oprofile isn't new and a lot of developers use it. Drop the
EXPERIMENTAL.
Signed-off-by: Chris Wedgwood <[email protected]>
diff --git a/arch/i386/Kconfig b/arch/i386/Kconfig
index 5b1a7d4..2b8657d 100644
Index: linux-2.6/arch/i386/Kconfig
===================================================================
--- linux-2.6.orig/arch/i386/Kconfig 2006-05-19 19:35:35.586394755 -0700
+++ linux-2.6/arch/i386/Kconfig 2006-05-19 19:35:38.866552357 -0700
@@ -1053,7 +1053,6 @@
source "fs/Kconfig"
menu "Instrumentation Support"
- depends on EXPERIMENTAL
source "arch/i386/oprofile/Kconfig"
Index: linux-2.6/arch/i386/oprofile/Kconfig
===================================================================
--- linux-2.6.orig/arch/i386/oprofile/Kconfig 2006-05-19 19:35:32.644253387 -0700
+++ linux-2.6/arch/i386/oprofile/Kconfig 2006-05-19 19:35:38.866552357 -0700
@@ -1,12 +1,12 @@
config PROFILING
- bool "Profiling support (EXPERIMENTAL)"
+ bool "Profiling support"
help
Say Y here to enable the extended profiling support mechanisms used
by profilers such as OProfile.
config OPROFILE
- tristate "OProfile system profiling (EXPERIMENTAL)"
+ tristate "OProfile system profiling"
depends on PROFILING
help
OProfile is a profiling system capable of profiling the
On Fri, May 19, 2006 at 07:53:22PM -0700, Chris Wedgwood wrote:
> oprofile isn't new and a lot of developers use it. Drop the
> EXPERIMENTAL.
Some feedback claims this is not the case and that "it's unmaintained,
rarely well-tested, and may still need changes for some forthcoming
JVM stuff. and the tools aren't yet at 1.0".
Given that it should probably stay EXPERIMENTAL for now. I wonder if
left unmaintained for much longer if it should be potentially marked
BROKEN and/or scheduled for removal?
MAINTAINERS claims it is maintained, if that is not long the case I'm
happy post a trivial patch to change that. Or perhaps we can find a
new maintainer?
On Sun, May 21, 2006 at 12:49:15PM -0700, Chris Wedgwood wrote:
> On Fri, May 19, 2006 at 07:53:22PM -0700, Chris Wedgwood wrote:
>
> > oprofile isn't new and a lot of developers use it. Drop the
> > EXPERIMENTAL.
>
> Some feedback claims this is not the case and that "it's unmaintained,
> rarely well-tested, and may still need changes for some forthcoming
> JVM stuff. and the tools aren't yet at 1.0".
>
> Given that it should probably stay EXPERIMENTAL for now. I wonder if
> left unmaintained for much longer if it should be potentially marked
> BROKEN and/or scheduled for removal?
Why on earth would we want to remove a working feature ?
Just because it isn't getting patched every release doesn't mean
it's rotting, Oprofile is actually one of the few kernel features which I
don't recall a single regression in since 2.6.0
BROKEN is for stuff that doesn't compile, or is fundamentally flawed
beyond repair at the current time (For example, needs infrastructure
work to happen before it can work correctly).
Oprofile fits neither of those categories.
Dave
--
http://www.codemonkey.org.uk
On Sun, May 21, 2006 at 03:57:10PM -0400, Dave Jones wrote:
> Why on earth would we want to remove a working feature ?
We don't. I personally don't want to see it go, I use it myself.
> Just because it isn't getting patched every release doesn't mean
> it's rotting, Oprofile is actually one of the few kernel features
> which I don't recall a single regression in since 2.6.0
It works for me, still others have claimed it's unmaintained and has
issues that have not been address.
> BROKEN is for stuff that doesn't compile, or is fundamentally flawed
> beyond repair at the current time (For example, needs infrastructure
> work to happen before it can work correctly).
Sure, based on my personal experience oprofile works fine, I'm just
putting an idea out there given comments from others.
> Oprofile fits neither of those categories.
So it should remain EXPERIMENTAL then in your view?
On Sun, May 21, 2006 at 03:57:10PM -0400, Dave Jones wrote:
> > Given that it should probably stay EXPERIMENTAL for now. I wonder if
> > left unmaintained for much longer if it should be potentially marked
> > BROKEN and/or scheduled for removal?
