7). Also avoid -read_only_relocs since it's not applicable on 64-bit ABI.
Switch from finite() to isfinite() since the former is flagged as deprecated on OS X.
Removed trailing spaces.
Configure: Fixed svalue size test.
Silence a ton of warnings in clang since the UNUSED() family of macros now depend on HAVE_FUNCTION_ATTRIBUTES being available, and the configure test refused to set that if it detected __declspec before.
Configure: TEST_BUILTIN{,_VOID} now know about the configure cache.
MPI: autohell butchering
Build [NT]: Reenable the configure cache by default for WIN32. Running uncached configure with rntcl et al takes considerable time...
build: complete removal of --enable-dll option in order to build successfully on Windows.
Build: Moved OS checking code to aclocal.m4.
Remove deprecated security system.
add popcount32 and popcount64
build: modules that invoke pike for building should now be able to do so.
Move strdup from pike_memory to port.
Build: Support compilers with union aliasing problems. There are apparently compilers that don't understand that the fields in unions overlap, and overcache overwritten fieldvalues. Also restores support for compilers without support for static union init.
Removed ATOMIC_SVALUE, it does not compile and does not work.
Faster svalue type/subtype setting The code generated for setting two shorts (on x86/x86_64 and arm at least) is very sub-optimal, especially with gcc, for some reason: Old push_int(0): movq Pike_interpreter_pointer(%rip), %rdx movq (%rdx), %rax leaq 16(%rax), %rcx movq %rcx, (%rdx) xorl %edx, %edx xorl %ecx, %ecx movw %dx, (%rax) movw %cx, 2(%rax) movq $0, 8(%rax) New push_int(0): movq Pike_interpreter_pointer(%rip), %rdx movq (%rdx), %rax leaq 16(%rax), %rcx movq %rcx, (%rdx) movq $0, (%rax) movq $0, 8(%rax) Except for the lower number of instructions there is an additional benefit: The old code triggered a read-modify-write operation on most modern x86 CPU:s, all to preserve the undefined data between subtype and the value of the svalue. This could be fixed by changing the type and subtype to be 32-bit instead of 16-bit, but that is a bigger change.
Removed the support for mmx.h This was a header file that existed back in ye good old time, allowing usage of MMX without much in the way of compiler support. However, it is not really available anywhere any more, and almost all code that used it was disabled. If we do feel the need for MMX (or rather SSE3+, really, MMX is rather old-school these days) we should use compiler intrinsics.
Removed the old typechecker. We really should stopping just keeping old code around. It only confuses things when you are reading the code.
Enable unused function warnings again.
Runtime: Increase the default thread C stack size to 1MB. The old default (256KB) is a bit too little on current 64-bit hardware. This essentially reverts 78797d06 (aka src/threads.c:1.157). Most likely fixes [bug 7061 (#7061)].
Added LinuxThreads configure test. HAVE_BROKEN_LINUX_THREAD_EUID is now only defined when euid is indeed broken.
Disabled unused-function. We have too many of those.
-ggdb gives more debug information than -ggdb3 with modern gcc:s.
Added some more warnings by default, and test for some more builtins. The __builtin_expect is now tested for together with the other builtins, which makes it available when compiling with clang. This helps the index_shared_string function be significantly faster when indexing narrow shared strings (which is assumed to be most common). Of course, if you generally speaking index more wide strings than shared strings it's slower. Then again, it you have "medium" wide strings (2 bytes per character) the performance is identical (both gcc and clang had the tests in 2,1,0 order).
Build system: Fixed some issues with semidynamic modules. The post_modules were not linked properly.
Build system: Improved simulation of dynamic modules. Dynamic modules are now simulated with static modules when not available. This includes having stub *.so files that are loaded with load_module(). This fixes issues with eg loading Nettle without dynamic modules.
Configure: More work-arounds for Autoconf 2.50... Autoconf 2.59 thinks it is a good idea to pass the original environment variables (like eg $CC) on the command line to sub-configures, which means that smartlink won't propagate causing failures in the sub-configures.
Configure: Halt the OOB check on poll error to prevent hang
Configure: Improved robustness of stack depth test. Don't trust the stack depth test if it determines the stack depth to be 32KB. Note that Pike requires at least 64KB of stack anyway. Fixes failure where Pike_INITIAL_STACK_SIZE was set to 32KB, causing a fatal error during initialization of the run-time.
Backend: Darwin's poll(2) is based on kqueue(2), so there's no benefit to supporting poll independently.
Now works without SelectBackend.
Configure: Disable use of config.cache by default. By popular demand.
Fixed tests for compiler intrisics when not using gcc
Removed non-DLL implementation of dynamic modules for Windows It has not actually compiled for quite some time, so it is not very likely to be used all that much.
Configure: Use AC_TRY_LINK to test for builtins Using AC_TRY_RUN makes it not work for cross compilation, and the runtime test didn't do anything anyway...
Configure: Fix putenv test for cross compilation
pike_memory: Use regular memalign if posix_memalign is unavailable
Configure: Added detection of liburcu.
Remove --without-bignums.
Added new block allocator. It dramatically speeds up free, when allocating many blocks and deallocation happens non linearly.
Removed the system for bundling 3rd party libraries with Pike.
Remove the remaining $.
Deprecate the security system.
Removed the experimental facets implementation.
Added a low-level wrapper for struct tm: System.TM This can be used to do (very) simple calendar operations. It is, as it stands, not 100% correct unless the local time is set to GMT, and does mirror functionality already available in gmtime() and localtime() and friends, but in a (perhaps) easier to use API.
Configure: Enable the full address space on NT. The default address space on NT is just 31 bits. Attempt to get the linker to enable the full addressspace.
configure: Don't execute $CFLAGS as a command
configure: Add -mcpu=niagara to CFLAGS on sun4v
Makefiles: Added a few more @EXEEXT@s. Hopefully the generated Makefiles should now be viable on OS/2. Thanks to Chris Angelico <rosuav@gmail.com> for the reminder.
Precompile.sh: Support other path separators than :.
Add detection of unsigned 128-bit integer. Define UINT128 as distinct type since "unsigned __int128_t" does not always work even when __int128_t does. Adds UINT64 as well for macro expansion in bignum.h.
bignum: added standard compliant overflow checks
Various clang-related changes to reduce compile-time warnings.
Configure: fixed test for __attribute__ support va_start and other macros expect the va_list as first argument. This test only produced warnings with most compilers, which is why this problem was never fixed.
Build: Added configure option --enable-werror. The configure option --enable-werror now adds -Werror to the set of WARN flags.
Less AUTO_BIGNUM
Removed security system (on this branch)
Removed MAX_FD configure test It only returns the current max anyway, which is usually smaller than the max if you are running as root, and can also be changed at any time at runtime. Instead, use sysconf(_SC_OPEN_MAX) when available if needed, and as a last resort a hardcoded value (on Windows), but as a secondary change all uses except two was removed: o Stdio.get_all_active_fd - On systems without /dev/fd and /proc/self/fd. o On NT, the size of FD_SET is now set to a hardcoded value (64k). The rewrite of get_all_active_fd also speeds it up a lot. Not that it is a frequently used function.
Unified the automatic bash selection We already did select bash in the toplevel makefile to run configure if desired, we now set the CONFIG_SHELL variable there as well. This makes the test in configure.in 100% instead of 99.9% redundant.
Change from 10 to 0.5 second timeout in OOB testing. If we have not received the data we are sending to ourselves in .5 seconds it is higly unlikely we will get it any time in the next 9.5 seconds.
Don't fail twice.
Runtime: Added check and fallback for accept4(2). accept4(2) is a system call available on current Linux and many BSDs, that allows for setting the close-on-exec and non-blocking flags on the resulting fd.
Merge remote-tracking branch 'origin/8.0' into string_alloc Conflicts: src/stralloc.c
signal is C89, so assume it exists.
strerror is C89 , so assume it exists.
setlocale is C89, so assume it exists.
setbuf and setvbuf are C89, so assume they exists.
perror is C89 and 4.3BSD (and the result of this test isn't used).
localtime, gmtime and mktime are C89 and 4.3BSD, so assume they exists
getenv is C89 and 4.3BSD, so assume it exists
isgraph is C89 and 4.3BSD, so assume it exists.
strcoll is C89 and 4.3BSD, so assume it exists.
clock is C89, so assume it exists.
ldexp is C89 and 4.3BSD
frexp is C89 and 4.3BSD
vsnprintf is C89
strchr is C89 and 4.3BSD
memchr is C89 and 4.3BSD
memcmp is C89 and 4.3BSD
memmem isn't used.
memmove is C89 and 4.3BSD
Don't cast memset void* argument.
memcpy is C89 and 4.3BSD
bcopy was deprecated 2001.
memset is C89 and 4.3BSD
STRNLEN fallback isn't used anywhere.
