Hi,
I'm running rawhide these days again and during yum update I quite often see messages like
Updating : gedit ##################### [173/382] gconfd-2: no process killed Updating : deskbar-applet ##################### [174/382] gconfd-2: Kein Prozess abgebrochen Updating : gnome-terminal ##################### [175/382]
Looking closer at the spec and at the Gconf session on http://fedoraproject.org/wiki/PackagingDrafts/ScriptletSnippets it seems they run
killall -HUP gconfd-2 || :
to make sure any running gconfd-2 pick up newly installed schemes. But well, as you can see from above output there might be cases when no gconfd-2 is running, thus killall will print a warning.
I think we should avoid such useless warnings -- thus it would seem better to me to use killall with "--quiet" or use
killall -HUP gconfd-2 2> /dev/null || :
instead. What do you guys think?
Cu knurd
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Thorsten Leemhuis wrote:
Hi,
I'm running rawhide these days again and during yum update I quite often see messages like
Updating : gedit ##################### [173/382] gconfd-2: no process killed Updating : deskbar-applet ##################### [174/382] gconfd-2: Kein Prozess abgebrochen Updating : gnome-terminal ##################### [175/382]
Looking closer at the spec and at the Gconf session on http://fedoraproject.org/wiki/PackagingDrafts/ScriptletSnippets it seems they run
killall -HUP gconfd-2 || :
to make sure any running gconfd-2 pick up newly installed schemes. But well, as you can see from above output there might be cases when no gconfd-2 is running, thus killall will print a warning.
I think we should avoid such useless warnings -- thus it would seem better to me to use killall with "--quiet" or use
killall -HUP gconfd-2 2> /dev/null || :
instead. What do you guys think?
This looks like a good change. Anyone object to me going ahead and changing it?
- -Toshio
"TK" == Toshio Kuratomi a.badger@gmail.com writes:
TK> This looks like a good change. Anyone object to me going ahead TK> and changing it?
No objection here; removing inconsistencies from ScriptletSnippets is always a good thing. Using --quiet is probably the best choice if it's supported as far back as we need it to be.
- J<
On Fri, 2007-09-21 at 23:08 -0500, Jason L Tibbitts III wrote:
"TK" == Toshio Kuratomi a.badger@gmail.com writes:
TK> This looks like a good change. Anyone object to me going ahead TK> and changing it?
No objection here; removing inconsistencies from ScriptletSnippets is always a good thing. Using --quiet is probably the best choice if it's supported as far back as we need it to be.
The kill is completely redundant nowadays, since we have:
--- GConf-2.18.0.1/gconf/gconftool.c.reload 2007-03-02 17:10:13.000000000 -0500 +++ GConf-2.18.0.1/gconf/gconftool.c 2007-03-13 02:21:29.000000000 -0400 @@ -3780,6 +3780,8 @@ ++args; }
+ g_spawn_command_line_sync ("/usr/bin/killall -q -TERM " GCONF_SERVERDIR "/" GCONFD, NULL, NULL, NULL, NULL); + retval |= do_sync (conf); return retval; }
On Sat, 2007-09-22 at 00:23 -0400, Matthias Clasen wrote:
On Fri, 2007-09-21 at 23:08 -0500, Jason L Tibbitts III wrote:
> "TK" == Toshio Kuratomi a.badger@gmail.com writes:
TK> This looks like a good change. Anyone object to me going ahead TK> and changing it?
No objection here; removing inconsistencies from ScriptletSnippets is always a good thing. Using --quiet is probably the best choice if it's supported as far back as we need it to be.
The kill is completely redundant nowadays, since we have:
--- GConf-2.18.0.1/gconf/gconftool.c.reload 2007-03-02 17:10:13.000000000 -0500 +++ GConf-2.18.0.1/gconf/gconftool.c 2007-03-13 02:21:29.000000000 -0400 @@ -3780,6 +3780,8 @@ ++args; }
- g_spawn_command_line_sync ("/usr/bin/killall -q -TERM "
GCONF_SERVERDIR "/" GCONFD, NULL, NULL, NULL, NULL);
- retval |= do_sync (conf); return retval;
}
Then gconftool should probably better Requires: /usr/bin/killall
Ralf
On Fri, 2007-09-21 at 23:49 -0500, Jason L Tibbitts III wrote:
"MC" == Matthias Clasen mclasen@redhat.com writes:
MC> The kill is completely redundant nowadays, since we have:
Does nowadays extend back to RHEL4? If not, we need to document the point at which it becomes unnecessary.
* Wed Feb 1 2006 Christopher Aillon caillon@redhat.com 2.13.5-2 - Add patch from Mandriva to reload GConf2 every time a schema is added or removed (solves bug 173869)
"MC" == Matthias Clasen mclasen@redhat.com writes:
MC> * Wed Feb 1 2006 Christopher Aillon caillon@redhat.com 2.13.5-2
So that's a definite no for RHEL4 (looks like 2.8.1), but yes for FC5 and later. Which means Fedora doesn't care but EPEL does, and I think we still try to keep our documents valid for EPEL.
- J<
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Matthias Clasen wrote:
On Fri, 2007-09-21 at 23:49 -0500, Jason L Tibbitts III wrote:
> "MC" == Matthias Clasen mclasen@redhat.com writes:
MC> The kill is completely redundant nowadays, since we have:
Does nowadays extend back to RHEL4? If not, we need to document the point at which it becomes unnecessary.
- Wed Feb 1 2006 Christopher Aillon caillon@redhat.com 2.13.5-2
- Add patch from Mandriva to reload GConf2 every time a schema is added or removed (solves bug 173869)
Actually, it wasn't fixed until this release::
* Mon Nov 6 2006 Ray Strode <rstrode redhat com> - 2.14.0-2.fc5 - - remove path name from hacky killall workaround, so that the work around works more reliably (bug 173869, comment 14)
Still fixed in FC5 and RHEL5, though, so I'll update the documentation to reflect the fact that EPEL4 is the only thing we're building for that's still affected.
- -Toshio
On 22/09/2007, Ville Skyttä ville.skytta@iki.fi wrote:
On Saturday 22 September 2007, Matthias Clasen wrote:
- g_spawn_command_line_sync ("/usr/bin/killall -q -TERM "
GCONF_SERVERDIR "/" GCONFD, NULL, NULL, NULL, NULL);
Hm, ScriptletSnippets advices to use -HUP; is -TERM instead of it intentional?
The next time an application uses GConf, the daemon is automatically respawned anyway, right? So there should be no difference between -HUP and -TERM.
On Monday 24 September 2007, Michel Salim wrote:
On 22/09/2007, Ville Skyttä ville.skytta@iki.fi wrote:
On Saturday 22 September 2007, Matthias Clasen wrote:
- g_spawn_command_line_sync ("/usr/bin/killall -q -TERM "
GCONF_SERVERDIR "/" GCONFD, NULL, NULL, NULL, NULL);
Hm, ScriptletSnippets advices to use -HUP; is -TERM instead of it intentional?
The next time an application uses GConf, the daemon is automatically respawned anyway, right?
I don't know, that's why I asked ;)
So there should be no difference between -HUP and -TERM.
Actually, if the behaviour you described is how it works, -TERM may be a better idea than -HUP performance-wise - no need to repeatedly re-read things eg. during a rpm transaction which installs many packages that have something to do with GConf.
packaging@lists.fedoraproject.org
-
Jason L Tibbitts III -
Matthias Clasen -
Michel Salim -
Ralf Corsepius -
Thorsten Leemhuis -
Toshio Kuratomi -
Ville Skyttä