>
> Why on earth would we want to remove a working feature ?
I certainly didn't mean to suggest that. But it should certainly stay at
EXPERIMENTAL until the project as a whole reaches a stable release.
regards,
john
On Sun, May 21, 2006 at 01:17:20PM -0700, Chris Wedgwood wrote:
> > BROKEN is for stuff that doesn't compile, or is fundamentally flawed
> > beyond repair at the current time (For example, needs infrastructure
> > work to happen before it can work correctly).
>
> Sure, based on my personal experience oprofile works fine, I'm just
> putting an idea out there given comments from others.
>
> > Oprofile fits neither of those categories.
>
> So it should remain EXPERIMENTAL then in your view?
Part of the problem with EXPERIMENTAL is that it's been taken
to mean all sorts of things over the years from "this isn't quite ready, but works"
to "if you enable this, you're a lunatic". It's meant that pretty much
everyone has to enable CONFIG_EXPERIMENTAL anyway for a kernel offering
the features everyone has come to expect. As means of example, I'll bet
there's no (or at least very few) distros that ship a kernel with
CONFIG_EXPERIMENTAL disabled.
Dave
--
http://www.codemonkey.org.uk
On Sun, May 21, 2006 at 04:49:47PM -0400, Dave Jones wrote:
> Part of the problem with EXPERIMENTAL is that it's been taken
> to mean all sorts of things over the years
yes
> from "this isn't quite ready, but works"
ok for -mm perhaps, not mainline
> to "if you enable this, you're a lunatic".
suitable for neither -mm not mainline
> It's meant that pretty much everyone has to enable
> CONFIG_EXPERIMENTAL anyway for a kernel offering the features
> everyone has come to expect. As means of example, I'll bet there's
> no (or at least very few) distros that ship a kernel with
> CONFIG_EXPERIMENTAL disabled.
i don't see why we can't audit EXPERIMENTAL users and try to clarify
which ones are really needed over time though
On Sul, 2006-05-21 at 12:49 -0700, Chris Wedgwood wrote:
> Some feedback claims this is not the case and that "it's unmaintained,
> rarely well-tested,
By which theory we should ditch ia-64 while we are at it.
Be serious, oprofile is good working code, even if you have some
personal problem with it.
On Mon, May 22, 2006 at 12:51:21PM +0100, Alan Cox wrote:
> Be serious, oprofile is good working code, even if you have some
> personal problem with it.
Does the opinion of the former mantainer count for nothing? If nothing
else, it should remain experimental on arches like Alpha, where there's
a whole bunch of events that can't possibly work.
regards
john
On Mon, May 22, 2006 at 12:51:21PM +0100, Alan Cox wrote:
> By which theory we should ditch ia-64 while we are at it.
ia64 is used, fairly well tested and has a large body of people
working on it
if the code isn't to your liking then I'm sure the people involved
are open to your constructive comments
> Be serious, oprofile is good working code, even if you have some
> personal problem with it.
Yes, oprofile does work, I use it myself. So the core issue is really
whether it should be EXPERIMENTAL or not then.
I've sorta come to the concluusion the EXPERIMENTAL is basically
pointless, it hides too many useful things.
John Levon <[email protected]> writes:
> On Mon, May 22, 2006 at 12:51:21PM +0100, Alan Cox wrote:
>
>> Be serious, oprofile is good working code, even if you have some
>> personal problem with it.
>
> Does the opinion of the former mantainer count for nothing? If nothing
> else, it should remain experimental on arches like Alpha, where there's
> a whole bunch of events that can't possibly work.
Why should be marked experimental only because of architecture limits?
If the parts that can work, work well, there's no reason to suggest
otherwise. (Speaking as an Alpha owner)
--
M?ns Rullg?rd
[email protected]
On Mon, May 22, 2006 at 06:50:28PM +0100, M?ns Rullg?rd wrote:
> Why should be marked experimental only because of architecture
> limits? If the parts that can work, work well, there's no reason to
> suggest otherwise. (Speaking as an Alpha owner)
The only conclusion I can draw from this is that EXPERIMENTAL has no
well defined semantics and as such is essentially pointless. Everyone
interprets it differently.
I think this is exactly what DaveJ was saying the other day.