This fallback never worked.
strtod is C89
strtol is C89 and 4.3BSD
We are not actually using HAVE_WORKING_STRTOL anywhere anymore.
time is C89 and 4.3BSD
BSD function index is deprecated from 2001, removed 2008.
Building without issapce has not worked since 1999
Building without vfprintf has not worked since 2003.
Removed some unused tests.
It hasn't been possible to compile Pike without strncmp since 2003. Remove port code for it.
decode_value: make old style encoding of programs optional This patch adds a new configure argument --with-compat-encode, which enables support for old style program decoding. This saves about 12 kB in decode_value2 with -Os.
Configure: Inhibit machine code with gcc 4.6.0 and later. The machine code generator is broken when compiled with gcc 4.6.0 and later, so disable it in that case.
Merge branch '8.0' into gobject-introspection
Merge remote-tracking branch 'origin/7.9' into pdf
new block alloc
Merge remote-tracking branch 'origin/7.9' into ba Conflicts: src/interpret.c src/interpret.h src/pike_embed.c
Merge remote-tracking branch 'origin/7.9' into ba
ammend this
added standard compliant overflow checks
included GJAlloc
Merge branch '7.9' into gobject-introspection
Runtime: Tune malloc(3C) in glibc. Alleviates [bug 6045 (#6045)]. Malloc(3C) in glibc defaults to having one pool (aka arena) of memory per thread. This means that by default memory released by one thread will not be reused by another thread. This in turn can cause a "pumping" effect where one thread temporarily may use a lot of memory, which causes that thread's arena to grow. If the same code is run again later, but by a different thread, that threads arena will grow even though there's plenty of free memory in the old arena. This means that there's a multiplication effect by the number of active threads. This implements the suggested work-around from the similar bug at http://sourceware.org/bugzilla/show_bug.cgi?id=11261
Build: exclude mach headers on Darwin releases where these conflict with threads. --HG-- extra : rebase_source : 9b288b4b9f515e6dad7b68f508cfc37d4fd08043
Merge remote branch 'origin/7.9' into block_alloc
Configure: Terminate the OOB test early on ETIMEDOUT.
Merge remote branch 'origin/7.9' into rblock_alloc Conflicts: src/post_modules/CritBit/floattree.cmod src/post_modules/CritBit/inttree.cmod src/post_modules/CritBit/stringtree.cmod
Avoid double define of PIKE_NEW_BLOCK_ALLOC.
configure: Fix issue with autoconf 2.63 and later. Autoconf 2.63 and later created a broken config.status on OSes where /bin/sh is not bash.
fix bootstrapping issue
add GJAlloc as a bundle
Merge remote branch 'origin/7.9' into breaking_into_pieces
Configure: Added check for mallinfo(3MALLOC).
Pike.PollDeviceBackend: add GOT_CORE_FOUNDATION and enable_core_foundation(). When enabled, backend will use CoreFoundation (CFRunLoop) to perform IO polling. This allows CF enabled functionality (eg. FSEvents, SDL, DNS_SD) to function properly.
block_allocator: initial commit
Configure: Potential fix for VCDiff failure.
Configure: Added --with-atomic-svalue.
Configure: Cosmetic fix.
Added support for bzip2'ed bundles as well.
Do not restart the configure with bash if already running bash. Backport from 7.8, commit d9caad7.
Implement --with-dtrace to enable source-level tracing of Pike programs using e.g. Apple Instruments. (It's not yet balancing entry/exit properly.)
Fixed configure test (and lots of spellos) for CRC32 intrinsics.
Use the CRC32 instrisics when available.
Compiler: Enable machine-code for amd64 by default.
Use -fvisibility=hidden. This allows gcc to optimize function calls to globally visible, but not exported (using PMOD_EXPORT) functions to a direct call. Saves a few % CPU, and also decreases the size of the binary. However, it is now important to use PMOD_EXPORT correctly on any system, not only Windows. It also significantly speeds up loading of dynamic modules, but that does not generally speaking use any significant amount of CPU anyway.
Do not look for the unused tallow tool. It's not even part of WiX any longer since version 3.
Autoconf: Ensure that $CC and $CXX are environment variables.
Make sure that $CXX is set before we start checking for ABI flags.
Improved support for modules needing C++ compilers.
Fixed rdtsc for msvc.
Limit Darwin fd count to OPEN_MAX instead of 2 * 1024 * 1024. Fixes [bug 5751 (#5751)].
Use rdtsc if available to cut syscalls in check_threads.
Added missing #include of <stdio.h> in feature summary program. Rev: src/configure.in:1.1093
Added missing #include of <stdio.h> in feature summary program. Rev: src/configure.in:1.873
Added missing #include of <stdio.h> in feature summary program. Rev: src/configure.in:1.704
Now attempts to set up the PKG_CONFIG_PATH. Rev: src/aclocal.m4:1.170 Rev: src/configure.in:1.1092
Lots of cosmetic fixes. Reduced the number of chained if-else statements by using elif in more places. Rev: src/configure.in:1.1091
Fix abundance of warnings when building on OS X 10.6 (at least on x86_64) since the compiler didn't know the minimum target OS version. Also add -mfpmath=sse on all 32-bit OS X/x86 builds since we always have SSE present (the 64-bit get this by default). Rev: src/configure.in:1.1090
Fixed a few warnings in a configure test. Rev: src/configure.in:1.1089
Put /usr/X before /opt/csw and its ilk in the search order of standard directories. Rev: src/configure.in:1.1088
When using gcc -print-search-dirs to add ABI library directories, make sure to do it _after_ adding other ABI directories, so that their relative position in the link order is preserved. If this is not done so, the order between lib directories will not match the order between include directories, leading to weird problems if a package exists in more than one location. Rev: src/configure.in:1.1087
Some potential fixes for --enable-pikelib mode. Rev: src/configure.in:1.1086
Improved libdir abi handling. Rev: src/Makefile.in:1.480 Rev: src/configure.in:1.1085
Added check for an abi indicator for the installation lib directory. Rev: src/configure.in:1.1084
Fixed some typos. Rev: src/configure.in:1.1083
Kludge for snprintf on WIN32. Rev: src/configure.in:1.1082 Rev: src/port.h:1.66
Request WinNT 5.0 APIs again. Rev: src/configure.in:1.1081 Rev: src/global.h:1.118
Added check for inet_ntop(). Rev: src/configure.in:1.1080
Bumped the WIN32 APIs from 5.0 to 6.0. Rev: src/configure.in:1.1079 Rev: src/global.h:1.117
Improve performance of get_cpu_time() on OS X by avoiding a kernel trap to get the ID of the underlying system thread. Rev: src/configure.in:1.1078 Rev: src/rusage.c:1.53
Added check for sys/uio.h. Rev: src/configure.in:1.1077
Moved {send,recv}msg() stuff from modules/files/configure.in to the main configure.in. Rev: src/acconfig.h:1.161 Rev: src/configure.in:1.1076 Rev: src/modules/files/acconfig.h:1.31 Rev: src/modules/files/configure.in:1.114
Check for and use sched_yield(2) for yielding, since pthread_yield is falling out of grace. Rev: src/configure.in:1.1075 Rev: src/pike_threadlib.h:1.67
Check for pthread_cond_reltimedwait_np(3T). Rev: src/configure.in:1.1074
Can just as well ensure -Wformat-security is enabled regardless of distro defaults. Rev: src/configure.in:1.1073
Added support for Mach high resolution time, using thread_info() for cpu time and host_get_clock_service(SYSTEM_CLOCK) for real time. Note that the resolution returned by clock_get_attributes for the system clock on OS X is very bad (10 ms), but the time values actually delivered appears to be much better (on nanosecond scale). Rev: src/configure.in:1.1072 Rev: src/rusage.c:1.51
