2011-01-03 20:06:28 +00:00
Name : pypy
2012-06-09 00:18:15 +00:00
Version : 1.9
2012-07-10 20:45:52 +00:00
Release : 3%{?dist}
2011-01-03 20:06:28 +00:00
Summary : Python implementation with a Just-In-Time compiler
Group : Development/Languages
# LGPL and another free license we'd need to ask spot about are present in some
# java jars that we're not building with atm (in fact, we're deleting them
# before building). If we restore those we'll have to work out the new
# licensing terms
License : MIT and Python and UCD
URL : http://pypy.org/
BuildRoot : %{_tmppath} /%{name} -%{version} -%{release} -root-%(%{__id_u} -n)
# High-level configuration of the build:
# PyPy consists of an implementation of an interpreter (with JIT compilation)
# for the full Python language written in a high-level language, leaving many
# of the implementation details as "pluggable" policies.
#
# The implementation language is then compiled down to .c code, from which we
# obtain a binary.
#
# This allows us to build a near-arbitrary collection of different
# implementations of Python with differing tradeoffs
#
# (As it happens, the implementation language is itself Python, albeit a
# restricted subset "RPython", chosen to making it amenable to being compiled.
# The result implements the full Python language though)
# We could build many different implementations of Python.
# For now, let's focus on the implementation that appears to be receiving the
# most attention upstream: the JIT-enabled build, with all standard
# optimizations
# Building a configuration can take significant time:
# A build of pypy (with jit) on i686 took 77 mins:
# [Timer] Timings:
# [Timer] annotate --- 583.3 s
# [Timer] rtype_lltype --- 760.9 s
# [Timer] pyjitpl_lltype --- 567.3 s
# [Timer] backendopt_lltype --- 375.6 s
# [Timer] stackcheckinsertion_lltype --- 54.1 s
# [Timer] database_c --- 852.2 s
# [Timer] source_c --- 1007.3 s
# [Timer] compile_c --- 419.9 s
# [Timer] ===========================================
# [Timer] Total: --- 4620.5 s
#
# A build of pypy (nojit) on x86_64 took about an hour:
# [Timer] Timings:
# [Timer] annotate --- 537.5 s
# [Timer] rtype_lltype --- 667.3 s
# [Timer] backendopt_lltype --- 385.4 s
# [Timer] stackcheckinsertion_lltype --- 42.5 s
# [Timer] database_c --- 625.3 s
# [Timer] source_c --- 1040.2 s
# [Timer] compile_c --- 273.9 s
# [Timer] ===========================================
# [Timer] Total: --- 3572.0 s
#
#
# A build of pypy-stackless on i686 took about 87 mins:
# [Timer] Timings:
# [Timer] annotate --- 584.2 s
# [Timer] rtype_lltype --- 777.3 s
# [Timer] backendopt_lltype --- 365.9 s
# [Timer] stackcheckinsertion_lltype --- 39.3 s
# [Timer] database_c --- 1089.6 s
# [Timer] source_c --- 1868.6 s
# [Timer] compile_c --- 490.4 s
# [Timer] ===========================================
# [Timer] Total: --- 5215.3 s
2011-04-30 22:24:07 +00:00
# We will build a "pypy" binary.
#
# Unfortunately, the JIT support is only available on some architectures.
#
2011-01-03 20:06:28 +00:00
# pypy-1.4/pypy/jit/backend/detect_cpu.py:getcpuclassname currently supports the
# following options:
# 'i386', 'x86'
# 'x86-without-sse2':
# 'x86_64'
# 'cli'
# 'llvm'
2011-04-30 22:24:07 +00:00
#
# We will only build with JIT support on those architectures, and build without
# it on the other archs. The resulting binary will typically be slower than
# CPython for the latter case.
#
2011-01-03 20:06:28 +00:00
%ifarch %{ix86} x86_64
# FIXME: is there a better way of expressing "intel" here?
%global with_jit 1
%else
%global with_jit 0
%endif
# Should we build a "pypy-stackless" binary?
%global with_stackless 0
2011-09-13 09:00:09 +00:00
# Should we build the emacs JIT-viewing mode?
%if 0%{?rhel} == 5 || 0%{?rhel} == 6
%global with_emacs 0
%else
%global with_emacs 1
%endif
2011-01-03 20:06:28 +00:00
# Easy way to enable/disable verbose logging:
%global verbose_logs 0
2011-12-16 22:15:41 +00:00
# Forcibly use the shadow-stack option for detecting GC roots, rather than
# relying on hacking up generated assembler with regexps:
%global shadow_stack 1
2011-08-19 23:04:38 +00:00
# Easy way to turn off the selftests:
%global run_selftests 1
2011-01-03 20:06:28 +00:00
%global pypyprefix %{_libdir} /pypy-%{version}
Rebase to 1.5 (rhbz#701121)
Update patches:
patch 0: pypy-1.4-config.patch -> pypy-1.5-config.patch
patch 4: pypy-1.4.1-more-readable-c-code.patch -> pypy-1.5-more-readable-c-code.patch
Remove references to *.inl files, no longer present
Add the following tests to the skip list:
test_audioop, test_capi, test_distutils, test_gc, test_gdb, test_generators,
test_getargs2, test_hotshot, test_io, test_multiprocessing, test_posix,
test_readline, test_scope, test_strop, test_structmembers, test_subprocess,
test_symtable, test_sys_settrace, test_tempfile, test_thread, test_uuid,
test_zipimport_support
Add a couple of text files to the payload (TODO, stdlib-version.txt)
2011-05-02 18:46:32 +00:00
%global pylibver 2.7
2011-01-03 20:06:28 +00:00
# We refer to this subdir of the source tree in a few places during the build:
%global goal_dir pypy/translator/goal
# Turn off the brp-python-bytecompile postprocessing script
# We manually invoke it later on, using the freshly built pypy binary
%global __os_install_post \
%(echo '%{__os_install_post}' | sed -e 's!/usr/lib[^[:space:]]*/brp-python-bytecompile[[:space:]].*$!!g' )
