This should mean we don't get invalid fds in the main loop.
The BSD (kqueue) and Windows code paths are untested, but follow the same
patterns as the tested Linux/generic Unix versions.
DBusTransportSocket was already OK (it called free_watches() before
_dbus_close_socket, and that did the remove, invalidate, unref dance).
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=33336
Reviewed-by: Will Thompson <will.thompson@collabora.co.uk>
Reviewed-by: Thiago Macieira <thiago@kde.org>
num_fds is the number of elements of dirs currently in use. This bug
meant that encountering a previously un-watched directory would cause j
to increment forever, and so dirs[j] would eventually segfault.
(I've checked the corresponding code for inotify, and it's correct. I
wonder if some of the duplication could be eliminated.)
Thanks to Pablo Martí Gamboa <pmarti@warp.es> for reporting this issue!
It's not expected to have to manually SIGHUP the bus after installing
a new .service file. Since our directory monitoring is already set
up to queue a full reload which includes service activation, simply
monitor the servicedirs too.
https://bugs.freedesktop.org/show_bug.cgi?id=23846
Substantially based on a patch by Matthias Clasen <mclasen@redhat.com>
kqueue implementation by Joe Marcus Clarke <marcus@freebsd.org>
Previously, when we detected a configuration change (which included
the set of config directories to monitor for changes), we would
simply drop all watches, then readd them.
The problem with this is that it introduced a race condition where
we might not be watching one of the config directories for changes.
Rather than dropping and readding, change the OS-dependent monitoring
API to simply take a new set of directories to monitor. Implicit
in this is that the OS-specific layer needs to keep track of the
previously monitored set.
* dbus/dbus-sysdeps.c (_dbus_set_errno_to_zero)
(_dbus_get_is_errno_nonzero, _dbus_get_is_errno_eintr)
(_dbus_strerror_from_errno): family of functions to abstract
errno, though these are somewhat bogus (really we should make our
socket wrappers not use errno probably - the issue is that any
usage of errno that isn't socket-related probably is not
cross-platform, so should either be in a unix-only file that can
use errno directly, or is a bug - these general errno wrappers
hide issues of this nature in non-socket code, while
socket-specific API changes would not since sockets are allowed
cross-platform)
bus/dir-watch-kqueue.c (bus_watch_directory): Pass in a BusContext
instead of a void *. kqueue uses this to get the context's loop
while the other modules ignore the parameter. This allows us to
avoid platform conditionals
* bus/bus.c (process_config_postinit): Pass in the context to the
watch
* dbus/dbus-sysdeps.c:
* dbus/dbus-string.c:
s/_dbus_printf_length/_dbus_printf_string_upper_bound to comform with
GLib's function which does the same thing
* configure.in:
* bus/Makefile.am:
* bus/dir-watch-default.c:
* bus/dir-watch-dnotify.c:
* bus/dir-watch-kqueue.c:
Add kqueue directory watching for freebsd and split the directory
watching code into seperate files per method/arch
(patches from Timothy Redaelli <drizzt at gufi dotorg>)