Backported windows fix from 7.8: Test winsock2.h before most other headers to avoid conflict with winsock.h. Rev: src/configure.in:1.703
Explain --with-dmalloc-c-stack-trace better. Rev: src/configure.in:1.1071
Made support for C stack traces in dmalloc optional, since it makes Pike more than ~2 times slower. Rev: src/acconfig.h:1.160 Rev: src/configure.in:1.1070 Rev: src/dmalloc.h:1.67
Enable two-level namespace in OS X linker (10.3 or newer) to avoid crashes caused by multiply defined library symbols. Rev: bin/smartlink:1.31 Rev: src/acconfig.h:1.159 Rev: src/configure.in:1.1069 Rev: src/smartlink.c:1.21
The smartlink binary is now built and used by default again. Rev: src/configure.in:1.1068
Disabled the windows dll load error dialogs by default and added an option -rl to enable them. Rev: src/configure.in:1.1067 Rev: src/main.c:1.232 Rev: src/pike_embed.h:1.11
Spello. Rev: src/configure.in:1.1066
Fixed problem with AC_PROG_CC not putting -g into CFLAGS due to the variable already being set by the icc workaround test. Rev: src/configure.in:1.1065
On Cell, use -mcpu=cell if available. Rev: src/configure.in:1.1064
Another attempt with enabled -O3. If it doesn't work then it hopefully can be disabled on a more specific basis. Rev: src/configure.in:1.1063
Need -m64 and more in order to make LDSHARED work on OS X in 64-bit mode. Rev: src/configure.in:1.1062
Added some tuning for ia64. Rev: src/configure.in:1.1061
Improve PPC G4/G5 scheduling flags slightly. Rev: src/configure.in:1.1060
Fix for linking GMP on Darwin PPC using GCC 4.x. Rev: src/configure.in:1.1059
Added checks for {,un}setenv(3). Rev: src/configure.in:1.1058
Building external modules normally doesn't require anything but the location of the Pike C headers. This patch tries to save the header location in the master and provides a pike -x cflags command to extract the compiler flags for using them. Rev: bin/install.pike:1.187 Rev: lib/master.pike.in:1.448 Rev: lib/modules/Tools.pmod/Standalone.pmod/cflags.pike:1.1 Rev: src/Makefile.in:1.466 Rev: src/configure.in:1.1057
If the path to the master is relative then resolve it relative to the exe file on windows. This allows us to skip the binary patching of the binary. Rev: src/configure.in:1.1056 Rev: src/main.c:1.230
Use the stack inspection feature of newer gnu libc's to add some handy C stack backtrace reporting to dmalloc. Rev: src/configure.in:1.1055 Rev: src/dmalloc.h:1.64 Rev: src/pike_memory.c:1.192
Fixed putenv on windows (hopefully). Rev: src/builtin.cmod:1.205 Rev: src/configure.in:1.1054
Even more paranoid... Rev: src/configure.in:1.1053
Increased paranoia in initial stack limit check. Rev: src/configure.in:1.1052
Fixed overflow error in mktime when sizeof(time_t) > sizeof(INT_TYPE). This fixes [bug 4525 (#4525)]. Fixed gmtime, localtime and ctime to handle bignums in the same circumstances. Rev: src/builtin.cmod:1.109 Rev: src/builtin_functions.c:1.481 Rev: src/configure.in:1.702 Rev: src/global.h:1.81
Fixed configure stuff for va_copy. Rev: src/acconfig.h:1.158 Rev: src/configure.in:1.1051 Rev: src/error.c:1.153
Use exit rather than return to terminate subshells. Rev: src/configure.in:1.1050
Compilation now nolonger halts when a bundle fails to compile. Rev: src/configure.in:1.1049
Improved support for bundles using pkg-config. Rev: src/configure.in:1.1048
Let's try once more... Rev: src/configure.in:1.1047
Bundles may now use pkgconfig. Rev: src/configure.in:1.1046
Added missing export of pike_bundle_prefix. Rev: src/configure.in:1.1045
Use bundle for libffi. Rev: bundles/.cvsignore:1.3 Rev: bundles/libffi.txt:1.1 Rev: src/configure.in:1.1044 Rev: src/modules/Java/configure.in:1.64
Mark bundles that fail to configure as installed so that the module configure scripts have a chance at not using the bundle on unsupported architectures. Rev: src/configure.in:1.1043
Help Nettle DTRT when cross-compiling (without rntcl). Rev: src/configure.in:1.1042
Help Nettle DTRT when cross-compiling (without rntcl). Rev: src/configure.in:1.872
Well, that didn't work... Rev: src/configure.in:1.1041
More paranoia... Rev: src/configure.in:1.1040
Added some protection against infinite looping in the bundle-extraction code. Rev: src/configure.in:1.1039
Eradicated all instances of setting type to T_INT without also setting the subtype in an svalue. From now on one should either set the subtype or use the new svalue type PIKE_T_FREE (see svalue.h). Rev: src/OCPikeInterpreter.m:1.3 Rev: src/apply_low.h:1.32 Rev: src/array.c:1.198 Rev: src/array.h:1.71 Rev: src/bignum.c:1.42 Rev: src/builtin.cmod:1.193 Rev: src/builtin_functions.c:1.650 Rev: src/configure.in:1.1038 Rev: src/cpp.c:1.164 Rev: src/encode.c:1.250 Rev: src/error.c:1.151 Rev: src/interpret.c:1.385 Rev: src/interpret_functions.h:1.197 Rev: src/iterators.cmod:1.64 Rev: src/las.c:1.403 Rev: src/main.c:1.228 Rev: src/mapping.c:1.195 Rev: src/module.c:1.47 Rev: src/modules/Gmp/mpz_glue.c:1.170 Rev: src/modules/HTTPLoop/accept_and_parse.c:1.39 Rev: src/modules/HTTPLoop/requestobject.c:1.31 Rev: src/modules/Image/colortable.c:1.130 Rev: src/modules/Java/jvm.c:1.84 Rev: src/modules/Parser/html.c:1.178 Rev: src/modules/Parser/xml.cmod:1.97 Rev: src/modules/Postgres/postgres.c:1.55 Rev: src/modules/_math/math.c:1.87 Rev: src/modules/files/sendfile.c:1.78 Rev: src/modules/spider/spider.c:1.134 Rev: src/multiset.c:1.106 Rev: src/object.c:1.281 Rev: src/operators.c:1.226 Rev: src/pike_error.h:1.42 Rev: src/post_modules/Shuffler/Shuffler.cmod:1.45 Rev: src/post_modules/_ADT/circular_list.cmod:1.18 Rev: src/post_modules/_ADT/sequence.cmod:1.19 Rev: src/program.c:1.642 Rev: src/signal_handler.c:1.331 Rev: src/sscanf.c:1.170 Rev: src/svalue.c:1.234 Rev: src/svalue.h:1.149
Disabled the stack ulimit check. Rev: src/configure.in:1.1037
Increased reduced stack from 4096 to 8192. Rev: src/configure.in:1.1036
Added code to reduce the stack ulimit if it exceeds 64MB. It seems the thread tests fail if the stack ulimit is too large on Linux 2.6/IA64. Rev: src/configure.in:1.1035
Moved test for libgcc to the module configure scripts, so that the installed dynamic_module_makefile avoids containing stale values for it. Rev: src/aclocal.m4:1.160 Rev: src/configure.in:1.1034 Rev: src/make_variables.in:1.29 Rev: src/modules/dynamic_module_makefile.in:1.124
Extracted the libgcc test from the main testsuite in preparation for integrating it with AC_MODULE_INIT. Rev: src/aclocal.m4:1.159 Rev: src/configure.in:1.1033
Backported PIKE_FUNCS_NEED_DECLS stuff from Pike 7.7. Rev: src/configure.in:1.701
Added check to enable gcc-warning for the C9x-ism of allowing variable declarations after statements. Rev: src/configure.in:1.1032
Removed the -ansi related stuff since it broke too many things.. Rev: src/acconfig.h:1.157 Rev: src/configure.in:1.1031
Bugfix. Rev: src/configure.in:1.1030
Added some symbols needed to survive strict ANSI-mode. Rev: src/configure.in:1.1029
Added some debug checks. Rev: src/configure.in:1.1028
* Added /opt/csw/lib to list of library directories to check for * If compiler is GCC, also add all library directories listed by gcc -print-search-dirs * Normalize library directory pathnames before adding them Rev: src/configure.in:1.871
Added some extra diagnostics for when posix threads aren't working. Rev: src/configure.in:1.1027
Avoid using pthread-config on FreeBSD 5.x or later. Rev: src/configure.in:1.1026
Moved down the tests using port.c as far as possible to ensure prerequisite tests have been made (specifically the IEEE float tests when compiling on windows). Rev: src/configure.in:1.870
Allow machine code for gcc 4.0 too; it appears to work on at least 4.0.4 20060630 (prerelease). Enable machine code regardless of gcc version if --with-machine-code is given explicitly. Rev: src/configure.in:1.700