# Source and patches:
2011-11-22 16:30:49 +00:00
Source0 : https://bitbucket.org/pypy/pypy/get/release-%{version} .tar.bz2
2011-01-03 20:06:28 +00:00
2011-08-25 15:58:55 +00:00
# Supply various useful RPM macros for building python modules against pypy:
# __pypy, pypy_sitelib, pypy_sitearch
Source2 : macros.pypy
2011-01-03 20:06:28 +00:00
# Edit a translator file for linux in order to configure our cflags and dynamic libffi
2011-11-22 16:30:49 +00:00
Patch0 : config.patch
2011-01-03 20:06:28 +00:00
# By default, if built at a tty, the translation process renders a Mandelbrot
# set to indicate progress.
# This obscures useful messages, and may waste CPU cycles, so suppress it, and
# merely render dots:
Patch1 : pypy-1.2-suppress-mandelbrot-set-during-tty-build.patch
2011-01-06 22:09:03 +00:00
# Try to improve the readability of the generated .c code, by adding in the
# RPython source as comments where possible.
# A version of this was sent upstream as:
# http://codespeak.net/pipermail/pypy-dev/2010q4/006532.html
# TODO: get this into the upstream bug tracker, and finish inlining
# support (rhbz#666963)
2011-11-22 16:30:49 +00:00
Patch4 : more-readable-c-code.patch
2011-01-06 22:09:03 +00:00
2011-08-20 15:59:48 +00:00
# In my koji builds, /root/bin is in the PATH for some reason
# This leads to test_subprocess.py failing, due to "test_leaking_fds_on_error"
# trying every dir in PATH for "nonexisting_i_hope", which leads to it raising
# OSError: [Errno 13] Permission denied
# when it tries to read /root/bin, rather than raising "No such file"
#
# Work around this by specifying an absolute path for the non-existant
# executable
# Not yet sent upstream
Patch5 : pypy-1.6-fix-test-subprocess-with-nonreadable-path-dir.patch
2011-01-06 22:09:03 +00:00
2012-07-10 20:45:52 +00:00
# Patch pypy.translator.platform so that stdout from "make" etc gets logged,
# rather than just stderr, so that the command-line invocations of the compiler
# and linker are captured:
Patch6 : 006-always-log-stdout.patch
# Disable the printing of a quote from IRC on startup (these are stored in
# ROT13 form in lib_pypy/_pypy_irc_topic.py). Some are cute, but some could
# cause confusion for end-users (and many are in-jokes within the PyPy
# community that won't make sense outside of it). [Sorry to be a killjoy]
Patch7 : 007-remove-startup-message.patch
# With pypy-1.9-1.fc17.x86_64, the pypy binary exposes about 200k symbols to
# the dynamic linker:
# $ eu-readelf -s $(which pypy) | head
# Symbol table [ 5] '.dynsym' contains 194163 entries:
# which is far more than necessary.
# Fix the version script for the linker as invoked thus in the Makefile:
# "-Wl,--export-dynamic,--version-script=../dynamic-symbols-6"
# so that it contains a "local: *;" clause, thus hiding the bulk of the
# symbols from the dynamic linker.
# Ideally we'd add:
# __attribute__ ((visibility ("hidden")))
# to most symbols, allowing the compiler to potentially generate better code.
# Not yet reported upstream
Patch8 : 008-fix-dynamic-symbols-script.patch
# Cherrypick upstream patch to add PyInt_AsUnsignedLongLongMask (used by
# the rpm python bindings); see https://bugs.pypy.org/issue1211
# This is https://bitbucket.org/pypy/pypy/changeset/542d481517d3
Patch9 : 009-add-PyInt_AsUnsignedLongLongMask.patch
2011-01-03 20:06:28 +00:00
# Build-time requirements:
2011-01-06 22:09:03 +00:00
# pypy's can be rebuilt using itself, rather than with CPython; doing so
# halves the build time.
#
# Turn it off with this boolean, to revert back to rebuilding using CPython
# and avoid a cycle in the build-time dependency graph:
2011-01-14 16:58:45 +00:00
# I'm disabling the self-hosting for now, due to a fatal error seen inside the
# JIT, presumably whilst JIT-compiling something within the translator's
# inliner.
#
# Specifically, building pypy-1.4.1-7.fc15.src.rpm on x86_64 using pypy-1.4.1-5.fc15.x86_64
# http://koji.fedoraproject.org/koji/taskinfo?taskID=2721517
# failed with this RPython traceback:
# ... snip ...
# [rtyper:WARNING] prebuilt instance <pypy.rpython.memory.gctransform.asmgcroot.ShapeDecompressor instance at 0x00000000f0b5bc80> has no attribute 'addr'
