-
-
Notifications
You must be signed in to change notification settings - Fork 31.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Reintroduce bootstrap_python for freezing #90108
Comments
bpo-45696 introduced a _bootstrap_python interpreter, which was used to create frozen and deepfrozen module files. bpo-45873 dropped the _bootstrap_python interpreter again. Instead Python used an existing Python installation to bootstrap the build. After some internal discussion we agreed that the approach is flawed and puts too much of a burden on users. Users have to install a fairly recent Python interpreter in order to build 3.11. This makes it harder to build Python on platforms with no or a too old Python version. It also complicated things for container builds and pyenv. Let's re-introduce a _bootstrap_python interpreter. Cross compiling will still require a build Python interpreter. |
Framework builds on macOS now fail with a bus error when trying to run the newly built _bootstrap_python. A simplified example: $ ./configure --enable-framework=$PWD/Library/Frameworks
checking for git... found
checking build system type... x86_64-apple-darwin20.6.0
checking host system type... x86_64-apple-darwin20.6.0
checking for Python interpreter freezing... ./_bootstrap_python
checking for python3.11... no
checking for python3.10... no
checking for python3.9... no
checking for python3.8... no
checking for python3.7... no
checking for python3.6... no
checking for python3... python3
checking Python for regen version... Python 3.8.9
[...]
creating Modules/Setup.local
creating Makefile pkg-config is missing. Some dependencies may not be detected correctly. If you want a release build with all stable optimizations active (PGO, etc), $ make
[...]
gcc -o _bootstrap_python Modules/getbuildinfo.o Parser/token.o Parser/pegen.o Parser/pegen_errors.o Parser/action_helpers.o Parser/parser.o Parser/string_parser.o Parser/peg_api.o Parser/myreadline.o Parser/tokenizer.o Objects/abstract.o Objects/accu.o Objects/boolobject.o Objects/bytes_methods.o Objects/bytearrayobject.o Objects/bytesobject.o Objects/call.o Objects/capsule.o Objects/cellobject.o Objects/classobject.o Objects/codeobject.o Objects/complexobject.o Objects/descrobject.o Objects/enumobject.o Objects/exceptions.o Objects/genericaliasobject.o Objects/genobject.o Objects/fileobject.o Objects/floatobject.o Objects/frameobject.o Objects/funcobject.o Objects/interpreteridobject.o Objects/iterobject.o Objects/listobject.o Objects/longobject.o Objects/dictobject.o Objects/odictobject.o Objects/memoryobject.o Objects/methodobject.o Objects/moduleobject.o Objects/namespaceobject.o Objects/object.o Objects/obmalloc.o Objects/picklebufobject.o Objects/rangeobject.o Objects/setobject.o Objects/sliceobject.o Objects/structseq.o Objects/tupleobject.o Objects/typeobject.o Objects/unicodeobject.o Objects/unicodectype.o Objects/unionobject.o Objects/weakrefobject.o Python/_warnings.o Python/Python-ast.o Python/Python-tokenize.o Python/asdl.o Python/ast.o Python/ast_opt.o Python/ast_unparse.o Python/bltinmodule.o Python/ceval.o Python/codecs.o Python/compile.o Python/context.o Python/dynamic_annotations.o Python/errors.o Python/frame.o Python/frozenmain.o Python/future.o Python/getargs.o Python/getcompiler.o Python/getcopyright.o Python/getplatform.o Python/getversion.o Python/hamt.o Python/hashtable.o Python/import.o Python/importdl.o Python/initconfig.o Python/marshal.o Python/modsupport.o Python/mysnprintf.o Python/mystrtoul.o Python/pathconfig.o Python/preconfig.o Python/pyarena.o Python/pyctype.o Python/pyfpe.o Python/pyhash.o Python/pylifecycle.o Python/pymath.o Python/pystate.o Python/pythonrun.o Python/pytime.o Python/bootstrap_hash.o Python/specialize.o Python/structmember.o Python/symtable.o Python/sysmodule.o Python/thread.o Python/traceback.o Python/getopt.o Python/pystrcmp.o Python/pystrtod.o Python/pystrhex.o Python/dtoa.o Python/formatter_unicode.o Python/fileutils.o Python/suggestions.o Python/dynload_shlib.o Modules/config.o Modules/main.o Modules/gcmodule.o Modules/atexitmodule.o Modules/faulthandler.o Modules/posixmodule.o Modules/signalmodule.o Modules/_tracemalloc.o Modules/_codecsmodule.o Modules/_collectionsmodule.o Modules/errnomodule.o Modules/_io/_iomodule.o Modules/_io/iobase.o Modules/_io/fileio.o Modules/_io/bytesio.o Modules/_io/bufferedio.o Modules/_io/textio.o Modules/_io/stringio.o Modules/itertoolsmodule.o Modules/_sre.o Modules/_threadmodule.o Modules/timemodule.o Modules/_weakref.o Modules/_abc.o Modules/_functoolsmodule.o Modules/_localemodule.o Modules/_operator.o Modules/_stat.o Modules/symtablemodule.o Modules/pwdmodule.o Modules/xxsubtype.o \
Programs/_bootstrap_python.o Modules/getpath.o -ldl -framework CoreFoundation
./_bootstrap_python ./Tools/scripts/deepfreeze.py Python/frozen_modules/importlib._bootstrap.h -m importlib._bootstrap -o Python/deepfreeze/importlib._bootstrap.c
make: *** [Python/deepfreeze/importlib._bootstrap.c] Bus error: 10
