Re: do we really need xrdb?

2005-10-31 Thread Lorenzo Colitti

James Henstridge wrote:

The question was: if gcc has already been paged in (e.g. by a prior xrdb
call), does this change provide any noticeable benefit?


No, it doesn't.


The previous discussions seemed to indicate that the largest performance
wins would come from not paging gcc in at all, which will only happen if
either (a) all xrdb calls use -nocpp or (b) xrdb uses a different C
preprocessor.


That's right. xrdb is already called by the X session startup scripts, 
and that's where you take the startup time penalty (because you have to 
page in cpp). So even if you removed xrdb completely from GNOME startup, 
you wouldn't gain anything.


The right approach is to tell xrdb to use a lightweight cpp, such as 
mcpp or tcpp (although I don't know if tcc has an option to preprocess 
only without compiling).


There is a patch in freedesktop.org bug 4325 to use other preprocessors 
by default if they're available at runtime, but the only feedback I got 
on that was that xrdb should detect the preprocessor at compile time 
instead of at runtime, which seems worse to me since if you then 
uninstall that processor your xrdb breaks completely.


https://bugs.freedesktop.org/show_bug.cgi?id=4325

Any suggestions?


Cheers,
Lorenzo
___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-18 Thread Rodrigo Moya
On Mon, 2005-10-17 at 11:54 -0500, Federico Mena Quintero wrote:
 On Mon, 2005-10-17 at 17:32 +0100, Ross Burton wrote:
 
  Using a smaller CPP and delaying xrdb's execution until everything else
  is started is the right solution here.
 
 You need to run xrdb before gnome-session (re)starts any Motif apps, for
 example.
 
I've just sent a patch
(http://mail.gnome.org/archives/gnomecc-list/2005-October/msg00024.html)
for using -nocpp where possible.

I was also trying to run the other xrdb command (which reads
~/.Xresources and all other appropriate files) in a 10/15 seconds
timeout, but since you say we need to run it before any motif app is
started (which can be started as part of the session, so we need to run
xrdb ASAP).

Anyway, gdm already runs xrdb with the user's .Xresources and other
system-wide files, from what I have seen. So, maybe we don't need that
other xrdb command?
-- 
Rodrigo Moya [EMAIL PROTECTED]

___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-18 Thread Ross Burton
On Tue, 2005-10-18 at 13:16 +0200, Rodrigo Moya wrote:
 I've just sent a patch
 (http://mail.gnome.org/archives/gnomecc-list/2005-October/msg00024.html)
 for using -nocpp where possible.

Cool, that will help.

 I was also trying to run the other xrdb command (which reads
 ~/.Xresources and all other appropriate files) in a 10/15 seconds
 timeout, but since you say we need to run it before any motif app is
 started (which can be started as part of the session, so we need to run
 xrdb ASAP).

As was said, just running it between the core desktop and the user tasks
would be best, but that isn't possible with the current session manager
as I understand it.

 Anyway, gdm already runs xrdb with the user's .Xresources and other
 system-wide files, from what I have seen. So, maybe we don't need that
 other xrdb command?

gdm doesn't generate defines from the GTK+ theme and read the .ad files
in /etc/gnome/config and ~/.gnome2/xrdb, it only loads ~/.Xresources and
the static files in /etc/X11/.

Ross
-- 
Ross Burton mail: [EMAIL PROTECTED]
  jabber: [EMAIL PROTECTED]
 www: http://www.burtonini.com./
 PGP Fingerprint: 1A21 F5B0 D8D0 CFE3 81D4 E25A 2D09 E447 D0B4 33DF



signature.asc
Description: This is a digitally signed message part
___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list

Re: do we really need xrdb?

2005-10-18 Thread James Henstridge
Rodrigo Moya wrote:

I've just sent a patch
(http://mail.gnome.org/archives/gnomecc-list/2005-October/msg00024.html)
for using -nocpp where possible.
  

Given the problems mentioned before with parsing Xresources without a
cpp, is this actually worth it?

From previous messages, I was under the impression that the main
slowdown came from paging gcc into memory.  Changing some, but not all,
invocations to not run the C preprocessor wouldn't affect that.