# [rtyper] specializing: 179300 / 180508 blocks (99%)
# [rtyper] specializing: 180500 / 180566 blocks (99%)
# [rtyper] -=- specialized 1363 more blocks -=-
# [rtyper] specializing: 180600 / 180777 blocks (99%)
# [rtyper] -=- specialized 211 more blocks -=-
# [backendopt:inlining] phase with threshold factor: 32.4
# [backendopt:inlining] heuristic: pypy.translator.backendopt.inline.inlining_heuristic
# [x86/regalloc] Bogus arg in operation 76 at 0
# RPython traceback:
# File "implement_62.c", line 39979, in send_bridge_to_backend
# File "implement_69.c", line 65301, in Assembler386_assemble_bridge
# File "implement_72.c", line 8078, in RegAlloc_prepare_bridge
# File "implement_40.c", line 53061, in RegAlloc__prepare
# File "implement_44.c", line 14305, in RegAlloc__compute_vars_longevity
# Fatal RPython error: NotImplementedError
2011-01-06 22:09:03 +00:00
#
2011-01-14 16:58:45 +00:00
# This appears to be deep within pypy/jit/backend/x86/regalloc.py which has
# called "not_implemented" to emit this message to stderr, before raising the
# exception:
# [x86/regalloc] Bogus arg in operation 76 at 0
%global use_self_when_building 0
2011-01-06 22:09:03 +00:00
%if 0%{use_self_when_building}
BuildRequires : pypy
%global bootstrap_python_interp pypy
%else
2011-09-13 01:01:16 +00:00
# Python 2.6 or later is needed, so on RHEL5 (2.4) we need to use the alternate
# python26 rpm:
%if 0%{?rhel} == 5
BuildRequires : python26-devel
%global bootstrap_python_interp python26
%else
2011-01-06 22:09:03 +00:00
BuildRequires : python-devel
%global bootstrap_python_interp python
%endif
2011-09-13 01:01:16 +00:00
%endif
2011-01-06 22:09:03 +00:00
2011-01-03 20:06:28 +00:00
# FIXME: I'm seeing errors like this in the logs:
# [translation:WARNING] The module '_rawffi' is disabled
# [translation:WARNING] because importing pypy.rlib.libffi raised ImportError
# [translation:WARNING] 'libffi.a' not found in ['/usr/lib/libffi', '/usr/lib']
# Presumably we need to fix things to support dynamically-linked libffi
BuildRequires : libffi-devel
BuildRequires : zlib-devel
BuildRequires : bzip2-devel
BuildRequires : ncurses-devel
BuildRequires : expat-devel
BuildRequires : openssl-devel
2011-01-07 16:56:06 +00:00
%ifarch %{ix86} x86_64 ppc ppc64 s390x
2011-01-03 20:06:28 +00:00
BuildRequires : valgrind-devel
2011-01-07 16:56:06 +00:00
%endif
2011-01-03 20:06:28 +00:00
2011-08-19 23:04:38 +00:00
%if %{run_selftests}
2011-01-03 20:06:28 +00:00
# Used by the selftests, though not by the build:
BuildRequires : gc-devel
2011-08-19 23:04:38 +00:00
# For use in the selftests, for recording stats:
BuildRequires : time
# For use in the selftests, for imposing a per-test timeout:
BuildRequires : perl
%endif
2011-01-03 20:06:28 +00:00
BuildRequires : /usr/bin/execstack
2011-08-03 17:02:57 +00:00
# For byte-compiling the JIT-viewing mode:
2011-09-13 09:00:09 +00:00
%if %{with_emacs}
2011-08-03 17:02:57 +00:00
BuildRequires : emacs
2011-09-13 09:00:09 +00:00
%endif
2011-08-03 17:02:57 +00:00
2011-01-03 20:06:28 +00:00
# pypy is bundling these so we delete them in %%prep. I don't think they are
# needed unless we build pypy targetted at running on the jvm.
#BuildRequires: jna
#BuildRequires: jasmin # Not yet in Fedora
# Metadata for the core package (the JIT build):
Requires : pypy-libs = %{version} -%{release}
%description
2011-04-30 22:24:07 +00:00
PyPy's implementation of Python, featuring a Just-In-Time compiler on some CPU
architectures, and various optimized implementations of the standard types
(strings, dictionaries, etc)
%if 0%{with_jit}
This build of PyPy has JIT-compilation enabled.
%else
This build of PyPy has JIT-compilation disabled, as it is not supported on this
CPU architecture.
%endif
2011-01-03 20:06:28 +00:00
%package libs
Group : Development/Languages
Summary : Run-time libraries used by PyPy implementations of Python
2011-08-03 17:02:57 +00:00
# We supply an emacs mode for the JIT viewer.
# (This doesn't bring in all of emacs, just the directory structure)
2011-09-13 09:00:09 +00:00
%if %{with_emacs}
2011-08-03 17:02:57 +00:00
Requires : emacs-filesystem >= %{_emacs_version}
2011-09-13 09:00:09 +00:00
%endif
2011-08-03 17:02:57 +00:00
2011-01-03 20:06:28 +00:00
%description libs
Libraries required by the various PyPy implementations of Python.
2011-01-14 00:51:26 +00:00
%package devel
Group : Development/Languages
Summary : Development tools for working with PyPy
%description devel
Header files for building C extension modules against PyPy
Requires : pypy = %{version} -%{release}
2011-01-03 20:06:28 +00:00
%if 0%{with_stackless}
%package stackless
Group : Development/Languages
Summary : Stackless Python interpreter built using PyPy
Requires : pypy-libs = %{version} -%{release}
%description stackless
Build of PyPy with support for micro-threads for massive concurrency
%endif
%if 0%{with_stackless}
%package stackless
Group : Development/Languages
Summary : Stackless Python interpreter built using PyPy
Requires : pypy-libs = %{version} -%{release}
%description stackless
Build of PyPy with support for micro-threads for massive concurrency
%endif
%prep
2012-06-09 00:18:15 +00:00
%setup -q -n pypy-pypy-341e1e3821ff
2011-01-03 20:06:28 +00:00
%patch0 -p1 -b .configure-fedora
%patch1 -p1 -b .suppress-mandelbrot-set-during-tty-build
2012-02-10 16:12:43 +00:00
# Disabled for now, as it needs regenerating for 1.8
#patch4 -p1 -b .more-readable-c-code
# Fails on 1.8 with this error:
# [translation:ERROR] Error:
# [translation:ERROR] Traceback (most recent call last):
# [translation:ERROR] File "translate.py", line 309, in main
# [translation:ERROR] drv.proceed(goals)
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/driver.py", line 811, in proceed
# [translation:ERROR] return self._execute(goals, task_skip = self._maybe_skip())
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/tool/taskengine.py", line 116, in _execute
# [translation:ERROR] res = self._do(goal, taskcallable, *args, **kwds)
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/driver.py", line 287, in _do
# [translation:ERROR] res = func()
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/driver.py", line 530, in task_source_c
# [translation:ERROR] exe_name=exe_name)
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/c/genc.py", line 252, in generate_source
# [translation:ERROR] split=self.split)
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/c/genc.py", line 989, in gen_source
# [translation:ERROR] sg.gen_readable_parts_of_source(f)
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/c/genc.py", line 843, in gen_readable_parts_of_source
# [translation:ERROR] for node, impl in nodeiter:
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/c/genc.py", line 729, in subiter
# [translation:ERROR] impl = '\n'.join(list(node.implementation())).split('\n')
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/c/node.py", line 867, in implementation
# [translation:ERROR] for s in self.funcgen_implementation(funcgen):
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/c/node.py", line 901, in funcgen_implementation
# [translation:ERROR] for line in bodyiter:
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/c/funcgen.py", line 272, in cfunction_body
# [translation:ERROR] blocks.sort(block_comparator)
# [translation:ERROR] File "/builddir/build/BUILD/pypy-pypy-2346207d9946/pypy/translator/c/funcgen.py", line 34, in block_comparator
# [translation:ERROR] if blk0.isstartblock:
# [translation:ERROR] AttributeError: 'Block' object has no attribute 'isstartblock'
2011-01-06 22:09:03 +00:00
2011-08-20 15:59:48 +00:00
%patch5 -p1
2012-07-10 20:45:52 +00:00
%patch6 -p1
%patch7 -p1
%patch8 -p1
%patch9 -p1
2011-01-03 20:06:28 +00:00
# Replace /usr/local/bin/python shebangs with /usr/bin/python:
find -name " * . p y " -exec \
sed \
-i -e " s | / u s r / l o c a l / b i n / p y t h o n | / u s r / b i n / p y t h o n | " \
" { } " \
\;
find . -name '*.jar' -exec rm \{\} \;
# Remove stray ".svn" directories present within the 1.4.1 tarball
# (reported as https://codespeak.net/issue/pypy-dev/issue612 )
find . -path '*/.svn*' -delete
# Remove DOS batch files:
find -name " * . b a t " |xargs rm -f
# The "demo" directory gets auto-installed by virture of being listed in %doc
# Remove shebang lines from demo .py files, and remove executability from them:
for f in demo/bpnn.py ; do
# Detect shebang lines && remove them:
sed -e '/^#!/Q 0' -e 'Q 1' $f \
&& sed -i '1d' $f
chmod a-x $f
done
%build
BuildPyPy() {
ExeName=$1
Options=$2
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " S T A R T I N G B U I L D O F : $ E x e N a m e "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
pushd %{goal_dir}
# The build involves invoking a python script, passing in particular
# arguments, environment variables, etc.
# Some notes on those follow:
# The generated binary embeds copies of the values of all environment
# variables. We need to unset "RPM_BUILD_ROOT" to avoid a fatal error from
# /usr/lib/rpm/check-buildroot
# during the postprocessing of the rpmbuild, complaining about this
# reference to the buildroot
# By default, pypy's autogenerated C code is placed in
# /tmp/usession-N
#
# and it appears that this stops rpm from extracting the source code to the
# debuginfo package
#
# The logic in pypy-1.4/pypy/tool/udir.py indicates that it is generated in:
# $PYPY_USESSION_DIR/usession-$PYPY_USESSION_BASENAME-N
# and so we set PYPY_USESSION_DIR so that this tempdir is within the build
# location, and set $PYPY_USESSION_BASENAME so that the tempdir is unique
# for each invocation of BuildPyPy
# Compilation flags for C code:
# pypy-1.4/pypy/translator/c/genc.py:gen_makefile
# assembles a Makefile within
# THE_UDIR/testing_1/Makefile
# calling out to platform.gen_makefile
# For us, that's
# pypy-1.4/pypy/translator/platform/linux.py: class BaseLinux(BasePosix):
# which by default has:
# CFLAGS = ['-O3', '-pthread', '-fomit-frame-pointer',
# '-Wall', '-Wno-unused']
# plus all substrings from CFLAGS in the environment.
# This is used to generate a value for CFLAGS that's written into the Makefile
2011-12-16 22:15:41 +00:00
# How will we track garbage-collection roots in the generated code?
# http://pypy.readthedocs.org/en/latest/config/translation.gcrootfinder.html
%if 0%{shadow_stack}
# This is the most portable option, and avoids a reliance on non-guaranteed
# behaviors within GCC's code generator: use an explicitly-maintained stack
# of root pointers:
%define gcrootfinder_options --gcrootfinder=shadowstack
export CFLAGS=$(echo " $RPM_OPT_FLAGS " )
%else
# Go with the default, which is "asmgcc"
%define gcrootfinder_options %{nil}
2011-01-03 20:06:28 +00:00
# https://bugzilla.redhat.com/show_bug.cgi?id=588941#c18
# The generated Makefile compiles the .c files into assembler (.s), rather
# than direct to .o It then post-processes this assembler to locate
# garbage-collection roots (building .lbl.s and .gcmap files, and a
# "gcmaptable.s"). (The modified .lbl.s files have extra code injected
# within them).
# Unfortunately, the code to do this:
# pypy-1.4/pypy/translator/c/gcc/trackgcroot.py
# doesn't interract well with the results of using our standard build flags.
# For now, filter our CFLAGS of everything that could be conflicting with
# pypy. Need to check these and reenable ones that are okay later.
2011-01-06 22:09:03 +00:00
# Filed as https://bugzilla.redhat.com/show_bug.cgi?id=666966
2011-01-03 20:06:28 +00:00
export CFLAGS=$(echo " $RPM_OPT_FLAGS " | sed -e 's/-Wp,-D_FORTIFY_SOURCE=2//' -e 's/-fexceptions//' -e 's/-fstack-protector//' -e 's/--param=ssp-buffer-size=4//' -e 's/-O2//' -e 's/-fasynchronous-unwind-tables//' -e 's/-march=i686//' -e 's/-mtune=atom//' )
2011-12-16 22:15:41 +00:00
%endif
# The generated C code leads to many thousands of warnings of the form:
# warning: variable 'l_v26003' set but not used [-Wunused-but-set-variable]
# Suppress them:
2012-06-18 08:50:36 +00:00
export CFLAGS=$(echo " $ C F L A G S " -Wno-unused -fPIC)
2011-12-16 22:15:41 +00:00
2011-01-03 20:06:28 +00:00
# If we're already built the JIT-enabled "pypy", then use it for subsequent
# builds (of other configurations):
if test -x './pypy' ; then
INTERP='./pypy'
else
2011-01-06 22:09:03 +00:00
# First pypy build within this rpm build?
# Fall back to using the bootstrap python interpreter, which might be a
# system copy of pypy from an earlier rpm, or be cpython's /usr/bin/python:
INTERP='%{bootstrap_python_interp}'
2011-01-03 20:06:28 +00:00
fi
# Here's where we actually invoke the build:
time \
RPM_BUILD_ROOT= \
PYPY_USESSION_DIR=$(pwd) \
PYPY_USESSION_BASENAME=$ExeName \
$INTERP translate.py \
%if 0%{verbose_logs}
--translation-verbose \
%endif
--cflags=" $ C F L A G S " \
--batch \
--output=$ExeName \
2011-12-16 22:15:41 +00:00
%{gcrootfinder_options} \
2011-01-03 20:06:28 +00:00
$Options
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " F I N I S H E D B U I L D I N G : $ E x e N a m e "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
popd
}
BuildPyPy \
pypy \
2011-04-30 22:24:07 +00:00
%if 0%{with_jit}
" - O j i t " \
fix build on non-jit arches
Build on non-jit arch like s390 fails with
[translation:info] Translating target as defined by targetpypystandalone
[platform:msg] Setting platform to 'host' cc=None
Traceback (most recent call last):
File "translate.py", line 322, in <module>
main()
File "translate.py", line 208, in main
targetspec_dic, translateconfig, config, args = parse_options_and_load_target()
File "translate.py", line 176, in parse_options_and_load_target
targetspec_dic['handle_config'](config, translateconfig)
File "targetpypystandalone.py", line 109, in handle_config
raise Exception("You have to specify the --opt level.\n"
Exception: You have to specify the --opt level.
Try --opt=2 or --opt=jit, or equivalently -O2 or -Ojit .