$ --enable-framework causes a number of changes in interpreter building and initialization. Since, as it stands, _bootstrap_python executable isn't and doesn't need to be built as a framework, there is probably some path confusion in getpath.c or elsewhere. Prior to these changes (at 3.11.0a2 at least), it was still possible to run a framework build executable in the build directory through some careful use of build variables in Makefile (like how "make test" works). Without delving deeper into it at the moment, I'm not sure what to suggest. It may be simpler to just do a clean separate non-framework build to provide _bootstrap_python. And/or possibly re-allow the use of --with-build-python to specify a user-supplied build python. I wonder if there are certain configurations on other platforms where there might be similar problems with the current approach. I'm marking this as a "release blocker" since it is preventing building the usual python.org macOS binary installer for 3.11.0a3. Since it is still early in the release cycle for 3.11, the release manager might reasonably decide to not delay the release for a fix and thus skip having a macOS installer for a3. |
Ned, does this change solve the issue with framework builds? $ git diff
diff --git a/Makefile.pre.in b/Makefile.pre.in
index 8e6e553554d..2068db30855 100644
--- a/Makefile.pre.in
+++ b/Makefile.pre.in
@@ -106,7 +106,7 @@ PY_CFLAGS_NODIST=$(CONFIGURE_CFLAGS_NODIST) $(CFLAGS_NODIST) -I$(srcdir)/Include
PY_CPPFLAGS= $(BASECPPFLAGS) -I. -I$(srcdir)/Include $(CONFIGURE_CPPFLAGS) $(CPPFLAGS)
PY_LDFLAGS= $(CONFIGURE_LDFLAGS) $(LDFLAGS)
PY_LDFLAGS_NODIST=$(CONFIGURE_LDFLAGS_NODIST) $(LDFLAGS_NODIST)
-PY_LDFLAGS_NOLTO=$(PY_LDFLAGS) $(CONFIGURE_LDFLAGS_NOLTO) $(LDFLAGS_NODIST)
+PY_LDFLAGS_NOLTO=$(PY_LDFLAGS_NODIST) $(CONFIGURE_LDFLAGS_NOLTO) $(LDFLAGS_NODIST)
NO_AS_NEEDED= @NO_AS_NEEDED@
CCSHARED= @CCSHARED@
# LINKFORSHARED are the flags passed to the $(CC) command that links |
No. I would guess at a minimum getpath.c for _bootstrap_python might need to be built without WITH_NEXT_FRAMEWORK defined. But I'm out of time for today and don't expect to have much time available in the immediate future. |
Ronald, Pablo, could you please test #74122 on macOS with --enable-framework? |
If the patch does not help then I can re-enable --with-build-python for non-cross builds. |
I added --enable-framework=yes to macOS GHA builder. #74125 has ./configure --enable-framework=yes without proposed bootstrap fix. The build is failing early: ./_bootstrap_python ./Tools/scripts/deepfreeze.py Python/frozen_modules/importlib._bootstrap.h -m importlib._bootstrap -o Python/deepfreeze/importlib._bootstrap.c #74122 has ./configure --enable-framework=yes and introduces a Modules/getpath_bootstrap.o file. It is the same build as Modules/getpath.o except it does not use the framework feature. The target compiles getpath.c with an additional -DPY_BOOTSTRAP_PYTHON=1 option. The build fails later: /bin/sh: line 1: 14772 Bus error: 10 DYLD_FRAMEWORK_PATH=/Users/runner/work/cpython/cpython ./python.exe -E -S -m sysconfig --generate-posix-vars |
Ned, I have implemented a fix for _bootstrap_python for framework builds and re-enabled --with-build-python for all builds. Previously I disabled the option for standard builds, because --with-build-python can cause build issues when a user runs ./configure --with-build-python with an older 3.11 alpha build. |
I've now verified that the _bootstrap_python fix for framework builds works - removing "release blocker" status. Thanks, Christian! |
The issue has been fixed, I am closing the issue. |
LTO optimization is nice to make Python faster, but _freeze_module and _testembed performance is not important. Using LTO to build these two programs make a whole Python build way slower, especially combined with a sanitizer (like ASAN).
LTO optimization is nice to make Python faster, but _freeze_module and _testembed performance is not important. Using LTO to build these two programs make a whole Python build way slower, especially combined with a sanitizer (like ASAN).
…109581) LTO optimization is nice to make Python faster, but _freeze_module and _testembed performance is not important. Using LTO to build these two programs make a whole Python build way slower, especially combined with a sanitizer (like ASAN).
…e and _testembed (python#109581)" This reverts commit 3e3a7da.
…109581) LTO optimization is nice to make Python faster, but _freeze_module and _testembed performance is not important. Using LTO to build these two programs make a whole Python build way slower, especially combined with a sanitizer (like ASAN).
…e and _teste… (python#110720) pythongh-110313: Revert "pythongh-90108: Disable LTO on _freeze_module and _testembed (python#109581)" This reverts commit 3e3a7da.
Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.
Show more details
GitHub fields:
bugs.python.org fields:
Linked PRs
The text was updated successfully, but these errors were encountered: