bug#14474: Using glib's g_file_monitor_file and g_file_monitor_directory

bug#14474: Using glib's g_file_monitor_file and g_file_monitor_directory

[Paul Eggert]

On 05/28/2013 11:12 PM, Michael Albinus wrote:
> DBUS_SESSION_BUS_ADDRESS="unix:path=/dev/null" seems to be reliable.

OK, thanks, I committed a patch along those lines to the trunk
(bzr 112795); please give it a try.

bug#14474: Using glib's g_file_monitor_file and g_file_monitor_directory

[Michael Albinus]

Paul Eggert <eggert@cs.ucla.edu> writes:

Hi Paul,

> On 05/28/2013 11:12 PM, Michael Albinus wrote:
>> DBUS_SESSION_BUS_ADDRESS="unix:path=/dev/null" seems to be reliable.
>
> OK, thanks, I committed a patch along those lines to the trunk
> (bzr 112795); please give it a try.

You've put the lines inside #ifdef USE_GTK. However, glib could be
linked to Emacs without using gtk. We might need another check, like
#ifdef USE_GLIB. This would be useful in other places, too.

Furthermore, I'd wrap these lines with #ifdef HAVE_DBUS. If Emacs is
compiled without D-Bus support, external processes could still use
D-Bus via autolaunch. You suppress this possibility by overwriting
$DBUS_SESSION_BUS_ADDRESS.

Best regards, Michael.

bug#14474: Using glib's g_file_monitor_file and g_file_monitor_directory

[Paul Eggert]

On 05/31/13 11:39, Michael Albinus wrote:

> You've put the lines inside #ifdef USE_GTK. However, glib could be
> linked to Emacs without using gtk. We might need another check, like
> #ifdef USE_GLIB. This would be useful in other places, too.

That might make sense in the long run, if Emacs is built that way.
But as I understand it, currently Emacs links to glib only when
using Gtk, so the workaround is OK under the current framework.

> Furthermore, I'd wrap these lines with #ifdef HAVE_DBUS. If Emacs is
> compiled without D-Bus support, external processes could still use
> D-Bus via autolaunch. You suppress this possibility by overwriting
> $DBUS_SESSION_BUS_ADDRESS.

Can't the Emacs process autolaunch D-bus subprocesses even when it
HAVE_DBUS is not defined?  That is, some other toolkit that Emacs links to,
could autolaunch D-Bus.  So it wouldn't be safe to wrap the lines
with #ifdef HAVE_DBUS.

I do see the problem that you mention.  I think a better fix, though,
is to redo Emacs to use the glib child watcher code -- that will fix
the problem that you mention, along with the problem for Emacs itself.

bug#14474: Using glib's g_file_monitor_file and g_file_monitor_directory

[Michael Albinus]

Paul Eggert <eggert@cs.ucla.edu> writes:

>> You've put the lines inside #ifdef USE_GTK. However, glib could be
>> linked to Emacs without using gtk. We might need another check, like
>> #ifdef USE_GLIB. This would be useful in other places, too.
>
> That might make sense in the long run, if Emacs is built that way.
> But as I understand it, currently Emacs links to glib only when
> using Gtk, so the workaround is OK under the current framework.

The g_file_monitor patch I've shown uses glib. No gtk. And the same is
true for gconf and gsettings, IIUC.

> I do see the problem that you mention.  I think a better fix, though,
> is to redo Emacs to use the glib child watcher code -- that will fix
> the problem that you mention, along with the problem for Emacs itself.

Yep, this might be better.

Best regards, Michael.

bug#14474: Using glib's g_file_monitor_file and g_file_monitor_directory

[Jan Djärv]

Hello.

31 maj 2013 kl. 21:24 skrev Paul Eggert <eggert@cs.ucla.edu>:

> On 05/31/13 11:39, Michael Albinus wrote:
> 
>> You've put the lines inside #ifdef USE_GTK. However, glib could be
>> linked to Emacs without using gtk. We might need another check, like
>> #ifdef USE_GLIB. This would be useful in other places, too.
> 
> That might make sense in the long run, if Emacs is built that way.
> But as I understand it, currently Emacs links to glib only when
> using Gtk, so the workaround is OK under the current framework.
> 

This is false. Emacs links with glib if one of GConf, Gsettings or rsvg is linked in. They are independent of Gtk.

      Jan D.