2011-05-11 08:15:52 +00:00
%else
" - O 2 " \
2011-01-03 20:06:28 +00:00
%endif
2011-04-30 22:24:07 +00:00
%{nil}
2011-01-03 20:06:28 +00:00
%if 0%{with_stackless}
BuildPyPy \
pypy-stackless \
" - - s t a c k l e s s "
%endif
2011-09-13 09:00:09 +00:00
%if %{with_emacs}
2011-08-03 17:02:57 +00:00
%{_emacs_bytecompile} pypy/jit/tool/pypytrace-mode.el
2011-09-13 09:00:09 +00:00
%endif
2011-08-03 17:02:57 +00:00
2011-01-03 20:06:28 +00:00
%install
rm -rf $RPM_BUILD_ROOT
# Install the various executables:
InstallPyPy() {
ExeName=$1
2012-01-31 09:53:38 +00:00
# To ensure compatibility with virtualenv, pypy finds its libraries
# relative to itself; this happens within
# pypy/translator/goal/app_main.py:get_library_path
# which calls sys.pypy_initial_path(dirname) on the dir containing
# the executable, with symlinks resolved.
#
# Hence we make /usr/bin/pypy be a symlink to the real binary, which we
# place within /usr/lib[64]/pypy-1.* as pypy
#
# This ought to enable our pypy build to work with virtualenv
# (rhbz#742641)
install -m 755 %{goal_dir} /$ExeName %{buildroot} /%{pypyprefix} /$ExeName
ln -s %{pypyprefix} /$ExeName %{buildroot} /%{_bindir}
2011-01-03 20:06:28 +00:00
# The generated machine code doesn't need an executable stack, but
# one of the assembler files (gcmaptable.s) doesn't have the necessary
# metadata to inform gcc of that, and thus gcc pessimistically assumes
# that the built binary does need an executable stack.
#
# Reported upstream as: https://codespeak.net/issue/pypy-dev/issue610
#
# I tried various approaches involving fixing the build, but the simplest
# approach is to postprocess the ELF file:
2012-01-31 09:53:38 +00:00
execstack --clear-execstack %{buildroot} /%{pypyprefix} /$ExeName
2011-01-03 20:06:28 +00:00
}
mkdir -p %{buildroot} /%{_bindir}
2012-01-31 09:53:38 +00:00
mkdir -p %{buildroot} /%{pypyprefix}
2011-01-03 20:06:28 +00:00
InstallPyPy pypy
%if 0%{with_stackless}
InstallPyPy pypy-stackless
%endif
# Install the various support libraries as described at:
# http://codespeak.net/pypy/dist/pypy/doc/getting-started-python.html#installation
# which refers to a "PREFIX" found relative to the location of the binary.
# Given that the pypy binaries will be in /usr/bin, PREFIX can be
# "../share/pypy-1.2" relative to that directory, i.e. /usr/share/pypy-1.2
#
# Running "strace" on a built binary indicates that it searches within
# PREFIX/lib-python/modified-2.5.2
# not
# PREFIX/lib-python/modified.2.5.2
# as given on the above page, i.e. it uses '-' not '.'
cp -a lib-python %{buildroot} /%{pypyprefix}
cp -a lib_pypy %{buildroot} /%{pypyprefix}
# Remove a text file that documents which selftests fail on Win32:
rm %{buildroot} /%{pypyprefix} /lib-python/win32-failures.txt
# Remove shebang lines from .py files that aren't executable, and
# remove executability from .py files that don't have a shebang line:
find \
%{buildroot} \
-name " * . p y " \
\( \
\( \! -perm /u+x,g+x,o+x -exec sed -e '/^#!/Q 0' -e 'Q 1' {} \; \
-print -exec sed -i '1d' {} \; \
\) \
-o \
\( \
-perm /u+x,g+x,o+x ! -exec grep -m 1 -q '^#!' {} \; \
-exec chmod a-x {} \; \
\) \
\)
