Closed
Description
Bug report
Bug description:
test_embed started failing for me recently on free threaded builds on both M1 Mac and Ubuntu (Intel). git bisect narrowed it down to a suspect commit. @colesbury this is probably for you.
Unfortunately, on my Mac I get this commit as the culprit:
% git bisect bad 5f1997896d9c3ecf92e9863177c452b468a6a2c8 is the first bad commit commit 5f1997896d9c3ecf92e9863177c452b468a6a2c8 Author: Sam Gross <colesbury@gmail.com> Date: Tue Jan 23 13:05:15 2024 -0500 gh-112984: Fix link error on free-threaded Windows build (GH-114455) The test_peg_generator test tried to link the python313_d.lib library, which failed because the library is now named python313t_d.lib. The underlying problem is that the "compiler" attribute was not set when we call get_libraries() from distutils. Tools/peg_generator/pegen/build.py | 3 +++ 1 file changed, 3 insertions(+)while in my Linux environment it says this:
% git bisect bad b331381485c1965d1c88b7aee7ae9604aca05758 is the first bad commit commit b331381485c1965d1c88b7aee7ae9604aca05758 Author: Sam Gross <colesbury@gmail.com> Date: Tue Jan 16 16:42:15 2024 -0500 gh-112529: Track if debug allocator is used as underlying allocator (#113747) * gh-112529: Track if debug allocator is used as underlying allocator The GC implementation for free-threaded builds will need to accurately detect if the debug allocator is used because it affects the offset of the Python object from the beginning of the memory allocation. The current implementation of `_PyMem_DebugEnabled` only considers if the debug allocator is the outer-most allocator; it doesn't handle the case of "hooks" like tracemalloc being used on top of the debug allocator. This change enables more accurate detection of the debug allocator by tracking when debug hooks are enabled. * Simplify _PyMem_DebugEnabled Include/internal/pycore_pymem.h | 3 +++ Include/internal/pycore_pymem_init.h | 2 ++ Include/internal/pycore_runtime_init.h | 1 + Objects/obmalloc.c | 21 +++++++++++++++------ 4 files changed, 21 insertions(+), 6 deletions(-) (I backed up farther on my Linux box to find a good commit.) 16 January was a long time ago. My inclination is that the later commit is the culprit.
I then tried with the GIL enabled on main. Still an error. Hmmm... I went through the bisect process again on my Mac and landed on this as the putative culprit commit:
% git bisect good 441affc9e7f419ef0b68f734505fa2f79fe653c7 is the first bad commit commit 441affc9e7f419ef0b68f734505fa2f79fe653c7 Author: Sam Gross <colesbury@gmail.com> Date: Tue Jan 23 13:08:23 2024 -0500 gh-111964: Implement stop-the-world pauses (gh-112471) The `--disable-gil` builds occasionally need to pause all but one thread. Some examples include: * Cyclic garbage collection, where this is often called a "stop the world event" * Before calling `fork()`, to ensure a consistent state for internal data structures * During interpreter shutdown, to ensure that daemon threads aren't accessing Python objects This adds the following functions to implement global and per-interpreter pauses: * `_PyEval_StopTheWorldAll()` and `_PyEval_StartTheWorldAll()` (for the global runtime) * `_PyEval_StopTheWorld()` and `_PyEval_StartTheWorld()` (per-interpreter) (The function names may change.) These functions are no-ops outside of the `--disable-gil` build. Include/cpython/pystate.h | 2 +- Include/internal/pycore_ceval.h | 1 + Include/internal/pycore_interp.h | 17 +++ Include/internal/pycore_llist.h | 3 +- Include/internal/pycore_pystate.h | 51 +++++-- Include/internal/pycore_runtime.h | 7 + Include/internal/pycore_runtime_init.h | 3 + Include/pymacro.h | 3 + Python/ceval_gil.c | 9 ++ Python/pystate.c | 269 +++++++++++++++++++++++++++++++-- 10 files changed, 336 insertions(+), 29 deletions(-) CPython versions tested on:
CPython main branch
Operating systems tested on:
Linux, macOS