Re: Odd gdb problem

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Perhaps I should just try to bisect the current trunk -- this
> started happening this past week.

Done.  And the culprit found at least makes some kind of sense as
something that could break gdb.

Philipp, after this change, I can no longer debug Emacs with gdb on
Debian/bookworm.  (It works fine with older gdb versions.)

Do you have any idea what in particular in this change might make gdb
just exit with:

Program terminated with signal SIGTRAP, Trace/breakpoint trap.


a60053f8368e058229721f1bf1567c2b1676b239 is the first bad commit
commit a60053f8368e058229721f1bf1567c2b1676b239
Author: Philipp Stephani <phst@google.com>
Date:   Wed Dec 30 14:42:01 2020 +0100

    Use posix_spawn if possible.
    
    posix_spawn is less error-prone than vfork + execve, and can make
    better use of system-specific enhancements like 'clone' on Linux.  Use
    it if we don't need to configure a pseudoterminal.
    
    * configure.ac (HAVE_SPAWN_H, HAVE_POSIX_SPAWN)
    (HAVE_POSIX_SPAWN_FILE_ACTIONS_ADDCHDIR)
    (HAVE_POSIX_SPAWN_FILE_ACTIONS_ADDCHDIR_NP)
    (HAVE_POSIX_SPAWNATTR_SETFLAGS, HAVE_DECL_POSIX_SPAWN_SETSID): New
    configuration variables.
    * src/callproc.c (USABLE_POSIX_SPAWN): New configuration macro.
    (emacs_posix_spawn_init_actions)
    (emacs_posix_spawn_init_attributes, emacs_posix_spawn_init): New
    helper functions.
    (emacs_spawn): Use posix_spawn if possible.

 configure.ac   |  17 ++++++
 src/callproc.c | 189 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 205 insertions(+), 1 deletion(-)



-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no