A patch to change all xrdb invocations to xrdb -cpp
/path/to/a/small/cpp would probably provide better results without the
compatibility problems.

James.
___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-18 Thread Ross Burton
On Tue, 2005-10-18 at 20:01 +0800, James Henstridge wrote:
 I've just sent a patch
 (http://mail.gnome.org/archives/gnomecc-list/2005-October/msg00024.html)
 for using -nocpp where possible.
   
 Given the problems mentioned before with parsing Xresources without a
 cpp, is this actually worth it?

The locations where CPP was removed don't need to do any pre-processing,
as they merge in a generated string and not .Xresources.

Ross
-- 
Ross Burton mail: [EMAIL PROTECTED]
  jabber: [EMAIL PROTECTED]
 www: http://www.burtonini.com./
 PGP Fingerprint: 1A21 F5B0 D8D0 CFE3 81D4 E25A 2D09 E447 D0B4 33DF


___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-18 Thread James Henstridge
Ross Burton wrote:

On Tue, 2005-10-18 at 20:01 +0800, James Henstridge wrote:
  

I've just sent a patch
(http://mail.gnome.org/archives/gnomecc-list/2005-October/msg00024.html)
for using -nocpp where possible.
 
  

Given the problems mentioned before with parsing Xresources without a
cpp, is this actually worth it?



The locations where CPP was removed don't need to do any pre-processing,
as they merge in a generated string and not .Xresources.
  

The question was: if gcc has already been paged in (e.g. by a prior xrdb
call), does this change provide any noticeable benefit?

The previous discussions seemed to indicate that the largest performance
wins would come from not paging gcc in at all, which will only happen if
either (a) all xrdb calls use -nocpp or (b) xrdb uses a different C
preprocessor.

James.
___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-18 Thread Xavier Bestel
On Mon, 2005-10-17 at 22:31, Alan Cox wrote:
 On Llu, 2005-10-17 at 13:18 +0200, Rodrigo Moya wrote:
  Hi
  
  One of the best improvements we've seen while working on speeding up the
  GNOME startup time has been the removal of the xsettings thing, via xrdb
  -merge execution.
 
 Make xrdb use decus cpp and it takes basically zero time. Trivial fix
 and doesn't break anything. Decus CPP is pd

And how about just inserting suitable calls to XChangeProperty() 
friends directly in gnome-session ? Beats a sytem() anytime.

Xav


___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


do we really need xrdb?

2005-10-17 Thread Rodrigo Moya
Hi

One of the best improvements we've seen while working on speeding up the
GNOME startup time has been the removal of the xsettings thing, via xrdb
-merge execution.

Ben Kahn came up with a list of apps that still need this:

acroread 5
Emacs
Anything written in Tk (aMSN, crossover office, much corporate custom
software)
Anything written in Motif (slowly being replaced, but still a lot of
software)
Anything written in XForms (oddly, a lot of scientific software...)
Apps that ship with the X server: xterm, xedit, xclock, etc.

Anything else?

So, do we really want to keep around the xrdb thing in
gnome-settings-daemon?
-- 
Rodrigo Moya [EMAIL PROTECTED]

___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-17 Thread Jeff Waugh
quote who=Rodrigo Moya

 So, do we really want to keep around the xrdb thing in
 gnome-settings-daemon?

How about an off-by-default GConf key? Lots of workstation deployments I
know about really benefit from the xrdb stuff (heaps of motif and xforms).

- Jeff

-- 
Ubuntu USA  Europe Tour: Oct-Nov 2005http://wiki.ubuntu.com/3BT
 
   Instead you're doing circle jerks with the Care Bears of Censorship.
- Siduri on Slashdot
___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-17 Thread Ross Burton
On Mon, 2005-10-17 at 13:18 +0200, Rodrigo Moya wrote:
 acroread 5
 Emacs
 Anything written in Tk (aMSN, crossover office, much corporate custom
 software)
 Anything written in Motif (slowly being replaced, but still a lot of
 software)
 Anything written in XForms (oddly, a lot of scientific software...)
 Apps that ship with the X server: xterm, xedit, xclock, etc.
 
 Anything else?

All X software which isn't written in GTK+ or Qt.