The ugly hack to fix machine code jumps by changing ((unsigned char **)__builtin_frame_address(0))[1] doesn't work with gcc 4.x on ia32 (at least). Machine code is therefore disabled for anything but gcc 3.x. (Later pikes use a better method.) Rev: src/configure.in:1.699
Keypair loop is currently broken (mapping.c:GC_RECUR() doesn't survive moving keypairs...). Rev: src/configure.in:1.1025
Fixed typo regarding type checker selector default. Rev: src/configure.in:1.1024
The new function call type checker is now enabled by default. Keypair loop is now enabled by default. Now selects the medium memory model when compiling with cc on Solaris/amd64. Rev: src/configure.in:1.1023
Moved down the tests using port.c as far as possible to ensure prerequisite tests have been made (specifically the IEEE float tests when compiling on windows). Rev: src/configure.in:1.1022
Yet another fix in the CONFIGURE_TEST tests. Rev: src/configure.in:1.1021
Further fixes in the CONFIGURE_TEST tests to avoid compilation trouble. Adaption in the thread local cpu time test for change in pike_rusage.h. Rev: src/configure.in:1.1020
Added support for POSIX style timers using clock_gettime(3). Notably this fixes nice high resolution thread local cpu time and monotonic real time on reasonably modern Linux systems. Also added a few more pike constants about various characteristics of the cpu/real time interfaces in use. Rev: src/acconfig.h:1.155 Rev: src/builtin_functions.c:1.640 Rev: src/configure.in:1.1019 Rev: src/gc.c:1.290 Rev: src/gc.h:1.129 Rev: src/modules/system/system.c:1.182 Rev: src/pike_rusage.h:1.18 Rev: src/rusage.c:1.46 Rev: src/threads.c:1.254 Rev: src/threads.h:1.132
Temporarily made a failure in a test break the configure run altogether, to easier see how often that happens in xenofarm. Rev: src/configure.in:1.1018
Hopefully work around a compilation problem in stralloc.h in CONFIGURE_TEST that oddly affects only some platforms in debug mode. Rev: src/configure.in:1.1017
Fixed decay that broke the configure tests using rusage.c and threads.c. Also added configure tests to make it more certain that this kind of thing doesn't go by unnoticed in the future. Rev: src/configure.in:1.869 Rev: src/rusage.c:1.39 Rev: src/threads.c:1.242
Fixed decay that broke the configure tests using rusage.c and threads.c. Also added configure tests to make it more certain that this kind of thing doesn't go by unnoticed in the future. Rev: src/configure.in:1.1016 Rev: src/pike_threadlib.h:1.61 Rev: src/rusage.c:1.45 Rev: src/svalue.h:1.146 Rev: src/threads.c:1.253
Avoid the inevitable core dump during the initial stack size test - takes a lot of time and space to dump a 2 Gb core. Rev: src/configure.in:1.1015
Avoid the inevitable core dump during the initial stack size test - takes a lot of time and space to dump a 2 Gb core. Rev: src/configure.in:1.868
Restored cleanup on exit for valgrind. Use --without-cleanup-on-exit if you don't want it. Rev: src/configure.in:1.1014
Actually, it's not really 100% needed to have PIKE_CLEANUP when running with valgrind. Especially not when you are trying to not run valgrind on pike, but a large library loaded into pike (as a dynamic module). Rev: src/configure.in:1.1013
Fixed broken test for epoll(). Rev: src/configure.in:1.1012
Added check for signal(2). Rev: src/configure.in:1.1011
Fixed a missing quote in an old HP-UX check. Rev: src/configure.in:1.698
Fixed a missing quote in an old HP-UX check. Rev: src/configure.in:1.867
Fixed a missing quote in an old HP-UX check that for some mysterious reason only recently caused ./config.status --recheck to fail with ./configure.lineno: 1: Syntax error: Unterminated quoted string Might be related to autoconf 2.61 but I really don't know.. :P Rev: src/configure.in:1.1010
Minor style fix. Rev: src/configure.in:1.1009
Make sure the installed bundles are still found. Rev: src/configure.in:1.1008
Added --with-type-checker=new. Rev: src/configure.in:1.1007
Machine code backend for PPC64 added. Rev: src/Makefile.in:1.459 Rev: src/code/ppc64.c:1.1 Rev: src/code/ppc64.h:1.1 Rev: src/configure.in:1.1006 Rev: src/pikecode.c:1.11 Rev: src/pikecode.h:1.15 Rev: src/program.h:1.228
Backported epoll availability checks from Pike 7.7. Rev: src/backend.cmod:1.153 Rev: src/configure.in:1.866
Fixed typo. Rev: src/configure.in:1.1005
Now checks for epoll availability. Rev: src/configure.in:1.1004
Made IA32 machine code compatible with Darwin IA32 ABI. Enabled machine code for Darwin. Rev: src/Makefile.in:1.392 Rev: src/code/ia32.h:1.24 Rev: src/configure.in:1.865
Made IA32 machine code compatible with Darwin IA32 ABI. Enabled machine code for Darwin. Rev: src/code/ia32.h:1.31 Rev: src/configure.in:1.1003
Made IA32 machine code compatible with Darwin IA32 ABI. Enabled machine code for Darwin. Rev: src/code/README.txt:1.20 Rev: src/code/ia32.h:1.22 Rev: src/configure.in:1.697 Rev: src/interpret.c:1.301
Now OSFLAGS actually works... Rev: src/Makefile.in:1.456 Rev: src/configure.in:1.1002
Several PIKE_CHECK_ABI_DIR-related fixes. Rev: src/aclocal.m4:1.148 Rev: src/configure.in:1.1001
Backported some ABI code from Pike 7.7. Hopefully fixes some issues with mixed 32/64-bit library paths. Rev: src/configure.in:1.864
Some more gcc LDFLAGS-related fixes. Rev: src/configure.in:1.1000
Some bugfixes in the default library path setup. Rev: src/configure.in:1.999
More fixes... Rev: src/configure.in:1.998
Fixed quoting typo. Rev: src/configure.in:1.997
typo fixed Rev: src/configure.in:1.996
Attempt to extract the default lib path from gcc. Rev: src/configure.in:1.995
Fixes to make malloc.c jibe with the system malloc.h. Rev: src/configure.in:1.994 Rev: src/malloc.c:1.3
Enable PIKE_EXTRA_DEBUG for OpenBSD. Rev: src/configure.in:1.993
Removing Ssleay from standard Pike Rev: src/configure.in:1.992 Rev: src/modules/Ssleay/.cvsignore:1.7(DEAD) Rev: src/modules/Ssleay/Makefile.in:1.8(DEAD) Rev: src/modules/Ssleay/acconfig.h:1.5(DEAD) Rev: src/modules/Ssleay/configure.in:1.9(DEAD) Rev: src/modules/Ssleay/ssleay.c:1.23(DEAD) Rev: src/modules/Ssleay/testsuite.in:1.2(DEAD)
Make some effort to find the real page size. Rev: src/configure.in:1.991 Rev: src/pike_embed.c:1.9 Rev: src/pike_memory.c:1.182 Rev: src/pike_memory.h:1.54 Rev: src/program.c:1.607
Slightly nicer. Rev: src/configure.in:1.990
Report dlmalloc in the config info. Rev: src/configure.in:1.989
Backported import of Doug Lea's malloc implementation. It's enabled by default on Windows to fix bad fragmentation. Rev: COPYRIGHT:1.11 Rev: lib/modules/Tools.pmod/Legal.pmod/Copyright.pmod:1.5 Rev: src/acconfig.h:1.120 Rev: src/configure.in:1.696 Rev: src/dmalloc.h:1.44 Rev: src/malloc.c:1.1 Rev: src/pike_memory.c:1.147
Backported import of Doug Lea's malloc implementation. It's enabled by default on Windows to fix bad fragmentation. Rev: COPYRIGHT:1.15 Rev: lib/modules/Tools.pmod/Legal.pmod/Copyright.pmod:1.9 Rev: src/acconfig.h:1.144 Rev: src/configure.in:1.863 Rev: src/dmalloc.h:1.53 Rev: src/malloc.c:1.1 Rev: src/pike_memory.c:1.160
Added Doug Lea's malloc, use --enable-dlmalloc to enable it. It's enabled by default on Windows since the OS malloc sucks. Rev: src/acconfig.h:1.153 Rev: src/configure.in:1.988 Rev: src/dmalloc.h:1.60 Rev: src/malloc.c:1.1 Rev: src/pike_memory.c:1.180
Fixed bugs when dlls are enabled but dynamic modules aren't. Rev: src/configure.in:1.987
Enabled DLL modules by default. Rev: src/configure.in:1.986
Test winsock2.h before most other headers to avoid conflict with winsock.h. Rev: src/configure.in:1.985
