Debian Patches
Status for babeltrace/1.5.11-6
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| fix-clean-bison-files.patch | Add the bison generated files to maintainer clean files so they are not deleted on a make distclean. |
no | ||||
| 0001-Fix-Hurd-compat-for-PATH_MAX.patch | [PATCH babeltrace] Fix: Hurd compat for PATH_MAX | Michael Jeanson <mjeanson@efficios.com> | no | 2016-07-08 | ||
| cross.patch | Fix the elfutils check to work with cross-compilation. | no | ||||
| fix-pkgconfig-private-requires.patch | Remove superfluous pkgconfig private requires. | no | ||||
| Fix-handle-NULL-list-pointer-to-avoid-Python-segfaul.patch | [PATCH 1/2] Fix: handle NULL list pointer to avoid Python segfault Before a change in SWIG 4.3.0 [0], `AppendOutput(Py_None, count_obj)` returned just `count_obj`, an `int`, when an error happens. That value is garbage because `*len` (`== count_obj`) is not initialized by `bt_ctf_get_field_list()` & friends when they return an error (e.g., when called for an event scope that does not exist on the event, like `EVENT_CONTEXT` on an event without per-event context). However, it's not a `list`, so `if not isinstance(ret, list): return` in `EventDeclaration.fields_scope()` (and other location) prevents it from resulting in a crash. After that change, SWIG returns `[None, count_obj]` (still with garbage `count_obj`) in this case since it's a non-void return (see the SWIG commit). This is unfortunately a `list`, so it keeps going, iterates `range(garbage_count)` and dereferences a `NULL` list pointer. ``` Current thread 0x000073d613434200 [python3] (most recent call first): File "/usr/lib/python3/dist-packages/babeltrace/babeltrace.py", line 83 in _bt_python_field_one_from_list File "/usr/lib/python3/dist-packages/babeltrace/babeltrace.py", line 839 in _field_list_with_scope File "/usr/lib/python3/dist-packages/babeltrace/babeltrace.py", line 719 in field_list_with_scope File "/usr/lib/python3/dist-packages/babeltrace/babeltrace.py", line 792 in keys ``` See the new Python regression test. Fix this on the Python caller side: `count` shouldn't be used if the call failed in the first place. If `[list_ptr, count]` is returned (always the case in SWIG >=4.3.0), make sure `list_ptr is not None` before interpreting `count`. The existing checks for `isinstance(ret, list)` handle SWIG <4.3.0 returning just `count` when the underlying call fails. I used Claude Opus 4.7 to identify and fix this issue. I've verified that this actually resolves the issue. [0] https://github.com/swig/swig/commit/cd39cf132c96a0887be07c826b80804d7677a701 |
Christophe Bedard <bedard.christophe@gmail.com> | no | 2026-04-29 | ||
| Print-a-clear-error-when-reading-a-CTF-2-trace.patch | [PATCH 2/2] Print a clear error when reading a CTF 2 trace A CTF 2 metadata stream starts with an ASCII RS (0x1e) byte, which Babeltrace 1.x can't parse as TSDL. Detect it and fail with a message pointing to Babeltrace 2.1+ instead of confusing syntax errors. |
Philippe Proulx <eeppeliteloop@gmail.com> | no | 2026-06-04 |