 So, do we really want to keep around the xrdb thing in
 gnome-settings-daemon?

Yes, really.  Lots of people use Emacs and this is responsible for
making it not stand out like a sore thumb.  How about a check box in
Prefences-Theme, [X] Apply colour theme to legacy applications?

Ross
-- 
Ross Burton mail: [EMAIL PROTECTED]
  jabber: [EMAIL PROTECTED]
 www: http://www.burtonini.com./
 PGP Fingerprint: 1A21 F5B0 D8D0 CFE3 81D4 E25A 2D09 E447 D0B4 33DF


___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-17 Thread Ross Burton
On Mon, 2005-10-17 at 13:18 +0200, Rodrigo Moya wrote:
 One of the best improvements we've seen while working on speeding up the
 GNOME startup time has been the removal of the xsettings thing, via xrdb
 -merge execution.

Oh, as the settings can be applied at any time in the startup, has
anyone tried moving the xrdb invocation to be the last thing, after the
Panel and Nautilus have been started?

Ross
-- 
Ross Burton mail: [EMAIL PROTECTED]
  jabber: [EMAIL PROTECTED]
 www: http://www.burtonini.com./
 PGP Fingerprint: 1A21 F5B0 D8D0 CFE3 81D4 E25A 2D09 E447 D0B4 33DF


___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-17 Thread Rodrigo Moya
On Mon, 2005-10-17 at 13:16 +0100, Ross Burton wrote:
 On Mon, 2005-10-17 at 13:18 +0200, Rodrigo Moya wrote:
  One of the best improvements we've seen while working on speeding up the
  GNOME startup time has been the removal of the xsettings thing, via xrdb
  -merge execution.
 
 Oh, as the settings can be applied at any time in the startup, has
 anyone tried moving the xrdb invocation to be the last thing, after the
 Panel and Nautilus have been started?
 
will try this next,although I first wanted to consolidate all xrdb calls
(3 places) into one only.
-- 
Rodrigo Moya [EMAIL PROTECTED]

___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-17 Thread Shaun McCance
On Mon, 2005-10-17 at 13:41 +0200, Jeff Waugh wrote:
 quote who=Rodrigo Moya
 
  So, do we really want to keep around the xrdb thing in
  gnome-settings-daemon?
 
 How about an off-by-default GConf key? Lots of workstation deployments I
 know about really benefit from the xrdb stuff (heaps of motif and xforms).

Why do we always assume large-scale workstation deployments
with sysadmins who have time to dig through GConf keys and
set different defaults?

Speaking as somebody employed by a Motif-app-producing ISV,
(but not speaking on behalf of said ISV, blah blah blah),
I'm telling you that all the people out there using Motif
and Tk software aren't doing so because they're ubergeeks
who like to dig through gconf-editor.  These are average
people, the sorts of folks who have to have their children
help them with email.

--
Shaun



___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-17 Thread Sebastien Bacher
Le lundi 17 octobre 2005 à 09:32 -0600, Elijah Newren a écrit :

 we could achieve the same benefit by just calling xrdb with the -nocpp

We tried that before Ubuntu 5.10 and had some issues:
https://bugzilla.ubuntu.com/show_bug.cgi?id=14268

Cheers,

Sebastien Bacher


___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-17 Thread Federico Mena Quintero
On Mon, 2005-10-17 at 17:32 +0100, Ross Burton wrote:

 Using a smaller CPP and delaying xrdb's execution until everything else
 is started is the right solution here.

You need to run xrdb before gnome-session (re)starts any Motif apps, for
example.

  Federico

___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list


Re: do we really need xrdb?

2005-10-17 Thread Alan Cox
On Llu, 2005-10-17 at 13:18 +0200, Rodrigo Moya wrote:
 Hi
 
 One of the best improvements we've seen while working on speeding up the
 GNOME startup time has been the removal of the xsettings thing, via xrdb
 -merge execution.

Make xrdb use decus cpp and it takes basically zero time. Trivial fix
and doesn't break anything. Decus CPP is pd

___
desktop-devel-list mailing list
desktop-devel-list@gnome.org
http://mail.gnome.org/mailman/listinfo/desktop-devel-list