Simplification. Rev: src/Makefile.in:1.450 Rev: src/configure.in:1.984
Link the icon into pike.exe. Rev: src/Makefile.in:1.448 Rev: src/configure.in:1.983
More accurate checking of windows tools, to avoid false positives. Rev: src/aclocal.m4:1.143 Rev: src/configure.in:1.982
Fixed debug/release crt selection. Added --with(out)-debug-crt. Rev: src/configure.in:1.981
Added support to bundle the dynamic Visual Studio crt in the install packages. Rev: src/Makefile.in:1.444 Rev: src/configure.in:1.980
Added fallback to link uncompress_header with a special static libz. Rev: src/Makefile.in:1.442 Rev: src/configure.in:1.979
Do the manifest embedding dance necessary to make the VC8 CRT happy. Rev: src/Makefile.in:1.441 Rev: src/configure.in:1.978 Rev: src/modules/dynamic_module_makefile.in:1.123
Gave the .pdb files used for debugging in MSVC unique names and fixed them to get copied correctly to the module lib directory. Rev: src/configure.in:1.977 Rev: src/modules/dynamic_module_makefile.in:1.121
Implemented a thorough separation of the different build orders necessary to build static modules (where a tpike is required) and dynamic modules (where a pike.lib that fits the pike core exe is required with --enable-dll). tpike is now built and used only when dynamic modules are disabled. Post modules can no longer affect the link options when the pike binary is built (something that wouldn't work reliably anyway). Rev: src/Makefile.in:1.438 Rev: src/configure.in:1.976 Rev: src/make_variables.in:1.26 Rev: src/modules/common_module_makefile.in:1.44 Rev: src/modules/dynamic_module_makefile.in:1.120 Rev: src/modules/static_module_makefile.in:1.104 Rev: src/post_modules/configure.in:1.7 Rev: src/tmodule.c:1.2
Reinstated some tpike stuff to use pike instead of tpike for various post-build targets and to avoid building tpike in some cases where a fresh external pike can be assumed. Rev: src/Makefile.in:1.437 Rev: src/configure.in:1.975
Try -mtune before -mcpu to avoid warnings from gcc 4.0. Rev: src/configure.in:1.974
Search for libraries when building with MinGW Rev: src/configure.in:1.973
Added check for PRINTOFFT. Rev: src/acconfig.h:1.151 Rev: src/configure.in:1.972
Fix. Rev: src/configure.in:1.971
Fixes for dll linking. Rev: src/Makefile.in:1.435 Rev: src/configure.in:1.970
Added facet configure option Rev: src/acconfig.h:1.150 Rev: src/configure.in:1.969
_getdrive is defined in direct.h Rev: src/configure.in:1.968 Rev: src/fdlib.c:1.81
Use thread safe variant of gmtime(3) if available. Rev: src/builtin_functions.c:1.618 Rev: src/configure.in:1.967
It's localtime_s, not _localtime_s. Rev: src/configure.in:1.966 Rev: src/fdlib.c:1.80
Corrected header tests for new AC_CHECK_FUNCS method. (It's e.g. necessary to test winsock2.h before winsock.h since they conflict and winsock2.h must be included before winsock.h to disable it properly.) Rev: src/configure.in:1.965 Rev: src/modules/HTTPLoop/configure.in:1.8 Rev: src/modules/Mysql/configure.in:1.49 Rev: src/modules/_math/configure.in:1.14 Rev: src/modules/system/configure.in:1.76
Enabled new AC_CHECK_FUNCS method for windows. Rev: src/configure.in:1.964
pike_cv_sys_os might contain spaces. Rev: src/configure.in:1.963
Added --enable-dll and basic support for linking dlls (a fresh rntcl is required if enabled). Rev: src/acconfig.h:1.149 Rev: src/configure.in:1.962 Rev: src/make_variables.in:1.25 Rev: src/modules/dynamic_module_makefile.in:1.118
Touch stamp-h only if machine.h is created. Rev: src/configure.in:1.961
Cleanup. Rev: src/configure.in:1.960
Fixed typo that caused the configure script to search for lots of nonapplicable include dirs on windows. Rev: src/configure.in:1.862
Fixed typo that caused the configure script to search for lots of nonapplicable include dirs on windows. Rev: src/configure.in:1.959
Have to propagate the base LIBS to the LDFLAGS used in the modules, to keep compatibility with module makefiles. Rev: src/aclocal.m4:1.135 Rev: src/configure.in:1.958 Rev: src/make_variables.in:1.24 Rev: src/modules/static_module_makefile.in:1.102
Propagate LIBS. Rev: src/aclocal.m4:1.133 Rev: src/configure.in:1.957 Rev: src/make_variables.in:1.23
Did away with the use of the undocumented internal MS CRT functions __loctotime_t and __dtoxmode. Rev: src/configure.in:1.956 Rev: src/fdlib.c:1.75
Did away with the use of the undocumented internal MS CRT functions __loctotime_t and __dtoxmode. Rev: src/configure.in:1.695 Rev: src/fdlib.c:1.62
Did away with the use of the undocumented internal MS CRT functions __loctotime_t and __dtoxmode. Rev: src/configure.in:1.861 Rev: src/fdlib.c:1.67
Cleaned up a little bit of dead code. Rev: src/configure.in:1.955
Cleaned up propagation of CFLAGS, CPPFLAGS and LDFLAGS. Rev: src/aclocal.m4:1.121 Rev: src/configure.in:1.954 Rev: src/make_variables.in:1.20 Rev: src/modules/common_module_makefile.in:1.36
Cleaned up propagation of BUILDDIR/TMP_BUILDDIR and BINDIR/TMP_BINDIR. Rev: src/aclocal.m4:1.120 Rev: src/configure.in:1.953 Rev: src/make_variables.in:1.19 Rev: src/modules/Makefile.in:1.53 Rev: src/modules/common_module_makefile.in:1.35 Rev: src/post_modules/Makefile.in:1.18
Do not restart the configure with bash if already running bash. This avoids that the --no-recursion flag from config.status gets lost since it's not part of $ac_configure_args. Rev: src/configure.in:1.952
Added detection and workaround for broken icc <=> gcc compatibility. Rev: src/configure.in:1.951
Fix for Visual Studio 8. Rev: src/configure.in:1.950 Rev: src/fdlib.c:1.74
Fix for Visual Studio 8. Rev: src/configure.in:1.860 Rev: src/fdlib.c:1.66
Fix for Visual Studio 8. Rev: src/configure.in:1.694 Rev: src/fdlib.c:1.61
Added checks for closefrom() and fdwalk(). Rev: src/configure.in:1.949
Always create the bundles directory. Rev: src/configure.in:1.948
Fixed m4-quoting... Rev: src/configure.in:1.947
Added workaround for Darwin/x86 assembler alignment limit. Rev: src/configure.in:1.946
Support Darwin x86. Rev: src/configure.in:1.859 Rev: src/modules/Image/configure.in:1.29
Support Darwin x86. Rev: src/configure.in:1.693 Rev: src/modules/Image/configure.in:1.26
Support Darwin x86. Rev: src/configure.in:1.945 Rev: src/modules/Image/configure.in:1.31
Disable -parallel for icc. Rev: src/configure.in:1.944
Darwin x86 doesn't support assembler yet. Rev: src/configure.in:1.858
Use just -ip instead of -ipo + -ipo_obj with icc. Rev: src/configure.in:1.943
Enable gradual underflow with cc/Solaris/sparc. Rev: src/configure.in:1.942
We want to use the most recent bundle file if there are multiple. Rev: src/configure.in:1.941
Fixed overcaching problem with computed goto switch. Rev: src/configure.in:1.940
Use pike_cv_abi_suffixes. Rev: src/configure.in:1.939
Backported some bundle fixes from Pike 7.7. Rev: src/configure.in:1.857
Backported configure-bugfix Rev: src/configure.in:1.856
Removed redundant AC_MSG_CHECKING. Rev: src/configure.in:1.938
Removed redundant command. Rev: src/configure.in:1.937
Now uses PIKE_CHECK_ABI_LIB_DIR. Rev: src/configure.in:1.936
Introduce PIKE_MSG_NOTE for messages that aren't really warnings, but should be prominently noted. Converted some warning offenders to PIKE_MSG_NOTE. Rev: src/aclocal.m4:1.108 Rev: src/configure.in:1.935
Clearified documentation for --with-site-prefixes. Rev: src/configure.in:1.934
Extracted ABI selection stuff to separate macros. Rev: src/aclocal.m4:1.104 Rev: src/configure.in:1.933
pike.so -> libpike.so Rev: Makefile:1.155 Rev: src/Makefile.in:1.413 Rev: src/configure.in:1.932
Backported MAP_ANON detection for Mac OS X from 7.7. Rev: src/configure.in:1.855
Backported MAP_ANON detection for Mac OS X from 7.7. Rev: src/configure.in:1.692
Fix suggested by Adam Montague <amontague@siriushosting.com>. Thanks. Rev: src/configure.in:1.931
Backported sys/param.h fix for OpenBSD from 7.7. Rev: src/configure.in:1.854 Rev: src/signal_handler.c:1.301
<sys/param.h> is needed for <sys/user.h> on OpenBSD. Rev: src/configure.in:1.930 Rev: src/signal_handler.c:1.315
kFreeBSD build detection fix Rev: src/configure.in:1.929
Updated kFreeBSD configure.in fix Rev: src/configure.in:1.853
kFreeBSD detection fix Rev: src/configure.in:1.852
Changes to be able to compile Pike 7.4 on kFreeBSD Rev: src/configure.in:1.691
Fixed typo. Rev: src/configure.in:1.928
Some OSes have an rm in interactive mode when running as root. This may cause a perceived hang of the configure script. Rev: src/configure.in:1.927
Added configure test for SetFilePointerEx(). Rev: src/configure.in:1.926 Rev: src/fdlib.c:1.73
Improved WIN32 support. Rev: src/configure.in:1.925
Automaticly include an icon in the Windows executable. This is a bit fragile as it will break if no 'rc' is available on the Windows build machine, but it shouldn'r be a problem in the real world. Rev: src/Makefile.in:1.389 Rev: src/configure.in:1.851
Oops, /usr/ccs/bin got lost in the bin-directory search. Rev: src/configure.in:1.924
Disable dependency tracking when configuring the bundled Nettle. This may fix compilation problems with BSD make. Rev: src/configure.in:1.923
Disable Nettle's openssl benchmarks. They do currently (nettle-1.13) not test libcrypt strictly enough. Rev: src/configure.in:1.922
Moved AC_SYS_COMPILER_FLAG et al to aclocal.m4. Rev: src/aclocal.m4:1.103 Rev: src/configure.in:1.921
Now cleans up y.tab.* after the YACC test. Rev: src/configure.in:1.920
Added timestamp adjustment to the bundle patching system. Rev: bundles/patches/nettle-1.13/.touch-list:1.1 Rev: src/configure.in:1.919
Added configure test to make the right short define for different bison versions. Rev: src/acconfig.h:1.119 Rev: src/configure.in:1.690 Rev: src/language.yacc:1.318
Added configure test to make the right short define for different bison versions. Rev: src/acconfig.h:1.148 Rev: src/configure.in:1.918 Rev: src/language.yacc:1.359
Added configure test to make the right short define for different bison versions. Rev: src/acconfig.h:1.143 Rev: src/configure.in:1.850 Rev: src/language.yacc:1.339
Improved ABI detection. Now uses --build in preference to --host when invoking configure for bundles. Rev: src/configure.in:1.917
Disable use of assembler for bundles when compiling with rntcl et al. Rev: src/configure.in:1.916
Improved ABI diagnostics. Rev: src/configure.in:1.915
Improved ABI detection. Rev: src/configure.in:1.914
Backported sigsetjmp() from 7.7 in order to avoid unwanted syscalls on OS X. Rev: src/configure.in:1.849 Rev: src/error.c:1.124 Rev: src/pike_error.h:1.33
Backported sigsetjmp() from 7.7 in order to avoid unwanted syscalls on OS X. Rev: src/configure.in:1.689 Rev: src/error.c:1.102 Rev: src/pike_error.h:1.26
Fix for newer Autoconf versions that broke memcmp optimization. Rev: src/configure.in:1.913
Fix for newer Autoconf versions that broke memcmp optimization. Rev: src/configure.in:1.848
Fix for newer Autoconf versions that broke memcmp optimization. Rev: src/configure.in:1.688
Fixed gcc 4.x compatibility on Mac OS X. Rev: src/configure.in:1.847
Fixed gcc 4.x compatibility on Mac OS X. Rev: src/configure.in:1.687
Fixed gcc 4.x compatibility on Mac OS X. Rev: src/configure.in:1.912
Remove major bottleneck on Mac OS X systems. This cuts Pike startup time almost in half and shaves many seconds off Roxen CMS startup on a G5. Rev: src/code/ppc32.c:1.36 Rev: src/configure.in:1.686
Remove major bottleneck on Mac OS X systems. This cuts Pike startup time almost in half and shaves many seconds off Roxen CMS startup on a G5. Rev: src/code/ppc32.c:1.38 Rev: src/configure.in:1.846
Remove major bottleneck on Mac OS X systems. This cuts Pike startup time almost in half and shaves 5+ seconds off Roxen CMS startup on a G5. Rev: src/code/ppc32.c:1.38 Rev: src/configure.in:1.911
Added extra ranlib pass to the bundle installer. Added multi install detection to the bundle installer. Rev: src/configure.in:1.910
Minor cleanup. Rev: src/configure.in:1.909
Added detection of compiler default ABI. Rev: src/configure.in:1.908
Improved diagnostics. Rev: src/configure.in:1.907
Oops, fixed typo when patching bundles. Rev: src/configure.in:1.906
Use a directory as patch source instead of *.ptar.gz. This makes the patches easier to verify. Rev: src/configure.in:1.905
Minor fix. Rev: src/configure.in:1.904
Potential fix for AIX 5L/ia64. Rev: src/configure.in:1.903
Fixed typo. Rev: src/configure.in:1.902
Bugfix... Rev: src/configure.in:1.901
Added some 64-bit ABI support for Sun Studio. Rev: src/configure.in:1.900
Bundle reconfiguration now utilizes RECONFIGURE_BUNDLES mode. This should speed up bundle reconfiguration considerably. Disabled #line warning on AIX 5L/ia64. Rev: src/configure.in:1.899
Joined lib and include directory checks to one pass. Now doesn't add include directories that correspond to libraries with the wrong abi. Rev: src/configure.in:1.898
Moved ABI options to CC, since having them in LDFLAGS is not a good idea if ld is used to link the binary. Rev: src/configure.in:1.897
Now sets ABI flags in both CPPFLAGS and LDFLAGS. Might fix ABI64 problems with Nettle/sparcv9. Rev: src/configure.in:1.896
Moved the compiler ABI flag from CFLAGS to CPPFLAGS. Rev: src/configure.in:1.895
OpenBSD now uses the same LDSHARED and LINKFORSHARED options as NetBSD. Fixes [bug 3941 (#3941)]. Thanks to Adam Montague <amontague@siriushosting.com>. Rev: src/configure.in:1.845
OpenBSD now uses the same LDSHARED and LINKFORSHARED options as NetBSD. Fixes [bug 3941 (#3941)]. Thanks to Adam Montague <amontague@siriushosting.com>. Rev: src/configure.in:1.894
Added detection of symlinks to /usr/include in include directory check. Probably fixes problems with icc on Gentoo/ia64 (/usr/X11R6 is symlink to /usr).. Rev: src/configure.in:1.893
Some more minor ABI-related fixes. Rev: src/configure.in:1.892
Fixed a few minor bugs. Rev: src/configure.in:1.891
Added some extra scanning for abi-specific libraries. Rev: src/configure.in:1.890
Added some paranoia to the ABI selector. Rev: src/configure.in:1.889
Added ABI check to library directory scan. Rev: src/configure.in:1.888
Minor cleanup of AC_SYS_COMPILER_FLAG. Obsoleted DO_IF_CPU. Obsoleted execstack code, since it shouldn't be needed anymore. Rev: src/configure.in:1.887
Added --with-abi. Added possibility to disable checks for specific compiler options. Rev: src/configure.in:1.886
Made stack direction test more paranoid (needed for gcc 3.4.4/ia64). Rev: src/configure.in:1.844
Made stack direction test more paranoid (needed for gcc 3.4.4/ia64). Rev: src/configure.in:1.685
Made stack direction test more paranoid (needed for gcc 3.4.4/ia64). Rev: src/configure.in:1.885
Fixed typo in variable name. Rev: src/configure.in:1.843
Fixed typo in variable name. Rev: src/configure.in:1.684
Fixed typo in variable name. Rev: src/configure.in:1.884
Improved mmap PROT_EXEC test. Rev: src/configure.in:1.883
Added detection of broken mmap on MacOS X. Rev: src/acconfig.h:1.118 Rev: src/configure.in:1.683 Rev: src/pike_memory.c:1.141
Added detection of broken mmap on MacOS X. Rev: src/acconfig.h:1.142 Rev: src/configure.in:1.842 Rev: src/pike_memory.c:1.154
Added detection of broken mmap on MacOS X. Rev: src/acconfig.h:1.147 Rev: src/configure.in:1.882 Rev: src/pike_memory.c:1.163
Merge 'tags/ad_work_sync': Removed duplicate row.
Removed duplicate row. Rev: src/configure.in:1.881
Merge 'tags/ad_work_sync': Reduce warning count on AIX.
