Skip to content

Building Programs/_freeze_module sometimes fails with missing symbols, like undefined reference to PyErr_CheckSignals #161364

Closed

Description

@vstinner

Bug report

Sometimes, when I build Python incrementally (build Python with make, modify C code, run make again), the build fails with errors such as:

gcc     -o Programs/_freeze_module Programs/_freeze_module.o Modules/getpath_noop.o (...) Modules/gcmodule.o  -ldl   -lm 
/usr/bin/ld.bfd: Parser/myreadline.o: in function `my_fgets':
/home/vstinner/python/main/Parser/myreadline.c:105:(.text+0x64): undefined reference to `PyErr_CheckSignals'

The PyErr_CheckSignals() function is implemented by Modules/signalmodule.c, but it seems like Modules/signalmodule.o is not used to build Programs/_freeze_module program.

Programs/_freeze_module rule in Makefile:

Programs/_freeze_module: Programs/_freeze_module.o Modules/getpath_noop.o $(BOOTSTRAP_LIBRARY_OBJS)
	$(LINKCC) $(PY_CORE_LDFLAGS) -o $@ Programs/_freeze_module.o Modules/getpath_noop.o $(BOOTSTRAP_LIBRARY_OBJS) $(LIBS) $(BOOTSTRAP_MODLIBS) $(SYSLIBS)

BOOTSTRAP_LIBRARY_OBJS variable:

BOOTSTRAP_LIBRARY_OBJS=	\
		Modules/getbuildinfo_bootstrap.o \
		$(PARSER_OBJS) \
		$(OBJECT_OBJS) \
		$(PYTHON_OBJS) \
		Modules/config_bootstrap.o \
		$(MODULE_OBJS_COMMON) \
		$(BOOTSTRAP_MODOBJS)

MODULE_OBJS_COMMON and BOOTSTRAP_MODOBJS variables:

MODULE_OBJS_COMMON=	\
		Modules/main.o \
		Modules/gcmodule.o
BOOTSTRAP_MODOBJS= 

Full Makefile file (created on Fedora 45): Makefile

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

Activity

  1. vstinner commented on Nov 4, 2026

    @vstinner
    MemberAuthor

    I removed all generated files and I built Python from scratch:

    git clean -fdx
    ./configure --with-pydebug
    make 
    

    It fixed the issue: BOOTSTRAP_MODOBJS now contains Modules/signalmodule.o.

    The question is why/how BOOTSTRAP_MODOBJS can become empty sometimes.

    BOOTSTRAP_MODOBJS variable:

    BOOTSTRAP_MODOBJS=   Modules/atexitmodule.o  Modules/faulthandler.o  Modules/posixmodule.o  Modules/signalmodule.o  Modules/_tracemalloc.o  Modules/_suggestions.o  Modules/_datetimemodule.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/sre.o  Modules/_sysconfig.o  Modules/_threadmodule.o  Modules/timemodule.o  Modules/_typesmodule.o  Modules/_typingmodule.o  Modules/_weakref.o  Modules/_abc.o  Modules/_functoolsmodule.o  Modules/_localemodule.o  Modules/_opcode.o  Modules/_operator.o  Modules/_stat.o  Modules/symtablemodule.o  Modules/pwdmodule.o
    
  2. vstinner commented on Nov 5, 2026

    @vstinner
    MemberAuthor

    Example of the issue on FreeBSD. I updated my Git checkout and ran make, but then the BOOTSTRAP_MODOBJS variable was empty.

    vstinner@freebsd$ git pull
    vstinner@freebsd$ gmake
    ./configure '--cache-file=../python-config.cache' '--with-pydebug' 'OPT=-O0' '--disable-gil' 'CFLAGS='
    (...)
    config.status: creating Makefile.pre
    (...)
    configure: creating Makefile
    (...)
    
    vstinner@freebsd$ grep ^BOOTSTRAP_MODOBJS Makefile
    BOOTSTRAP_MODOBJS=
    

    If I modify Makefile.pre.in, make rebuilds Makefile and this time, it generates properly the BOOTSTRAP_MODOBJS variable:

    vstinner@freebsd$ touch Makefile.pre.in 
    
    vstinner@freebsd$ LANG= LC_ALL= gmake Objects/object.o
    CONFIG_FILES=Makefile.pre CONFIG_HEADERS= ./config.status
    config.status: creating Makefile.pre
    gmake -f Makefile.pre Makefile
    gmake[1]: Entering directory '/home/vstinner/python/main'
    ./Modules/makesetup -c ./Modules/config.c.in \
    			-s Modules \
    			Modules/Setup.local \
    			Modules/Setup.stdlib \
    			-b Modules/Setup.bootstrap \
    			./Modules/Setup
    The Makefile was updated, you may need to re-run make.
    gmake[1]: Leaving directory '/home/vstinner/python/main'
    ./Modules/makesetup -c ./Modules/config.c.in \
    			-s Modules \
    			Modules/Setup.local \
    			Modules/Setup.stdlib \
    			-b Modules/Setup.bootstrap \
    			./Modules/Setup
    The Makefile was updated, you may need to re-run make.
    gmake: 'Objects/object.o' is up to date.
    
    vstinner@freebsd$ grep ^BOOTSTRAP_MODOBJS Makefile
    BOOTSTRAP_MODOBJS=   Modules/atexitmodule.o  Modules/faulthandler.o  (...)
    
  3. self-assigned this
    on Nov 10, 2026
  4. itamaro commented on Nov 10, 2026

    @itamaro
    Contributor

    oops, sorry about that, it's a side effect of my gh-160537!

    thanks @aozmekik for fixing it!

  5. added a commit that references this issue on Nov 10, 2026
  6. vstinner commented on Nov 10, 2026

    @vstinner
    MemberAuthor

    Oh it's ok, I only saw the issue a few times in the last 10 days. Great that it has been fixed!

  7. added a commit that references this issue on Nov 12, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

buildThe build process and cross-buildtype-bugAn unexpected behavior, bug, or error

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions