[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Matthew Wilcox]

On Thu, Apr 19, 2001 at 18:50:34 EDT, Eric S. Raymond wrote:
> Remove dead CONFIG_BINFMT_JAVA symbol.

Please don't do this, it just makes merging our patches with Linus harder.
This has already been done in our tree since Feb 1.  In fact, please
don't touch anything in the tree which is PA specific; we have a large
arch update pending.

http://puffin.external.hp.com/cvs/linux/arch/parisc/config.in?log=y

shows the current state of our config.in, if you're curious.  If you
have any changes you want to make, don't hesitate to coordinate with us
by mailing parisc-linux@parisc-linux.org.

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Matthew Wilcox <willy@ldl.fc.hp.com>:
> On Thu, Apr 19, 2001 at 18:50:34 EDT, Eric S. Raymond wrote:
> > Remove dead CONFIG_BINFMT_JAVA symbol.
> 
> Please don't do this, it just makes merging our patches with Linus harder.

Bother.  I've now heard "don't touch that tree!" from you and the ARM
folks.  I'm trying to be a good neighbor, here, but there is some
cleanup I want to do that crosses port boundaries.  (None of this is CML2,
BTW; I'm now addressing problems that are common to CML1 as well.)

What is the right procedure for doing changes like this?  Is "don't
touch that tree" a permanent condition, or am I going to get a chance
to clean up the global CONFIG_ namespace after your next merge-down?

Could I ask you to audit your tree and change the prefix on any 
CONFIG_ symbols that are private over there?  This would make life 
easier for my auditing tools (kxref and Stephen Cole's ach script).

That's the main thing I'm after right now -- I want to cut down on
the false positives in my orphaned-symbol reports so that the actual
bugs will stand out.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

"They that can give up essential liberty to obtain a little temporary 
safety deserve neither liberty nor safety."
	-- Benjamin Franklin, Historical Review of Pennsylvania, 1759.

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Matthew Wilcox]

On Thu, Apr 19, 2001 at 11:00:09PM -0400, Eric S. Raymond wrote:
> What is the right procedure for doing changes like this?  Is "don't
> touch that tree" a permanent condition, or am I going to get a chance
> to clean up the global CONFIG_ namespace after your next merge-down?

Our current status is that we've got a patch with Alan that's been sitting
in his tree for a while (things got trickier than he expected and he
hasn't been able to merge that upstream to Linus yet).  Meanwhile we've
carried on development as normal.  So even after the patches in Alan's
tree land, we've still got a fair hunk of changes to go in.

My preference would be for you to fetch our tree 

cvs -d :pserver:anonymous@puffin.external.hp.com:/home/cvs/parisc login
[no password]

cvs -d :pserver:anonymous@puffin.external.hp.com:/home/cvs/parisc co linux

and submit patches to us, which will get to Linus in the fullness of time.
I'm aware this might not be terribly satisfactory for you, but we're
doing our best not to lose our way amid the churn of development right
now and having patches which haven't followed a progression through
us makes that significantly harder.

> Could I ask you to audit your tree and change the prefix on any 
> CONFIG_ symbols that are private over there?  This would make life 
> easier for my auditing tools (kxref and Stephen Cole's ach script).

I don't think we have any of those.  We certainly have symbols which are
defined for symmetry and may not actually be used yet (CONFIG_PA11 might not
be, perhaps).  But that's what happens when you're developing software :-)

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[james rich]

On Thu, 19 Apr 2001, Matthew Wilcox wrote:

> On Thu, Apr 19, 2001 at 11:00:09PM -0400, Eric S. Raymond wrote:
> > What is the right procedure for doing changes like this?  Is "don't
> > touch that tree" a permanent condition, or am I going to get a chance
> > to clean up the global CONFIG_ namespace after your next merge-down?
> 

[snip]

> My preference would be for you to fetch our tree 

> and submit patches to us, which will get to Linus in the fullness of time.

Truly this is not meant to be negative - don't take it as such.

Doesn't this seem a little like the problems occurring with lvm right now?
A separate tree maintained with the maintainers not wanting others
submitting patches that conflict with their particular tree?  It seems
that any project should be able to submit any patch against The One True
Tree: Linus' tree.

James Rich
james.rich@m.cc.utah.edu

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Matthew Wilcox]

On Thu, Apr 19, 2001 at 10:07:22PM -0600, james rich wrote:
> Doesn't this seem a little like the problems occurring with lvm right now?
> A separate tree maintained with the maintainers not wanting others
> submitting patches that conflict with their particular tree?  It seems
> that any project should be able to submit any patch against The One True
> Tree: Linus' tree.

every single architecture has their own development tree.  the pa project
has not been running as long as the other ports, and has a large amount of
development going on.  i count 28 commits for april (so far), 75 commits
for march, 187 for february and 112 for january (to the kernel tree, other
parts of the port also have commit messages).  linus would go insane if
we sent him every single one of those patches individually.  and we'd
go insane trying to keep up with what he'd taken and what he'd dropped.

until you've actually tried doing this, please don't attempt to criticise.

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Albert D. Cahalan]

Matthew Wilcox writes:
> On Thu, Apr 19, 2001 at 10:07:22PM -0600, james rich wrote:

>> Doesn't this seem a little like the problems occurring with lvm right now?
>> A separate tree maintained with the maintainers not wanting others
>> submitting patches that conflict with their particular tree?  It seems
>> that any project should be able to submit any patch against The One True
>> Tree: Linus' tree.
>
> every single architecture has their own development tree.

This sucks for users of that architecture. Also, though not
applicable to PA-RISC, it sucks for sub-architecture porters.
(by sub-architecture I mean: Mac, PReP, PowerCore, BeBox, etc.)

It's hard enough deciding between Linus and Alan. I'm not at all
happy trying to pick through obscure CVS and BitKeeper trees that
might not be up-to-data with the latest mainstream bug fixes.

> the pa project
> has not been running as long as the other ports, and has a large amount of
> development going on.  i count 28 commits for april (so far), 75 commits
> for march, 187 for february and 112 for january (to the kernel tree, other
> parts of the port also have commit messages).  linus would go insane if
> we sent him every single one of those patches individually.  and we'd
> go insane trying to keep up with what he'd taken and what he'd dropped.
> 
> until you've actually tried doing this, please don't attempt to criticise.

Have _you_ tried? If I recall correctly, Linus spoke out against the
PowerPC people doing the exact same thing. So unless you get told to
quit annoying him with patches, sending them is the safe bet.

Well here we go. It's about IrDA though, not PowerPC. Read it!
http://lwn.net/2000/1109/a/lt-IrDA.php3

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Rik van Riel]

On Fri, 20 Apr 2001, Albert D. Cahalan wrote:

> This sucks for users of that architecture. Also, though not
> applicable to PA-RISC, it sucks for sub-architecture porters.
> (by sub-architecture I mean: Mac, PReP, PowerCore, BeBox, etc.)

As you said it so eloquently a few paragraphs down:

	"send patches!"

cheers,

Rik
--
Virtual memory is like a game you can't win;
However, without VM there's truly nothing to lose...

		http://www.surriel.com/
http://www.conectiva.com/	http://distro.conectiva.com.br/

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[David Woodhouse]

james.rich@m.cc.utah.edu said:
>  Doesn't this seem a little like the problems occurring with lvm right
> now? A separate tree maintained with the maintainers not wanting
> others submitting patches that conflict with their particular tree?
> It seems that any project should be able to submit any patch against
> The One True Tree: Linus' tree. 

Of course they can. Linus does apply them too. People are asking nicely 
that ESR not do so in this case, because merges are being planned. 

The contents of drivers/mtd/ are in the same situation. For some reason, I
felt it inappropriate to give every patch at every stage of development to
Linus for inclusion in the 2.4.0-test and 2.4.[123] kernels. Now I'm vaguely
happy with it all and it's stable, I'm working on cleaning up some of the 
cosmetics and breaking it up into digestible patches.

Doing primary development in CVS seems to work OK for me, and allows me to 
continue development without destabilising the One True Tree. During such 
times, it's useful to have a branch for the code which is in the One True 
Tree, so urgent fixes can be merged, and the diff against the One True Tree 
after each release has something to diff against to catch patches where 
people didn't even bother to Cc the maintainer.

I believe people were _told_ to hold off until 2.4.5-ish, or when the tree
became stable. Violent imagery was used to reinforce this instruction.
That being the case, how about holding the config changes back until after 
everyone else who's been waiting has merged their pending changes?

--
dwmw2

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Alan Cox]

> What is the right procedure for doing changes like this?  Is "don't
> touch that tree" a permanent condition, or am I going to get a chance
> to clean up the global CONFIG_ namespace after your next merge-down?

Feeding arch related stuff to the architecture maintainers.

> That's the main thing I'm after right now -- I want to cut down on
> the false positives in my orphaned-symbol reports so that the actual
> bugs will stand out.

Teach it to read a 'symbolstoignore' file.

Part of the problem you are hitting right now is that most architectures are
not yet fully in sync with 2.4 nor likely to all be for another few iterations.

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Alan Cox]

> > we sent him every single one of those patches individually.  and we'd
> > go insane trying to keep up with what he'd taken and what he'd dropped.
> > 
> > until you've actually tried doing this, please don't attempt to criticise.
> 
> Have _you_ tried? If I recall correctly, Linus spoke out against the

I have for one. Its definitely the wrong approach to bomb Linus with patches
when doing the merge of an architecture. All the architecture folk with in
their own trees for good reason.

Once the code is in a fit state to merge (ie actually works well with the new
2.4.x stuff and 2.4.x core stops shifting around) then the merge can get done
piece by piece.

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Alan Cox <alan@lxorguk.ukuu.org.uk>:
> > What is the right procedure for doing changes like this?  Is "don't
> > touch that tree" a permanent condition, or am I going to get a chance
> > to clean up the global CONFIG_ namespace after your next merge-down?
> 
> Feeding arch related stuff to the architecture maintainers.

I shall attempt it.

> > That's the main thing I'm after right now -- I want to cut down on
> > the false positives in my orphaned-symbol reports so that the actual
> > bugs will stand out.
> 
> Teach it to read a 'symbolstoignore' file.

Someone else has already pointed out that this is not a solution that will
scale well.  It would substitute a continuing management headache for the
cleanup that's really needed.  In fact I'm reluctant to do this even for 
cases where it's clearly legitimate (CONFIG_BOOM, CONFIG_BOGUS :-)) partly
because later on it might provide an excuse for people not to do the cleanup.

> Part of the problem you are hitting right now is that most
> architectures are not yet fully in sync with 2.4 nor likely to all
> be for another few iterations.

Understood.  I'll do what I can in the architecture-independent code, then.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

"Boys who own legal firearms have much lower rates of delinquency and
drug use and are even slightly less delinquent than nonowners of guns."
	-- U.S. Department of Justice, National Institute of
	   Justice, Office of Juvenile Justice and Delinquency Prevention,
	   NCJ-143454, "Urban Delinquency and Substance Abuse," August 1995.

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Alan Cox <alan@lxorguk.ukuu.org.uk>:
> I have for one. Its definitely the wrong approach to bomb Linus with patches
> when doing the merge of an architecture. All the architecture folk with in
> their own trees for good reason.

On the other hand, Linus has objected to the One-Big-Patch approach in
the past with respect to things like the networking and VM code.  How
are people to know what the right thing is?
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

Love your country, but never trust its government.
	-- Robert A. Heinlein.

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Alan Cox]

> Alan Cox <alan@lxorguk.ukuu.org.uk>:
> > I have for one. Its definitely the wrong approach to bomb Linus with patches
> > when doing the merge of an architecture. All the architecture folk with in
> > their own trees for good reason.
> 
> On the other hand, Linus has objected to the One-Big-Patch approach in
> the past with respect to things like the networking and VM code.  How
> are people to know what the right thing is?