mkdir -p %{buildroot} /%{pypyprefix} /site-packages
# pypy uses .pyc files by default (--objspace-usepycfiles), but has a slightly
# different bytecode format to CPython. It doesn't use .pyo files: the -O flag
# is treated as a "dummy optimization flag for compatibility with C Python"
#
# pypy-1.4/pypy/module/imp/importing.py has this comment:
# XXX picking a magic number is a mess. So far it works because we
# have only two extra opcodes, which bump the magic number by +1 and
# +2 respectively, and CPython leaves a gap of 10 when it increases
# its own magic number. To avoid assigning exactly the same numbers
# as CPython we always add a +2. We'll have to think again when we
# get at the fourth new opcode :-(
#
# * CALL_LIKELY_BUILTIN +1
# * CALL_METHOD +2
#
# In other words:
#
# default_magic -- used by CPython without the -U option
# default_magic + 1 -- used by CPython with the -U option
# default_magic + 2 -- used by PyPy without any extra opcode
# ...
# default_magic + 5 -- used by PyPy with both extra opcodes
#
# pypy-1.4/pypy/interpreter/pycode.py has:
#
# default_magic = (62141+2) | 0x0a0d0000 # this PyPy's magic
# # (62131=CPython 2.5.1)
# giving a value for "default_magic" for PyPy of 0xa0df2bf.
# Note that this corresponds to the "default_magic + 2" from the comment above
# In my builds:
# $ ./pypy --info | grep objspace.opcodes
# objspace.opcodes.CALL_LIKELY_BUILTIN: False
# objspace.opcodes.CALL_METHOD: True
# so I'd expect the magic number to be:
# 0x0a0df2bf + 2 (the flag for CALL_METHOD)
# giving
# 0x0a0df2c1
#
# I'm seeing
# c1 f2 0d 0a
# as the first four bytes of the .pyc files, which is consistent with this.
# Bytecompile all of the .py files we ship, using our pypy binary, giving us
# .pyc files for pypy. The script actually does the work twice (passing in -O
# the second time) but it's simplest to reuse that script.
#
# The script has special-casing for .py files below
# /usr/lib{64}/python[0-9].[0-9]
# but given that we're installing into a different path, the supplied "default"
# implementation gets used instead.
#
# Note that some of the test files deliberately contain syntax errors, so
# we pass 0 for the second argument ("errors_terminate"):
/usr/lib/rpm/brp-python-bytecompile \
%{buildroot} /%{_bindir} /pypy \
0
2011-01-14 00:51:26 +00:00
# Header files for C extension modules.
# Upstream's packaging process (pypy/tool/release/package.py)
# creates an "include" subdir and copies all *.h/*.inl from "include" there
# (it also has an apparently out-of-date comment about copying them from
# pypy/_interfaces, but this directory doesn't seem to exist, and it doesn't
# seem to do this as of 2011-01-13)
# FIXME: arguably these should be instead put into a subdir below /usr/include,
# it's not yet clear to me how upstream plan to deal with the C extension
# interface going forward, so let's just mimic upstream for now.
%global pypy_include_dir %{pypyprefix} /include
mkdir -p %{buildroot} /%{pypy_include_dir}
Rebase to 1.5 (rhbz#701121)
Update patches:
patch 0: pypy-1.4-config.patch -> pypy-1.5-config.patch
patch 4: pypy-1.4.1-more-readable-c-code.patch -> pypy-1.5-more-readable-c-code.patch
Remove references to *.inl files, no longer present
Add the following tests to the skip list:
test_audioop, test_capi, test_distutils, test_gc, test_gdb, test_generators,
test_getargs2, test_hotshot, test_io, test_multiprocessing, test_posix,
test_readline, test_scope, test_strop, test_structmembers, test_subprocess,
test_symtable, test_sys_settrace, test_tempfile, test_thread, test_uuid,
test_zipimport_support
Add a couple of text files to the payload (TODO, stdlib-version.txt)
2011-05-02 18:46:32 +00:00
cp include/*.h %{buildroot} /%{pypy_include_dir}
2011-01-14 00:51:26 +00:00
2011-01-06 22:09:03 +00:00
# Capture the RPython source code files from the build within the debuginfo
# package (rhbz#666975)
%global pypy_debuginfo_dir /usr/src/debug/pypy-%{version} -src
mkdir -p %{buildroot} %{pypy_debuginfo_dir}
# copy over everything:
cp -a pypy %{buildroot} %{pypy_debuginfo_dir}
2012-07-10 20:45:52 +00:00
# ...then delete files that aren't:
# - *.py files
# - the Makefile
# - typeids.txt
# - dynamic-symbols-*
2011-01-06 22:09:03 +00:00
find \
2012-07-10 20:45:52 +00:00
%{buildroot} %{pypy_debuginfo_dir} \
\( -type f \
-a \
\! \( -name " * . p y " \
-o \
-name " M a k e f i l e " \
-o \
-name " t y p e i d s . t x t " \
-o \
-name " d y n a m i c - s y m b o l s - * " \
\) \
\) \
2011-01-06 22:09:03 +00:00
-delete
2012-07-10 20:45:52 +00:00
# Alternatively, we could simply keep everything. This leads to a ~350MB
# debuginfo package, but it makes it easy to hack on the Makefile and C build
# flags by rebuilding/linking the sources.
# To do so, remove the above "find" command.
2011-01-06 22:09:03 +00:00
# We don't need bytecode for these files; they are being included for reference
# purposes.
# There are some rpmlint warnings from these files:
# non-executable-script
# wrong-script-interpreter
# zero-length
# script-without-shebang
# dangling-symlink
# but given that the objective is to preserve a copy of the source code, those
# are acceptable.
2011-08-03 17:02:57 +00:00
# Install the JIT trace mode for Emacs:
2011-09-13 09:00:09 +00:00
%if %{with_emacs}
2011-08-03 17:02:57 +00:00
mkdir -p %{buildroot} /%{_emacs_sitelispdir}
cp -a pypy/jit/tool/pypytrace-mode.el* %{buildroot} /%{_emacs_sitelispdir}
2011-09-13 09:00:09 +00:00
%endif
2011-08-03 17:02:57 +00:00
2011-08-25 15:58:55 +00:00
# Install macros for rpm:
mkdir -p %{buildroot} /%{_sysconfdir} /rpm
install -m 644 %{SOURCE2} %{buildroot} /%{_sysconfdir} /rpm
2011-01-03 20:06:28 +00:00
%check
topdir=$(pwd)
SkipTest() {
TEST_NAME=$1
2011-08-25 18:50:53 +00:00
sed -i -e" s | ^ $ T E S T _ N A M E $ | | g " testnames.txt
2011-01-03 20:06:28 +00:00
}
CheckPyPy() {
# We'll be exercising one of the freshly-built binaries using the
# test suite from the standard library (overridden in places by pypy's
# modified version)
ExeName=$1
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " S T A R T I N G T E S T O F : $ E x e N a m e "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
pushd %{goal_dir}
2011-08-19 23:04:38 +00:00
# I'm seeing numerous cases where tests seem to hang, or fail unpredictably
# So we'll run each test in its own process, with a timeout
# Use regrtest to explicitly list all tests:
( ./$ExeName -c \
" f r o m t e s t . r e g r t e s t i m p o r t f i n d t e s t s ; p r i n t ' \n ' . j o i n ( f i n d t e s t s ( ) ) "
) > testnames.txt
2011-08-25 15:58:55 +00:00
# Skip some tests:
# "audioop" doesn't exist for pypy yet:
SkipTest test_audioop
# The gdb CPython hooks haven't been ported to cpyext:
SkipTest test_gdb
# hotshot relies heavily on _hotshot, which doesn't exist:
SkipTest test_hotshot
# "strop" module doesn't exist for pypy yet:
SkipTest test_strop
2011-10-04 18:37:40 +00:00
# I'm seeing Koji builds hanging e.g.:
# http://koji.fedoraproject.org/koji/getfile?taskID=3386821&name=build.log
# The only test that seems to have timed out in that log is
# test_multiprocessing, so skip it for now:
SkipTest test_multiprocessing
2011-08-19 23:04:38 +00:00
echo " = = T e s t n a m e s = = "
cat testnames.txt
echo " = = = = = = = = = = = = = = = = = "
echo " " > failed-tests.txt
for TestName in $(cat testnames.txt) ; do
echo " = = = = = = = = = = = = = = = = = = = " $TestName " = = = = = = = = = = = = = = = = = = = = "
