Fabian Keil found the reason for a SIGBUS:
In the "gpg-agent --daemon" case, main() calls pth_kill()
after the client has been forked, so when es_deinit() is
called on exit, acquiring the estream_list_lock seems to
cause pth to dereference a pointer located in a memory
region that has previously been free()'d.
My approach to fix it is different than his suggestion. It should
allow to continue all estream operations after a pth_kill except for
restarting pth.
char * vs. unsigned char * warnings. The GNU coding standards used to
say that these mismatches are okay and better than a bunch of casts.
Obviously this has changed now.
1.9.
* agent.h (agent_exit): Add JNLIB_GCC_A_NR to indicate that this
function won't return.
* gpg-agent.c (check_for_running_agent): Initialize pid to a
default value if not needed.
* command-ssh.c: Removed stdint.h. s/byte_t/unsigned char/,
s/uint32/u32/ becuase that is what we have always used in GnuPG.
(ssh_request_specs): Moved to top of file.
(ssh_key_types): Ditto.
(make_cstring): Ditto.
(data_sign): Don't use a variable for the passphrase prompt, make
it translatable.
(ssh_request_process):
* findkey.c (modify_description): Renamed arguments for clarity,
polished documentation. Make comment a C-string. Fixed case of
DESCRIPTION being just "%".
(agent_key_from_file): Make sure comment string to a C-string.
* gpg-agent.c (create_socket_name): Cleanup the implemntation, use
DIMof, agent_exit, removed superflous args and return the
allocated string as value. Documented. Changed callers.
(create_server_socket): Cleanups similar to above. Changed callers.
(cleanup_do): Renamed to ..
(remove_socket): .. this. Changed caller.
(handle_connections): The signals are to be handled in the select
and not in the accept. Test all FDs after returning from a
select. Remove the event tests from the accept calls. The select
already assured that the accept won't block.