Who said anything about one big patch ?

Just because you have a lot of differences doesnt mean you send Linus one giant
splat of code. I don't send Linus -ac for example.

Alan

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Alan Cox <alan@lxorguk.ukuu.org.uk>:
> > Alan Cox <alan@lxorguk.ukuu.org.uk>:
> > > I have for one. Its definitely the wrong approach to bomb Linus
> > > with patches when doing the merge of an architecture. All the
> > > architecture folk with in their own trees for good reason.
> > 
> > On the other hand, Linus has objected to the One-Big-Patch approach in
> > the past with respect to things like the networking and VM code.  How
> > are people to know what the right thing is?
> 
> Who said anything about one big patch ?  Just because you have a lot
> of differences doesnt mean you send Linus one giant splat of code. I
> don't send Linus -ac for example.

OK, so maybe I'm being stupid.  But the implication of this talk of separate
port trees and architecture merges is that these guys periodically send big
resync patches to you and Linus.

If that's not what's going on, what is?
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

Never could an increase of comfort or security be a sufficient good to be
bought at the price of liberty.
	-- Hillaire Belloc

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Alan Cox]

> OK, so maybe I'm being stupid.  But the implication of this talk of separate
> port trees and architecture merges is that these guys periodically send big
> resync patches to you and Linus.
> 
> If that's not what's going on, what is?

People send batches of small fixes to Linus or to me. So for example the S/390
folks send me things like 'fix the mm layer to match the changes in 2.4.3'
and 'Update the DASD storage driver'. Each of which fixes one thing or one
set of things and is easy to check on its own

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Alan Cox <alan@lxorguk.ukuu.org.uk>:
> People send batches of small fixes to Linus or to me. So for example
> the S/390 folks send me things like 'fix the mm layer to match the
> changes in 2.4.3' and 'Update the DASD storage driver'. Each of
> which fixes one thing or one set of things and is easy to check on
> its own

I'll continue asking stupid questions, then.  Like, under this system how
can either you or the port maintainers maintain a good representation of 
how far out of sync they are with the main tree?

The implied workflow (developers in general, up to port maintainers,
up to you and Linus) makes both technological and sociological sense.
It kind of reminds me of Anglo-Norman feudalism post-1066 ("No lord
without land, no land without a lord.").

There are a couple of funny edge cases that it doesn't seem to handle
well, though.  One is the kind I'm bumping into right now, where
somebody legitimately needs to make small (almost trivial) changes
scattered all through the tree.

Another is the case where a piece of code that needs to be changed doesn't
have an active maintainer for a third party like me to go to.

What's the neighborly way to deal with these?
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

"The best we can hope for concerning the people at large is that they be
properly armed."
        -- Alexander Hamilton, The Federalist Papers at 184-188

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Alan Cox]

> I'll continue asking stupid questions, then.  Like, under this system how
> can either you or the port maintainers maintain a good representation of 
> how far out of sync they are with the main tree?

diff and read the output.

[bizzare sociopolitical mumble deleted]

> well, though.  One is the kind I'm bumping into right now, where
> somebody legitimately needs to make small (almost trivial) changes
> scattered all through the tree.

Yep. But such changes are rare. Or should be. 

> Another is the case where a piece of code that needs to be changed doesn't
> have an active maintainer for a third party like me to go to.
> What's the neighborly way to deal with these?

If I get patches for stuff that doesnt seem to have a maintainer I apply them.
On the odd occasion a scream is heard in the distance it means I now know
there is an active maintainer.

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Alan Cox <alan@lxorguk.ukuu.org.uk>:
> > well, though.  One is the kind I'm bumping into right now, where
> > somebody legitimately needs to make small (almost trivial) changes
> > scattered all through the tree.
> 
> Yep. But such changes are rare. Or should be. 

Knowing that doesn't help me much, since I'm trying to fix up a global
namespace that touches everybody :-(.
 
> If I get patches for stuff that doesnt seem to have a maintainer I
> apply them.  On the odd occasion a scream is heard in the distance
> it means I now know there is an active maintainer.

All right then.  I'm going to send you a bunch of dead-symbol cleanup
patches.  I'll try to stay in the mainline code and out of the port
trees.  Would you please do me the kindness of telling me which ones
can go in and which ones you think have to go through maintainers?

You should have received one such patch already, fixes for two
documentation files.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

Sometimes the law defends plunder and participates in it. Sometimes
the law places the whole apparatus of judges, police, prisons and
gendarmes at the service of the plunderers, and treats the victim --
when he defends himself -- as a criminal.
	-- Frederic Bastiat, "The Law"

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Tom Rini]

On Fri, Apr 20, 2001 at 10:59:34AM -0400, Eric S. Raymond wrote:
> Alan Cox <alan@lxorguk.ukuu.org.uk>:
> > > well, though.  One is the kind I'm bumping into right now, where
> > > somebody legitimately needs to make small (almost trivial) changes
> > > scattered all through the tree.
> > 
> > Yep. But such changes are rare. Or should be. 
> 
> Knowing that doesn't help me much, since I'm trying to fix up a global
> namespace that touches everybody :-(.

Which does boil down to having to work with trees other than Linus or
Alans.  Remember, the official tree is not always the up-to-date tree,
or in the case of other arches, the most relevant tree.  But if you send
something off to a maintainer, there's a good chance (if they're still active)
they'll do what you ask, and it'll get to Linus/Alan the next time they sync.
As long as the problem gets fixed, it shouldn't be as important if it's fixed
in everyones tree right now, or in a release or two.  If it's some sort of
huge bug it just might get fixed sooner.

-- 
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Jeff Garzik]

Tom Rini wrote:
> Which does boil down to having to work with trees other than Linus or
> Alans.  Remember, the official tree is not always the up-to-date tree,
> or in the case of other arches, the most relevant tree.

Yep.  You could even look at Linus as simply the x86 port maintainer :)

Except for alpha and x86, AFAIK, most people wind up going through
arch-specific channels to get their kernels...

-- 
Jeff Garzik      | The difference between America and England is that
Building 1024    | the English think 100 miles is a long distance and
MandrakeSoft     | the Americans think 100 years is a long time.
                 |      (random fortune)

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Bob McElrath]

[-- Attachment #1: Type: text/plain, Size: 980 bytes --]

Jeff Garzik [jgarzik@mandrakesoft.com] wrote:
> Tom Rini wrote:
> > Which does boil down to having to work with trees other than Linus or
> > Alans.  Remember, the official tree is not always the up-to-date tree,
> > or in the case of other arches, the most relevant tree.
> 
> Yep.  You could even look at Linus as simply the x86 port maintainer :)
> 
> Except for alpha and x86, AFAIK, most people wind up going through
> arch-specific channels to get their kernels...

This may be a dumb question, but is there some place where the arch
maintainers are listed?  Where the arch-specific trees are kept?  Where
would I go to get the latest set of relevant patches for alpha?

Grepping the Documentation/ directory for "alpha" I see nothing
relevant.  IMHO this should all be listend in one place.  Maybe
Documentation/Arch-Maintainers.txt.

Cheers,
-- Bob

Bob McElrath (rsmcelrath@students.wisc.edu) 
Univ. of Wisconsin at Madison, Department of Physics

[-- Attachment #2: Type: application/pgp-signature, Size: 240 bytes --]

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Matthew Wilcox]

On Fri, Apr 20, 2001 at 11:15:12AM -0500, Bob McElrath wrote:
> This may be a dumb question, but is there some place where the arch
> maintainers are listed?  Where the arch-specific trees are kept?  Where
> would I go to get the latest set of relevant patches for alpha?

http://www.kernel.org/ has a list of architecture websites.  Also the
CREDITS / MAINTAINERS files tend to list the people who are involved.

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Jeff Garzik]

Bob McElrath wrote:
> 
> Jeff Garzik [jgarzik@mandrakesoft.com] wrote:
> > Tom Rini wrote:
> > > Which does boil down to having to work with trees other than Linus or
> > > Alans.  Remember, the official tree is not always the up-to-date tree,
> > > or in the case of other arches, the most relevant tree.
> >
> > Yep.  You could even look at Linus as simply the x86 port maintainer :)
> >
> > Except for alpha and x86, AFAIK, most people wind up going through
> > arch-specific channels to get their kernels...
> 
> This may be a dumb question, but is there some place where the arch
> maintainers are listed?  Where the arch-specific trees are kept?  Where
> would I go to get the latest set of relevant patches for alpha?

As I noted in the e-mail to which you replied, there is no separate
alpha tree nor arch-specific channel for alpha kernels.  Generally fixes
to the Alpha tree appear quickly and get merged quickly, and the tree
stays in sync quite well.  Watch linux-kernel or Alan Cox's patchkit for
Alpha fixes that may be in transmit to Linus.

There are of course RedHat's alpha distro, and various mailing lists on
http://www.alphalinux.org/

-- 
Jeff Garzik      | The difference between America and England is that
Building 1024    | the English think 100 miles is a long distance and
MandrakeSoft     | the Americans think 100 years is a long time.
                 |      (random fortune)

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Nicolas Pitre]

On Fri, 20 Apr 2001, Tom Rini wrote:

> On Fri, Apr 20, 2001 at 10:59:34AM -0400, Eric S. Raymond wrote:
> > Alan Cox <alan@lxorguk.ukuu.org.uk>:
> > > > well, though.  One is the kind I'm bumping into right now, where
> > > > somebody legitimately needs to make small (almost trivial) changes
> > > > scattered all through the tree.
> > >
> > > Yep. But such changes are rare. Or should be.
> >
> > Knowing that doesn't help me much, since I'm trying to fix up a global
> > namespace that touches everybody :-(.
>
> Which does boil down to having to work with trees other than Linus or
> Alans.  Remember, the official tree is not always the up-to-date tree,
> or in the case of other arches, the most relevant tree.  But if you send
> something off to a maintainer, there's a good chance (if they're still active)
> they'll do what you ask, and it'll get to Linus/Alan the next time they sync.
> As long as the problem gets fixed, it shouldn't be as important if it's fixed
> in everyones tree right now, or in a release or two.  If it's some sort of
> huge bug it just might get fixed sooner.

Guys,

There is kind of a ridiculous situation here where people want to withhold
their own changes in their own trees for all good reasons until it is mature
and stable enough to be fed upstream in the appropriate way, while insisting
for Linus' tree to remain incomplete and inconsistent.  And we're not
talking about deep architectural changes here -- only about configure
symbols and help text.

Why not having everybody's tree consistent with themselves and have whatever
CONFIGURE_* symbols and help text be merged along with the very code it
refers to?  It's worthless to have config symbols be merged into Linus' or
Alan's tree if the code isn't there (yet).  It simply makes no sense.

This might shift some of the namespace consistency work to architecture
maintainers (which is a good thing IMHO) and establish the basis for yet a
more sanitized kernel.org tree at all times for before and after any
further patches are merged.

I'm myself maintainer of a subarchitecture and removing currently
unreferenced SA1100 config symbols from the official Linux tree would
probably give me a one-time effort to bring them back in my tree but this is
certainly for a saner code/namespace distribution in general.  Why should
the symbols I maintain remain there if I'm not ready yet to sync up the code
they refer to?

Hey this is only CONFIG_ symbols after all.  If they get removed now, they
will only reappear _with_ the code they refer to eventually when it'll get
merged.


Nicolas

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Nicolas Pitre <nico@cam.org>:
> Why not having everybody's tree consistent with themselves and have whatever
> CONFIGURE_* symbols and help text be merged along with the very code it
> refers to?  It's worthless to have config symbols be merged into Linus' or
> Alan's tree if the code isn't there (yet).  It simply makes no sense.