# Use /usr/bin/time (rather than the shell "time" builtin) to gather
# info on the process (time/CPU/memory). This passes on the exit
# status of the underlying command
#
# Use perl's alarm command to impose a timeout
# 900 seconds is 15 minutes per test.
# If a test hangs, that test should get terminated, allowing the build
# to continue.
#
# Invoke pypy on test.regrtest to run the specific test suite
# verbosely
#
# For now, || true, so that any failures don't halt the build:
( /usr/bin/time \
perl -e 'alarm shift @ARGV; exec @ARGV' 900 \
./$ExeName -m test.regrtest -v $TestName ) \
|| (echo $TestName >> failed-tests.txt) \
|| true
done
echo " = = F a i l e d t e s t s = = "
cat failed-tests.txt
echo " = = = = = = = = = = = = = = = = = "
2011-01-03 20:06:28 +00:00
popd
# Doublecheck pypy's own test suite, using the built pypy binary:
# Disabled for now:
# x86_64 shows various failures inside:
# jit/backend/x86/test
# followed by a segfault inside
# jit/backend/x86/test/test_runner.py
#
# i686 shows various failures inside:
# jit/backend/x86/test
# with the x86_64 failure leading to cancellation of the i686 build
# Here's the disabled code:
# pushd pypy
# time translator/goal/$ExeName test_all.py
# popd
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " F I N I S H E D T E S T I N G : $ E x e N a m e "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
echo " - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - "
}
2011-08-19 23:04:38 +00:00
%if %{run_selftests}
2011-01-03 20:06:28 +00:00
CheckPyPy pypy
%if 0%{with_stackless}
CheckPyPy pypy-stackless
%endif
2011-08-19 23:04:38 +00:00
%endif # run_selftests
2011-01-03 20:06:28 +00:00
%clean
rm -rf $RPM_BUILD_ROOT
%files libs
%defattr (-,root,root,-)
%doc LICENSE README demo
%dir %{pypyprefix}
%dir %{pypyprefix} /lib-python
Rebase to 1.5 (rhbz#701121)
Update patches:
patch 0: pypy-1.4-config.patch -> pypy-1.5-config.patch
patch 4: pypy-1.4.1-more-readable-c-code.patch -> pypy-1.5-more-readable-c-code.patch
Remove references to *.inl files, no longer present
Add the following tests to the skip list:
test_audioop, test_capi, test_distutils, test_gc, test_gdb, test_generators,
test_getargs2, test_hotshot, test_io, test_multiprocessing, test_posix,
test_readline, test_scope, test_strop, test_structmembers, test_subprocess,
test_symtable, test_sys_settrace, test_tempfile, test_thread, test_uuid,
test_zipimport_support
Add a couple of text files to the payload (TODO, stdlib-version.txt)
2011-05-02 18:46:32 +00:00
%{pypyprefix} /lib-python/stdlib-version.txt
2011-01-03 20:06:28 +00:00
%{pypyprefix} /lib-python/%{pylibver} /
%{pypyprefix} /lib-python/conftest.py*
%{pypyprefix} /lib_pypy/
%{pypyprefix} /site-packages/
2011-09-13 09:00:09 +00:00
%if %{with_emacs}
2011-08-03 17:02:57 +00:00
%{_emacs_sitelispdir} /pypytrace-mode.el
%{_emacs_sitelispdir} /pypytrace-mode.elc
2011-09-13 09:00:09 +00:00
%endif
2011-01-03 20:06:28 +00:00
%files
%defattr (-,root,root,-)
%doc LICENSE README
%{_bindir} /pypy
2012-01-31 09:53:38 +00:00
%{pypyprefix} /pypy
2011-01-03 20:06:28 +00:00
2011-01-14 00:51:26 +00:00
%files devel
%defattr (-,root,root,-)
%dir %{pypy_include_dir}
%{pypy_include_dir} /*.h
2011-08-25 15:58:55 +00:00
%config (noreplace) %{_sysconfdir} /rpm/macros.pypy
2011-01-14 00:51:26 +00:00
2011-01-03 20:06:28 +00:00
%if 0%{with_stackless}
%files stackless
%defattr (-,root,root,-)
%doc LICENSE README
%{_bindir} /pypy-stackless
%endif
%changelog
2012-07-10 20:45:52 +00:00
* Tue Jul 10 2012 David Malcolm <dmalcolm@redhat.com> - 1.9-3
- log all output from " m a k e " (patch 6)
- disable the MOTD at startup (patch 7)
- hide symbols from the dynamic linker (patch 8)
- add PyInt_AsUnsignedLongLongMask (patch 9)
- capture the Makefile, the typeids.txt, and the dynamic-symbols file within
the debuginfo package
2012-06-18 08:50:36 +00:00
* Mon Jun 18 2012 Peter Robinson <pbrobinson@fedoraproject.org> - 1.9-2
- Compile with PIC, fixes FTBFS on ARM
2012-06-09 00:18:15 +00:00
* Fri Jun 8 2012 David Malcolm <dmalcolm@redhat.com> - 1.9-1
- 1.9
2012-02-10 16:12:43 +00:00
* Fri Feb 10 2012 David Malcolm <dmalcolm@redhat.com> - 1.8-2
- disable C readability patch for now (patch 4)
2012-02-10 14:19:56 +00:00
* Thu Feb 9 2012 David Malcolm <dmalcolm@redhat.com> - 1.8-1
- 1.8; regenerate config patch (patch 0); drop selinux patch (patch 2);
regenerate patch 5
2012-01-31 09:53:38 +00:00
* Tue Jan 31 2012 David Malcolm <dmalcolm@redhat.com> - 1.7-4
- fix an incompatibility with virtualenv (rhbz#742641)
2012-01-14 01:29:18 +00:00
* Sat Jan 14 2012 Fedora Release Engineering <rel-eng@lists.fedoraproject.org> - 1.7-3
- Rebuilt for https://fedoraproject.org/wiki/Fedora_17_Mass_Rebuild
2011-12-16 22:15:41 +00:00
* Fri Dec 16 2011 David Malcolm <dmalcolm@redhat.com> - 1.7-2
- use --gcrootfinder=shadowstack, and use standard Fedora compilation flags,
with -Wno-unused (rhbz#666966 and rhbz#707707)
2011-11-22 16:30:49 +00:00
* Mon Nov 21 2011 David Malcolm <dmalcolm@redhat.com> - 1.7-1
- 1.7: refresh patch 0 (configuration) and patch 4 (readability of generated
code)
2011-10-04 18:37:40 +00:00
* Tue Oct 4 2011 David Malcolm <dmalcolm@redhat.com> - 1.6-7
- skip test_multiprocessing
2011-09-13 09:00:09 +00:00
* Tue Sep 13 2011 David Malcolm <dmalcolm@redhat.com> - 1.6-6
- don't ship the emacs JIT-viewer on el5 and el6 (missing emacs-filesystem;
missing _emacs_bytecompile macro on el5)
2011-09-13 01:01:16 +00:00
* Mon Sep 12 2011 David Malcolm <dmalcolm@redhat.com> - 1.6-5
- build using python26 on el5 (2.4 is too early)
2011-08-25 18:50:53 +00:00
* Thu Aug 25 2011 David Malcolm <dmalcolm@redhat.com> - 1.6-4
- fix SkipTest function to avoid corrupting the name of " t e s t _ g d b m "
2011-08-25 15:58:55 +00:00