Reduce warning count on AIX. Rev: src/configure.in:1.880
strrchr, strtok and strstr isn't used in Pike. Rev: src/configure.in:1.879 Rev: src/port.c:1.83 Rev: src/port.h:1.61
Added detection and some support for Solaris 10/amd64. Rev: src/configure.in:1.878
Oops, fixed typo. Rev: src/configure.in:1.877
Now knows about /opt/csw/. Rev: src/configure.in:1.876
Prefer /usr/sfw to /opt/sfw when both exist (eg Solaris 10). Rev: src/configure.in:1.875
Added --with-extra-debug-checks Rev: src/acconfig.h:1.146 Rev: src/configure.in:1.874
Oops, forgot to m4-quote the brackets. Rev: src/configure.in:1.873
Potential fix for spaces in paths mounted via MinGW. Rev: src/configure.in:1.872
Oops, fixed typo. Rev: src/configure.in:1.871
More fixes. Rev: src/configure.in:1.870
More posix_to_native.sh fixes for the MinGW case. Rev: src/configure.in:1.869
Bugfix in posix_to_native.sh for MinGW. Rev: src/configure.in:1.868
Some MinGW-related fixes. Added creation of posix_to_native.sh. Rev: src/configure.in:1.867
Removed broken test. Rev: src/configure.in:1.866
Added configure test for dlinit() which doesn't exist on many platforms. Rev: src/configure.in:1.865 Rev: src/main.c:1.223 Rev: src/pike_embed.h:1.5
dlopen.o is now in DLOPEN_O. Rev: src/Makefile.in:1.397 Rev: src/configure.in:1.864
Fixed typo in previous commit. Added PIKE_LIB. Rev: src/configure.in:1.863
Changes to main function to allow embedding. Note that this is very much experimental. It should not break anything when compiling a normal binary, but no promises are made. Rev: src/acconfig.h:1.145 Rev: src/configure.in:1.862 Rev: src/main.c:1.217
More icc-related fixes. Rev: src/configure.in:1.861
Improved detection of supported icc options. Rev: src/configure.in:1.860
Oops, bugfix. Rev: src/configure.in:1.859
Backported improved overflow handling in f_mktime to avoid regression. Rev: src/builtin_functions.c:1.551 Rev: src/configure.in:1.841 Rev: src/pike_int_types.h:1.3
Workaround for Autoconf 2.5x. Thanks to Dan Nelson. Rev: src/configure.in:1.858
Yet another typo... Rev: src/configure.in:1.857
Oops, fixed typo. Rev: src/configure.in:1.856
Improved detection of and options for icc/ecc. Rev: src/configure.in:1.855
Added SIZEOF_TIME_T, MAX_TIME_T and MIN_TIME_T. Rev: src/configure.in:1.854 Rev: src/pike_int_types.h:1.3
Disabled some more icc warnings and remarks. Rev: src/configure.in:1.853
Backported icc warning fix from Pike 7.7. Rev: src/configure.in:1.840
Disable excessive warnings for icc/Linux. Rev: src/configure.in:1.852
More Autoconf 2.5x compat. Thanks to Mattias Wingstedt <wing@lysator.liu.se>. Rev: src/configure.in:1.851
Improved Autoconf 2.59 compat. Rev: src/configure.in:1.850
Backported wix support from Pike 7.6. Rev: bin/install.pike:1.114 Rev: lib/modules/Standards.pmod/UUID.pmod:1.2 Rev: lib/modules/Standards.pmod/XML.pmod/Wix.pmod:1.1 Rev: lib/modules/Tools.pmod/Standalone.pmod/make_wxs.pike:1.1 Rev: src/Makefile.in:1.342 Rev: src/Pike_ui.wxs:1.1 Rev: src/configure.in:1.682 Rev: src/make_banner.pike:1.1
Workaround for gcc/x86_64 and va_list *. Rev: src/acconfig.h:1.144 Rev: src/configure.in:1.849 Rev: src/stralloc.c:1.188
Backported Wix support from Pike 7.7. Rev: bin/install.pike:1.130 Rev: src/Makefile.in:1.384 Rev: src/Pike_ui.wxs:1.1 Rev: src/configure.in:1.839
Added some tallow. Rev: src/configure.in:1.848
Backported Autoconf 2.5x stupidity fix from Pike 7.7. Rev: src/configure.in:1.681
Backports from 7.7 by Grubba. Rev: src/aclocal.m4:1.62 Rev: src/configure.in:1.680
Machine code make FreeBSD compilers getting crazy on non x86 platforms. Disabling it per default Rev: src/configure.in:1.847
FreeBSD and Pike machine code make some compiler crazy on 5.x branch. Disabling per default. Rev: src/configure.in:1.838
Added --with-dmalloc-track-free to reduce the default dmalloc memory print somewhat. Rev: src/acconfig.h:1.143 Rev: src/configure.in:1.846 Rev: src/pike_memory.c:1.157
It's not a good idea to put single quotes in cache variables. Rev: src/configure.in:1.845
Now searches for candles and lights. Rev: src/configure.in:1.844
Note that --with-valgrind implies --with-cleanup-at-exit, and made that possible to override with --without-cleanup-at-exit. Rev: src/configure.in:1.843 Rev: src/global.h:1.105
Note that --with-valgrind implies --with-cleanup-at-exit, and made that possible to override with --without-cleanup-at-exit. Rev: src/configure.in:1.837 Rev: src/global.h:1.100
Provide a default value for pike_cv_initial_stack_size when cross-compiling. Rev: src/configure.in:1.836
Provide a default value for pike_cv_initial_stack_size when cross-compiling. Rev: src/configure.in:1.842
Include <devices/timer.h> first on AmigaOS to avoid timeval confusion. Rev: src/configure.in:1.835 Rev: src/global.h:1.99
Include <devices/timer.h> first on AmigaOS to avoid timeval confusion. Rev: src/configure.in:1.841 Rev: src/global.h:1.104
Don't check for initial stack limit on AmigaOS. Rev: src/configure.in:1.840
Don't check for initial stack limit on AmigaOS. Rev: src/configure.in:1.834
Avoid using $CFLAGS when testing for .uaword, so that we don't get -g. Rev: src/configure.in:1.679
Avoid using $CFLAGS when testing for .uaword, so that we don't get -g. Rev: src/configure.in:1.833
Avoid using $CFLAGS when testing for .uaword, so that we don't get -g. Rev: src/configure.in:1.839
Added support for more variants of setjmp() et al to avoid excessive system calls on BSDs. Thanks to Dan Nelson for the suggestion. Rev: src/configure.in:1.838 Rev: src/error.c:1.131 Rev: src/pike_error.h:1.34
Fixed a few typos in the last checkin. Rev: src/configure.in:1.832
Fixed a few typos in the last checkin. Rev: src/configure.in:1.837
Avoid default usage of machine code on OpenBSD. Seems to violate some checks on the compiler. Rev: src/configure.in:1.831
Avoid default usage of machine code on OpenBSD. Seems to violate some checks on the compiler. Rev: src/configure.in:1.836
Bugfix. Rev: src/configure.in:1.678
Backported mprotect workaround from Pike 7.7. Rev: src/configure.in:1.677
Backported mprotect workaround from Pike 7.7. Rev: src/configure.in:1.830
Added workaround for broken mprotect(PROT_EXEC) on Linux. Rev: src/configure.in:1.835
Added --without-system-locale. Rev: src/configure.in:1.676
PIKE_PORTABLE_BYTECODE is now on by default. Rev: src/configure.in:1.834
Added --with-portable-bytecode. Rev: src/configure.in:1.675
Preprocessor flags are supposed to be in CPPFLAGS. Fixed some comments. Rev: src/configure.in:1.833
Effort to make Pike compile on native win32 using Mingw32. This patch does still have some problems. For example, detection of the fpclass method under NT does not mean that the constants used are present. It also lacks the correct translation of paths. There is still lots to be done, but it is a start. Rev: src/configure.in:1.832 Rev: src/fdlib.h:1.52 Rev: src/global.h:1.97 Rev: src/main.c:1.201 Rev: src/modules/system/nt.c:1.64 Rev: src/port.c:1.77 Rev: src/port.h:1.56 Rev: src/smartlink.c:1.16
Added option to disable smartlink binary. This is useful for win32 native builds since smartlink binary does not work correctly under Mingw32. Rev: src/configure.in:1.831
Now searches /usr/sfw/* in addition to /opt/sfw/*. Rev: src/configure.in:1.830
Improved detection of icc and ecc. Rev: src/configure.in:1.829
Improved support for icc 7.0/7.1. Rev: src/configure.in:1.828
--without-new-multisets no longer builds and no one would really be interested in doing that anymore. Removing PIKE_NEW_MULTISETS Rev: src/acconfig.h:1.140 Rev: src/array.c:1.158 Rev: src/builtin.cmod:1.152 Rev: src/builtin_functions.c:1.537 Rev: src/configure.in:1.827 Rev: src/encode.c:1.211 Rev: src/gc.c:1.251 Rev: src/iterators.cmod:1.54 Rev: src/main.c:1.199 Rev: src/modules/Oracle/oracle.c:1.81 Rev: src/multiset.c:1.84 Rev: src/multiset.h:1.33 Rev: src/operators.c:1.188 Rev: src/svalue.c:1.192
Support for kqueue seems to work now. Rev: src/backend.cmod:1.116 Rev: src/configure.in:1.826
Now defaults to using /dev/epoll if it exists. --with-epoll is obsoleted. Use --without-devpoll instead. Solaris 7/x86 support for /dev/poll seems to work now. Rev: src/configure.in:1.825