And now it has a cost, too.  It makes finding real bugs more difficult.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

Every election is a sort of advance auction sale of stolen goods. 
	-- H.L. Mencken 

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Tom Rini]

On Fri, Apr 20, 2001 at 12:35:12PM -0400, Nicolas Pitre wrote:

> There is kind of a ridiculous situation here where people want to withhold
> their own changes in their own trees for all good reasons until it is mature
> and stable enough to be fed upstream in the appropriate way, while insisting
> for Linus' tree to remain incomplete and inconsistent.  And we're not
> talking about deep architectural changes here -- only about configure
> symbols and help text.

The answer is simple, noise.

> Why not having everybody's tree consistent with themselves and have whatever
> CONFIGURE_* symbols and help text be merged along with the very code it
> refers to?  It's worthless to have config symbols be merged into Linus' or
> Alan's tree if the code isn't there (yet).  It simply makes no sense.

Well, this depends a lot on a) The project to be merged (arch, mtd, whatever)
and b) how far something has gotten in being merged someplace else, and of
course c) the maintainer(s).  Whatever the exact case, and in general, it 
should be handled via the maintainer.  Why? They maintain the code.

-- 
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Matthew Wilcox]

On Fri, Apr 20, 2001 at 02:00:00PM -0500, Jeff Dike wrote:
> matthew@wil.cx said:
> > http://www.kernel.org/ has a list of architecture websites.  Also the
> > CREDITS / MAINTAINERS files tend to list the people who are involved. 
> 
> Except it's restricted to processor ports, which would leave you not knowing 
> about UML.

Have you tried mailing webmaster@kernel.org and asking to be added?

-- 
Revolutions do not require corporate support.

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Nicolas Pitre]

On Fri, 20 Apr 2001, Tom Rini wrote:

> On Fri, Apr 20, 2001 at 12:35:12PM -0400, Nicolas Pitre wrote:
>
> > Why not having everybody's tree consistent with themselves and have whatever
> > CONFIGURE_* symbols and help text be merged along with the very code it
> > refers to?  It's worthless to have config symbols be merged into Linus' or
> > Alan's tree if the code isn't there (yet).  It simply makes no sense.
>
> Well, this depends a lot on a) The project to be merged (arch, mtd, whatever)
> and b) how far something has gotten in being merged someplace else, and of
> course c) the maintainer(s).  Whatever the exact case, and in general, it
> should be handled via the maintainer.  Why? They maintain the code.

Therefore it's the maintainer's job to submit coherent patches and accept to
see inconsistent CONFIG_* references be removed from the official tree until
further patch submission is due.  It's only a question of discipline.
Otherwise how can you distinguish between dead wood which must be removed
and potentially valid symbols referring to code existing only in a remote
tree?


Nicolas

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Russell King]

On Fri, Apr 20, 2001 at 10:59:34AM -0400, Eric S. Raymond wrote:
> All right then.  I'm going to send you a bunch of dead-symbol cleanup
> patches.  I'll try to stay in the mainline code and out of the port
> trees.  Would you please do me the kindness of telling me which ones
> can go in and which ones you think have to go through maintainers?

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Jes Sorensen]

>>>>> "Jeff" == Jeff Dike <jdike@karaya.com> writes:

Jeff> matthew@wil.cx said:
>> http://www.kernel.org/ has a list of architecture websites.  Also
>> the CREDITS / MAINTAINERS files tend to list the people who are
>> involved.

Jeff> Except it's restricted to processor ports, which would leave you
Jeff> not knowing about UML.

I'd be highly surprised if they said no to adding UML to the list if
you mailed them a request to update the page.

Jes

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Tom Rini]

On Fri, Apr 20, 2001 at 02:48:18PM -0400, Nicolas Pitre wrote:
> 
> 
> On Fri, 20 Apr 2001, Tom Rini wrote:
> 
> > On Fri, Apr 20, 2001 at 12:35:12PM -0400, Nicolas Pitre wrote:
> >
> > > Why not having everybody's tree consistent with themselves and have whatever
> > > CONFIGURE_* symbols and help text be merged along with the very code it
> > > refers to?  It's worthless to have config symbols be merged into Linus' or
> > > Alan's tree if the code isn't there (yet).  It simply makes no sense.
> >
> > Well, this depends a lot on a) The project to be merged (arch, mtd, whatever)
> > and b) how far something has gotten in being merged someplace else, and of
> > course c) the maintainer(s).  Whatever the exact case, and in general, it
> > should be handled via the maintainer.  Why? They maintain the code.
> 
> Therefore it's the maintainer's job to submit coherent patches and accept to
> see inconsistent CONFIG_* references be removed from the official tree until
> further patch submission is due.  It's only a question of discipline.
> Otherwise how can you distinguish between dead wood which must be removed
> and potentially valid symbols referring to code existing only in a remote
> tree?

Er, I think we agree, but I'm not sure. :)
The only people who actually know the difference between dead wood and partily
merged code are the maintainers.  IMHO it's silly to remove a piece of code
like:
#ifdef CONFIG_SOMETHING_NOT_MERGED
...
#endif
If the rest of the code, which would make the above useful is heading toward
Linus.

-- 
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Jeff Dike]

matthew@wil.cx said:
> http://www.kernel.org/ has a list of architecture websites.  Also the
> CREDITS / MAINTAINERS files tend to list the people who are involved. 

Except it's restricted to processor ports, which would leave you not knowing 
about UML.

				Jeff

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Russell King]

On Fri, Apr 20, 2001 at 12:50:05PM -0400, Eric S. Raymond wrote:
> Nicolas Pitre <nico@cam.org>:
> > Why not having everybody's tree consistent with themselves and have whatever
> > CONFIGURE_* symbols and help text be merged along with the very code it
> > refers to?  It's worthless to have config symbols be merged into Linus' or
> > Alan's tree if the code isn't there (yet).  It simply makes no sense.