* Thu Aug 25 2011 David Malcolm <dmalcolm@redhat.com> - 1.6-3
- add rpm macros file to the devel subpackage (source 2)
- skip some tests that can't pass yet
2011-08-20 15:59:48 +00:00
* Sat Aug 20 2011 David Malcolm <dmalcolm@redhat.com> - 1.6-2
- work around test_subprocess failure seen in koji (patch 5)
2011-08-19 23:04:38 +00:00
* Thu Aug 18 2011 David Malcolm <dmalcolm@redhat.com> - 1.6-1
- 1.6
- rewrite the %%check section, introducing per-test timeouts
2011-08-03 17:02:57 +00:00
* Tue Aug 2 2011 David Malcolm <dmalcolm@redhat.com> - 1.5-2
- add pypytrace-mode.el to the pypy-libs subpackage, for viewing JIT trace
logs in emacs
Rebase to 1.5 (rhbz#701121)
Update patches:
patch 0: pypy-1.4-config.patch -> pypy-1.5-config.patch
patch 4: pypy-1.4.1-more-readable-c-code.patch -> pypy-1.5-more-readable-c-code.patch
Remove references to *.inl files, no longer present
Add the following tests to the skip list:
test_audioop, test_capi, test_distutils, test_gc, test_gdb, test_generators,
test_getargs2, test_hotshot, test_io, test_multiprocessing, test_posix,
test_readline, test_scope, test_strop, test_structmembers, test_subprocess,
test_symtable, test_sys_settrace, test_tempfile, test_thread, test_uuid,
test_zipimport_support
Add a couple of text files to the payload (TODO, stdlib-version.txt)
2011-05-02 18:46:32 +00:00
* Mon May 2 2011 David Malcolm <dmalcolm@redhat.com> - 1.5-1
- 1.5
2011-04-30 22:24:07 +00:00
* Wed Apr 20 2011 David Malcolm <dmalcolm@redhat.com> - 1.4.1-10
- build a /usr/bin/pypy (but without the JIT compiler) on architectures that
don't support the JIT, so that they do at least have something that runs
2011-02-09 02:23:05 +00:00
* Tue Feb 08 2011 Fedora Release Engineering <rel-eng@lists.fedoraproject.org> - 1.4.1-9
- Rebuilt for https://fedoraproject.org/wiki/Fedora_15_Mass_Rebuild
2011-01-14 16:58:45 +00:00
* Fri Jan 14 2011 David Malcolm <dmalcolm@redhat.com> - 1.4.1-8
- disable self-hosting for now, due to fatal error seen JIT-compiling the
translator
2011-01-14 16:06:07 +00:00
* Fri Jan 14 2011 David Malcolm <dmalcolm@redhat.com> - 1.4.1-7
- skip test_ioctl for now
2011-01-14 00:51:26 +00:00
* Thu Jan 13 2011 David Malcolm <dmalcolm@redhat.com> - 1.4.1-6
- add a " p y p y - d e v e l " subpackage, and install the header files there
- in %%check, re-run failed tests in verbose mode
2011-01-07 16:56:06 +00:00
* Fri Jan 7 2011 Dan Horák <dan[at]danny.cz> - 1.4.1-5
- valgrind available only on selected architectures
2011-01-06 22:09:03 +00:00
* Wed Jan 5 2011 David Malcolm <dmalcolm@redhat.com> - 1.4.1-4
- rebuild pypy using itself, for speed, with a boolean to break this cycle in
the build-requirement graph (falling back to using " p y t h o n - d e v e l " aka CPython)
- add work-in-progress patch to try to make generated c more readable
(rhbz#666963)
- capture the RPython source code files from the build within the debuginfo
package (rhbz#666975)
2011-01-03 20:06:28 +00:00
* Wed Dec 22 2010 David Malcolm <dmalcolm@redhat.com> - 1.4.1-3
- try to respect the FHS by installing libraries below libdir, rather than
datadir; patch app_main.py to look in this installation location first when
scanning for the pypy library directories.
- clarifications and corrections to the comments in the specfile
* Wed Dec 22 2010 David Malcolm <dmalcolm@redhat.com> - 1.4.1-2
- remove .svn directories
- disable verbose logging
- add a %%check section
- introduce %%goal_dir variable, to avoid repetition
- remove shebang line from demo/bpnn.py, as we're treating this as a
documentation file
- regenerate patch 2 to apply without generating a .orig file
* Tue Dec 21 2010 David Malcolm <dmalcolm@redhat.com> - 1.4.1-1
- 1.4.1; fixup %%setup to reflect change in toplevel directory in upstream
source tarball
- apply SELinux fix to the bundled test_commands.py (patch 2)
* Wed Dec 15 2010 David Malcolm <dmalcolm@redhat.com> - 1.4-4
- rename the jit build and subpackge to just " p y p y " , and remove the nojit and
sandbox builds, as upstream now seems to be focussing on the JIT build (with
only stackless called out in the getting-started-python docs); disable
stackless for now
- add a verbose_logs specfile boolean; leave it enabled for now (whilst fixing
build issues)
- add more comments, and update others to reflect 1.2 -> 1.4 changes
- re-enable debuginfo within CFLAGS (" - g " )
- add the LICENSE and README to all subpackages
- ensure the built binaries don't have the " I n e e d a n e x e c u t a b l e s t a c k " flag
- remove DOS batch files during %%prep (idlelib.bat)
- remove shebang lines from .py files that aren't executable, and remove
executability from .py files that don't have a shebang line (taken from
our python3.spec)
- bytecompile the .py files into .pyc files in pypy's bytecode format
* Sun Nov 28 2010 Toshio Kuratomi <toshio@fedoraproject.org> - 1.4-3
- BuildRequire valgrind-devel
- Install pypy library from the new directory
- Disable building with our CFLAGS for now because they are causing a build failure.
- Include site-packages directory
* Sat Nov 27 2010 Toshio Kuratomi <toshio@fedoraproject.org> - 1.4-2
- Add patch to configure the build to use our CFLAGS and link libffi
dynamically
* Sat Nov 27 2010 Toshio Kuratomi <toshio@fedoraproject.org> - 1.4-1
- Update to 1.4
- Drop patch for py2.6 that's in this build
- Switch to building pypy with itself once pypy is built once as recommended by
upstream
- Remove bundled, prebuilt java libraries
- Fix license tag
- Fix source url
- Version pypy-libs Req
* Tue May 4 2010 David Malcolm <dmalcolm@redhat.com> - 1.2-2
- cherrypick r72073 from upstream SVN in order to fix the build against
python 2.6.5 (patch 2)
* Wed Apr 28 2010 David Malcolm <dmalcolm@redhat.com> - 1.2-1
- initial packaging