I assume that if float.h is needed to find _isnan it is also needed to use it. Rev: src/configure.in:1.824 Rev: src/pike_float.h:1.4
Latest valgrind has no default tool. Rev: src/configure.in:1.823
Added some paranoia to the pike_cv_libgcc_filename test. Rev: src/configure.in:1.822
Improved bundle Mekefile patcher some more. Rev: src/configure.in:1.821
Since the cleanup code is fairly decent now I figured it could be interesting to have a configure option --with-cleanup-on-exit to enable it explicitly. Rev: src/acconfig.h:1.139 Rev: src/configure.in:1.820 Rev: src/global.h:1.96
Added support for --with-thread-library. Removed broken code for --with-named-thread-lib. Moved some stack depth checking code out from FreeBSD specific. Added some notes. Rev: src/configure.in:1.819
Some C99 compilers (eg icc7.0) do not fully support the restrict keyword. Rev: src/acconfig.h:1.138 Rev: src/configure.in:1.818
Improved bundle Makefile patcher. Rev: src/configure.in:1.817
Fixed egrep pattern used to detect bundle Makefiles that need patching. Improved diagnostics for bundle Makefile patching. Reverted find command change. Rev: src/configure.in:1.816
More paranoid use of find in bundle Makefile patcher. Rev: src/configure.in:1.815
Fixed parenthesis error in bundle Makefile patching rule. Rev: src/configure.in:1.814
Improved portability for the spaces in SHELL test. Rev: src/configure.in:1.813
Added check for /dev/poll and definition of PIKE_POLL_DEVICE. Rev: src/configure.in:1.812
Improved diagnostics. Rev: src/configure.in:1.811
More exhaustive make supports SHELL test. Now uses symbolic name AC_FD_CC for fd 5. Rev: src/configure.in:1.810
Added support for --without-devpoll. Rev: src/configure.in:1.809
Improved diagnostics. Rev: src/configure.in:1.808
Added some support for patching of bundles. Rev: src/configure.in:1.807
Attempt to force specification for thread library on freebsd. (used for free port). Rev: src/configure.in:1.806
Disable make.conf. Rev: src/configure.in:1.674
make.conf should not be used. Rev: src/configure.in:1.805
Support for OSF/1 /bin/sh in bundle system. Rev: src/configure.in:1.804
Use a /bin/true lookalike instead of a /bin/false on to simulate makeinfo. Rev: src/configure.in:1.803
Added hiding of makeinfo for the bundle build. Rev: src/configure.in:1.802
Fixed bashism. Rev: src/configure.in:1.801
Added patching with respect to SHELL of Makefiles in bundles. Rev: src/configure.in:1.800
Added check for support of SHELL by make. Rev: src/configure.in:1.799
Keep the extended $PATH for a bit longer... Rev: src/configure.in:1.798
Survive bash not being executable. Rev: src/configure.in:1.797
Fixed typo. Rev: src/configure.in:1.796
Some more diagnostics. Rev: src/configure.in:1.795
Fix for *.lib being installed by most install targets. Rev: src/configure.in:1.794
Improved bundle behaviour. Rev: src/configure.in:1.793
Bundles should now work again. Rev: src/configure.in:1.792
Fixed typo in variable name. Rev: src/configure.in:1.791
Restore CFLAGS, CPPFLAGS and LDFLAGS to their initial values on second configure pass. /usr/bin/ed is the "standard editor", but some systems don't have it anyway. Now uses sed instead. Added some more export directives. Rev: src/configure.in:1.790
Don't look for testsuites in bundles. Rev: src/configure.in:1.789
Some more kludges for helping bundles along. Rev: src/configure.in:1.788
The bundle system now helps stuff that uses config.guess with rntcl. Bundled libraries are now built with PIC code if possible. Improved diagnostics. Rev: src/configure.in:1.787
Now passes --build to the bundle configure in case of compiling with rntcl. This should help with stuff that uses config.guess. Rev: src/configure.in:1.786
Even more fixes. Rev: src/configure.in:1.785
Improved whitespace fix. Rev: src/configure.in:1.784
Fixed whitespace bug in bundle cache cleaner. Rev: src/configure.in:1.783
Even more bundle handling fixes. Rev: src/configure.in:1.782
Minor fix. Rev: src/configure.in:1.781
Fixed another typo. Rev: src/configure.in:1.780
Fixed typo. Rev: src/configure.in:1.779
Fixed bug in handling of --with-bundles. Rev: src/configure.in:1.778
Diagnostic change. Rev: src/configure.in:1.777
Automatic building of bundled libraries might now work. Rev: src/configure.in:1.776
Bugfix in bundle handling. Rev: src/configure.in:1.775
The bundle handling now performs some more cache cleaning. Rev: src/configure.in:1.774
Moved the default location for the bundle directory one level closer to the root. Rev: src/configure.in:1.773
Oops... Rev: src/configure.in:1.772
Bugfix. Rev: src/configure.in:1.771
Oops, fixed another typo. Rev: src/configure.in:1.770
Fixed quoting bug. Rev: src/configure.in:1.769
Some more infrastructure for handling bundled libraries. Not working yet. Rev: src/configure.in:1.768
Start bash with -norc to avoid surprises. Rev: src/configure.in:1.767
Added checks for <syscall.h> and <sys/syscall.h>. Rev: src/configure.in:1.766
Added --with-epoll. Rev: src/configure.in:1.765
Added check for <sys/epoll.h>. Rev: src/configure.in:1.764
Added support for Solaris-style /dev/poll. Rev: src/backend.cmod:1.76 Rev: src/configure.in:1.763
Potential fix for dynamic loading with icc/FreeBSD. Rev: src/configure.in:1.762
Introduce proper isordered test. Rev: src/acconfig.h:1.114 Rev: src/configure.in:1.673 Rev: src/svalue.c:1.160
Introduce proper isordered test. Rev: src/acconfig.h:1.134 Rev: src/configure.in:1.761 Rev: src/pike_float.h:1.3 Rev: src/svalue.c:1.183
changed the way int,float,pointer is reported in config.info (from compiled-and-run to more pretty), also added info if the float type is in ieee Rev: src/configure.in:1.760
Added support for --with-gcov. Rev: src/configure.in:1.759
Fixed typo in use of test on FreeBSD. Thanks to Dan Nelson <dan@emsphone.com>. Rev: src/configure.in:1.758
Don't use DEFAULT_RUNPIKE when cross-compiling. Rev: src/configure.in:1.757
Use prefix for *-config when cross-compiling. Rev: src/configure.in:1.756 Rev: src/modules/_Image_FreeType/configure.in:1.12 Rev: src/post_modules/GTK/configure.in:1.46 Rev: src/post_modules/SDL/configure.in:1.7 Rev: src/post_modules/_Image_SVG/configure.in:1.2
that fixes a "hang" on (at least) Debian/Sid. It's probably (ehm) an autoconf bug since AC_EGREP_CPP(yes, ...) invokes /usr/bin/yes now which, in turn, produces an endless stream of y<LF>. <rant>Hail EPITA!</rant> Rev: src/configure.in:1.755 Rev: src/modules/_Image_JPEG/configure.in:1.20
Backported Mac OS X fixes and performance improvements from 7.4/7.5. We now use pthread_wait_np() and use Mach kernel calls to greatly reduce thread scheduling overhead. Rev: src/acconfig.h:1.81 Rev: src/configure.in:1.502 Rev: src/threads.c:1.171 Rev: src/threads.h:1.120
Removed broken and redundant code. Fixes dynamic loading on MacOS X. Rev: src/configure.in:1.672
Removed broken and redundant code. Fixes dynamic loading on MacOS X. Rev: src/configure.in:1.754
Backported Mac OS X fixes and performance improvements from 7.5: We now use pthread_wait_np() and use Mach kernel calls to greatly reduce thread scheduling overhead. Included OSF/1 fixes by Grubba. Rev: src/acconfig.h:1.113 Rev: src/configure.in:1.671 Rev: src/pike_threadlib.h:1.28 Rev: src/threads.c:1.200
AmigaOS signal handler update (incomplete). Rev: src/configure.in:1.753 Rev: src/signal_handler.c:1.288
Detect target OS when cross-compiling. Rev: src/configure.in:1.752
Backport for head -1 -> head -n -1 Rev: src/aclocal.m4:1.60 Rev: src/configure.in:1.670 Rev: src/precompile.sh.in:1.34 Rev: src/precompile2.sh:1.2 Rev: src/run_autoconfig:1.35
m4_pattern_allow() doesn't exist in autoconf 2.13. Rev: src/configure.in:1.751
Fedora support (added -Wa,--execstack check, added another new-autoconf-workaround). Removed warning for 'using --with-poll', since it's not really anything to warn about. Rev: src/configure.in:1.750
Really new head don't understand -1. Rev: src/aclocal.m4:1.84 Rev: src/configure.in:1.749 Rev: src/precompile.sh.in:1.37 Rev: src/precompile2.sh:1.2 Rev: src/run_autoconfig:1.38
Use pthread_yield_np() on OS X. Added checks for sys/ioctl.h to fix missing prototype. Added checks for various Mach kernel headers which improve thread performance on OS X. Rev: src/configure.in:1.748