Really, the above issue is down to the sub-architecture maintainers splitting
up their patches into the "one feature, one bug" thing, rather than "one
set of files" (which, incidentally I'm guilty of as well).  That way, when
stuff gets added, you get:

1. The C source changes for that item
2. The configure script stuff for that one item
3. The help text for that one item.

Currently, stuff that comes to me appears mostly as "here's a configure
update", "here's a PCMCIA update", etc.  I'll pull out an instance from
my patch tracking system (sorry, Philip, yours is the first one I found):

Patch 413/1 (see http://www.arm.linux.org.uk/developer/patches/?id=413/1&mode=patch)
This patch adds the defconfig file for the CLPS7500 architecture, and it
contains symbols such as:

	CONFIG_BLK_DEV_FLD7500
	CONFIG_CLPS7500_FLASH

Neither of these two drivers are currently in Linus' tree, or in fact my
tree.  Should I reject the patch?  Should I accept it and edit these out,
or what?

> And now it has a cost, too.  It makes finding real bugs more difficult.

Well, if they get removed in Linus tree, then when I next sync, they'll get
re-added, or maybe they won't.  Then someone else will remove them, then
they'll get re-added ad infinitum.

This also touches on another issue - patch.  I've had several times where
I've sent Alan stuff, its gone up to Linus, I receive it back, and during
the merge, patch does its stuff without complaining (because there is not
enough context in the diff).  Typically, this happens in the Configure.help
file.

Generally it seems like diff needs to produce one more line of context, and
most of these problems will go away.  Yes, there will still be the odd
problem, so then it becomes the "how much do you crank the setting" problem.

--
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Matthew Wilcox <willy@ldl.fc.hp.com>:
> > Could I ask you to audit your tree and change the prefix on any 
> > CONFIG_ symbols that are private over there?  This would make life 
> > easier for my auditing tools (kxref and Stephen Cole's ach script).
> 
> I don't think we have any of those.  We certainly have symbols which are
> defined for symmetry and may not actually be used yet (CONFIG_PA11 might not
> be, perhaps).  But that's what happens when you're developing software :-)

Here's what I have for you guys:

CONFIG_BINFMT_JAVA: arch/parisc/config.in arch/parisc/defconfig arch/cris/config.in arch/cris/defconfig

You've already gotten rid of that one.

CONFIG_BINFMT_SOM: arch/parisc/config.in arch/parisc/defconfig

Not used in code anywhere.  Can you get rid of this one?

CONFIG_DMB_TRAP: arch/parisc/kernel/sba_iommu.c
CONFIG_FUNC_SIZE: arch/parisc/kernel/sba_iommu.c

Would you please take these out of the CONFIG_ namespace?  Changing the 
prefix to CONFIGURE would do nicely.

CONFIG_GENRTC: arch/parisc/defconfig

This is a typo for GEN_RTC.  Please fix it or get rid of it.

CONFIG_HIL: arch/parisc/defconfig

Looks like an orphan.  Can you get rid of it?

CONFIG_GSC: arch/parisc/config.in arch/parisc/defconfig
CONFIG_GSC_DINO: arch/parisc/config.in arch/parisc/defconfig
CONFIG_GSC_LASI: arch/parisc/config.in arch/parisc/defconfig arch/parisc/kernel/led.c
CONFIG_GSC_PS2: arch/parisc/config.in arch/parisc/defconfig
CONFIG_IODC_CONSOLE: arch/parisc/config.in arch/parisc/kernel/setup.c
CONFIG_IOMMU_CCIO: arch/parisc/config.in arch/parisc/defconfig arch/parisc/kernel/Makefile
CONFIG_IOMMU_SBA: arch/parisc/config.in arch/parisc/defconfig arch/parisc/kernel/Makefile
CONFIG_IOSAPIC: arch/parisc/config.in arch/parisc/defconfig arch/parisc/kernel/Makefile
CONFIG_KWDB: arch/parisc/Makefile arch/parisc/config.in arch/parisc/defconfig arch/parisc/kernel/entry.S arch/parisc/kernel/traps.c arch/parisc/mm/init.c
CONFIG_LASI_82596: arch/parisc/config.in arch/parisc/defconfig
CONFIG_PARPORT_GSC: drivers/parport/Makefile arch/parisc/config.in arch/parisc/defconfig
CONFIG_PCI_LBA: arch/parisc/config.in arch/parisc/defconfig arch/parisc/kernel/Makefile
CONFIG_SCSI_LASI: arch/parisc/config.in arch/parisc/defconfig
CONFIG_SCSI_ZALON: arch/parisc/config.in arch/parisc/defconfig
CONFIG_STI_CONSOLE: arch/parisc/Makefile arch/parisc/config.in arch/parisc/defconfig arch/parisc/kernel/setup.c arch/parisc/mm/init.c

Looks like these need Configure.help entries.

CONFIG_SERIAL_GSC: drivers/char/serial.c arch/parisc/defconfig

That reference pattern looks kind of weird.  Is this a bug?

If you could clean these up, that's everything I need from the parisc tree.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

(Those) who are trying to read the Second Amendment out of the Constitution by
claiming it's not an individual right (are) courting disaster by encouraging
others to use the same means to eliminate portions of the Constitution they
don't like.
	-- Alan Dershowitz, Harvard Law School

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Matthew Wilcox]

On Fri, Apr 20, 2001 at 03:47:43PM -0400, Eric S. Raymond wrote:
> CONFIG_BINFMT_SOM: arch/parisc/config.in arch/parisc/defconfig
> 
> Not used in code anywhere.  Can you get rid of this one?

Code not merged yet.

> CONFIG_DMB_TRAP: arch/parisc/kernel/sba_iommu.c
> CONFIG_FUNC_SIZE: arch/parisc/kernel/sba_iommu.c
> 
> Would you please take these out of the CONFIG_ namespace?  Changing the 
> prefix to CONFIGURE would do nicely.

Grant?  This is your code.

> CONFIG_GENRTC: arch/parisc/defconfig
> 
> This is a typo for GEN_RTC.  Please fix it or get rid of it.

in our tree it's in drivers/char/Makefile:

obj-$(CONFIG_GENRTC) += genrtc.o

this code was written by Sam Creasey as part of the sun3 port, and he
schlepped it into our tree too.  we have no GEN_RTC in our tree.

> CONFIG_HIL: arch/parisc/defconfig
> 
> Looks like an orphan.  Can you get rid of it?

code not yet merged.

> CONFIG_SERIAL_GSC: drivers/char/serial.c arch/parisc/defconfig
> 
> That reference pattern looks kind of weird.  Is this a bug?

it's old and needs to die properly.  i haven't had time to fix that yet.

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Matthew Wilcox <willy@ldl.fc.hp.com>:
> Code not merged yet.
    :
> it's old and needs to die properly.  i haven't had time to fix that yet.

Thanks for the information.  Actually the parisc tree is one of the ones
that leaks the fewest of these symbols... 
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

Ideology, politics and journalism, which luxuriate in failure, are
impotent in the face of hope and joy.
	-- P. J. O'Rourke

[parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Alan Cox]

> CONFIG_BINFMT_SOM: arch/parisc/config.in arch/parisc/defconfig
> Not used in code anywhere.  Can you get rid of this one?

Its used in the parisc tree as are most of the others you see. You probably want
to simply skip processing arch/parisc

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[David Woodhouse]

nico@cam.org said:
> Therefore it's the maintainer's job to submit coherent patches and
> accept to see inconsistent CONFIG_* references be removed from the
> official tree until further patch submission is due. 

Maybe. But you tend to include the latest MTD code in your tree, for
example, and hence the defconfigs have the new options in. Is it really
worth your time to produce separate defconfigs without it before feeding 
your changes upstream?


> Otherwise how can you distinguish between dead wood which must be
> removed and potentially valid symbols referring to code existing only
> in a remote tree?

By periodically publishing a list of the potentially-obsolete symbols as ESR
has done, and _not_ removing the ones which people speak up about. It's not
as if this is something which needs to be done more than about once a year.

--
dwmw2

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Andreas Dilger]

Russell King writes:
> - Secondly, its very easy to miss stuff in the lkml hunk of email each
>   day when you have less than 4 hours to read it and think about it.
>   (note that architecture maintainers have to read mail from their
>   side which may not be on lkml, think about that, think about bug fixes,
>   possible impacts of fixes on other machines, etc etc).  Therefore,
>   copying their email address registered in the MAINTAINER file means
>   that they should not overlook your patch.

One of the issues for contacting each MAINTAINER is that this information
is out-of-line from the actual kernel tree.  The other is that the
description of what a maintainer is actually controlling is somewhat
vague.

How about the following:
- each directory has a MAINTAINERS file which lists parties with a
  vested interest in files in that directory (format is mostly the
  same as current)
- subdirectories which don't have a MAINTAINERS file use the MAINTAINERS
  file of the parent (or grandparent) directory
- each maintainer entry explicitly lists each file/directory that this
  person is interested in, maybe "F: {file | directory} ...".

I'm sure Eric can come up with a simple program to parse the MAINTAINER
file/tree.  If the program takes a kernel-tree relative filename and
spit out the name/email of the relevant maintainer (subsystem and port
specific mailing lists should also be included), that would make the 
job of finding out who to send patches to a whole lot easier.

My one gripe about the MAINTAINERS file is that it still lists Remy
Card as EXT2 maintainer, so we would probably need to do a find on
the whole kernel tree, email each address a list of files that they
"maintain" and wait until they complain, agree, or time out.  Once
the database is up-to-date, it simplifies the job of keeping maintainers
(and other interested parties) in the loop.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

David Woodhouse <dwmw2@infradead.org>:
> > Otherwise how can you distinguish between dead wood which must be
> > removed and potentially valid symbols referring to code existing only
> > in a remote tree?
> 
> By periodically publishing a list of the potentially-obsolete symbols as ESR
> has done, and _not_ removing the ones which people speak up about. It's not
> as if this is something which needs to be done more than about once a year.

Not good enough.  In a year, the pile of false positives would get high enough
to make it too hard to spot real bugs like the Aironet mismatch.  The whole 
point of the cleanup is to be able to mechanize the consistency checks so they
require a minimum of human judgment.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

"This country, with its institutions, belongs to the people who
inhabit it. Whenever they shall grow weary of the existing government,
they can exercise their constitutional right of amending it or their
revolutionary right to dismember it or overthrow it."
	-- Abraham Lincoln, 4 April 1861

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[David Woodhouse]

esr@thyrsus.com said:
>  Not good enough.  In a year, the pile of false positives would get
> high enough to make it too hard to spot real bugs like the Aironet
> mismatch.  The whole  point of the cleanup is to be able to mechanize
> the consistency checks so they require a minimum of human judgment.

I'm not sure that's the case. The nature of the false positives is that 
they're generally _temporary_ aberrations, caused by the loss of 
synchronisation of various maintainers w.r.t submitting patches to Linus.

I'd be very surprised if the number of false positives isn't fairly stable, 
with new ones being introduced at a similar rate to the rate at which old 
ones finally become correct. 

Might be interesting to check a few older kernels to see if this is true. 
Actually I might expect it to be roughly proportional to the number of 
separately-maintained bodies of code - so it'll grow over time, as the size 
of the Linux kernel grows. 

--
dwmw2

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

David Woodhouse <dwmw2@infradead.org>:
> I'd be very surprised if the number of false positives isn't fairly stable, 
> with new ones being introduced at a similar rate to the rate at which old 
> ones finally become correct. 

Even supposing that's so, a 36% rate of broken symbols is way too high. 
It argues that we need to do a thorough housecleaning at least once in
order to get back to an acceptably low stable rate.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

The same applies for other kinds of long-lasting low-level pain. [...]
The body's response to being jabbed, pierced, and cut is to produce
endorphins. [...]  So here's my programme for breaking that cycle of
dependency on Windows: get left arm tattooed with dragon motif, buy a
crate of Jamaican Hot! Pepper Sauce, get nipples pierced.  With any
luck that will produce enough endorphins to make Windows completely
redundant, and I can then upgrade to Linux and get on with things.
	-- Pieter Hintjens

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[David Woodhouse]

esr@thyrsus.com said:
>  Even supposing that's so, a 36% rate of broken symbols is way too
> high.  It argues that we need to do a thorough housecleaning at least
> once in order to get back to an acceptably low stable rate.

Accepted. Can we let the 2.4 "angry penguin"-enforced stabilising period
finish, and give the arch and subsystem maintainers a chance to finally
brave the wrath of Linus and submit their patches, before we attempt do to
it though?

--
dwmw2

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Jeff Dike]

matthew@wil.cx said:
> Have you tried mailing webmaster@kernel.org and asking to be added? 

Yes.

jes@linuxcare.com said:
> I'd be highly surprised if they said no to adding UML to the list if
> you mailed them a request to update the page. 

Well, be surprised then.  The reply from hpa was that that list was for 
processor ports.  He did say that there might at some point in the future be a 
separate list (off the main page) of other things.

				Jeff

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Alan Cox]

> Even supposing that's so, a 36% rate of broken symbols is way too high. 
> It argues that we need to do a thorough housecleaning at least once in
> order to get back to an acceptably low stable rate.

Many of your 'broken' symbols arent. We have no idea what the real amount is

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

David Woodhouse <dwmw2@infradead.org>:
> esr@thyrsus.com said:
> >  Even supposing that's so, a 36% rate of broken symbols is way too
> > high.  It argues that we need to do a thorough housecleaning at least
> > once in order to get back to an acceptably low stable rate.
> 
> Accepted. Can we let the 2.4 "angry penguin"-enforced stabilising period
> finish, and give the arch and subsystem maintainers a chance to finally
> brave the wrath of Linus and submit their patches, before we attempt do to
> it though?

I guess so.  We don't particularly have a choice, do we? :-)
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

"Government is not reason, it is not eloquence, it is force; like fire, a
troublesome servant and a fearful master. Never for a moment should it be left
to irresponsible action."
	-- George Washington, in a speech of January 7, 1790

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Alan Cox <alan@lxorguk.ukuu.org.uk>:
> > Even supposing that's so, a 36% rate of broken symbols is way too high. 
> > It argues that we need to do a thorough housecleaning at least once in
> > order to get back to an acceptably low stable rate.
> 
> Many of your 'broken' symbols arent. We have no idea what the real amount is

If it can't be mechanically verified that the symbol has a correct reference 
pattern within the tree, then it's broken.  That's a definition.

The fact that it might become un-broken someday, by somebody's
intention to merge in future code, is interesting but irrelevant to
the fact that symbols broken in present time *mask bugs* in present time.

I'm not being a compulsive neatnik -- that wouldn't be worth my time.  What I'm
trying to do is reduce the number of crevices in which bugs can hide.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

This would be the best of all possible worlds, if there were
no religion in it.
	-- John Adams, in a letter to Thomas Jefferson.

[parisc-linux] Proposal for better attribution structure

[Eric S. Raymond]

Andreas Dilger <adilger@turbolinux.com>:
> One of the issues for contacting each MAINTAINER is that this information
> is out-of-line from the actual kernel tree.  The other is that the
> description of what a maintainer is actually controlling is somewhat
> vague.

I strongly agree.  I first tripped over this problem when I was trying
to identify the responsible parties for [Cc]onfig.in files.  It's
biting me again now that I'm trying to clean up the CONFIG_ space.
It's one that's going to cause grief for anybody trying to do *global*
work on the kernel, stuff that crosses boundaries between maintainer
jurisdictions.  
 
> How about the following:
> - each directory has a MAINTAINERS file which lists parties with a
>   vested interest in files in that directory (format is mostly the
>   same as current)
> - subdirectories which don't have a MAINTAINERS file use the MAINTAINERS
>   file of the parent (or grandparent) directory
> - each maintainer entry explicitly lists each file/directory that this
>   person is interested in, maybe "F: {file | directory} ...".
> 
> I'm sure Eric can come up with a simple program to parse the MAINTAINER
> file/tree.  If the program takes a kernel-tree relative filename and
> spit out the name/email of the relevant maintainer (subsystem and port
> specific mailing lists should also be included), that would make the 
> job of finding out who to send patches to a whole lot easier.

The spirit of this proposal is, IMO, excellent.  I like the idea that if
maintainer information for a particular piece of the hierarchy doesn't
exist, you float up to the next higher level.  Search always ends at
the root MAINTAINERS file.

And I could indeed write a program such as Andreas describes, and would
be most willing to do so.

I have one objection, however.  I think the maintainers information
should normally be inline of the file in question, so there won't
be a need for an explicit F: link that could become invalid.  So I
think the search order should look like this:

	1. Look for maintainer markup in the file itself.
	2. Then look for a NAINTAINERS file in the current directory.
	3. Then look upwards for MAINTAINERS files in enclosing directories.

> My one gripe about the MAINTAINERS file is that it still lists Remy
> Card as EXT2 maintainer, so we would probably need to do a find on
> the whole kernel tree, email each address a list of files that they
> "maintain" and wait until they complain, agree, or time out.  Once
> the database is up-to-date, it simplifies the job of keeping maintainers
> (and other interested parties) in the loop.

I have until 6 May at least to work on this, if there is consensus that it's
a good idea.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

Where rights secured by the Constitution are involved, there can be no
rule making or legislation which would abrogate them.
        -- Miranda vs. Arizona, 384 US 436 p. 491

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Tom Leete]

[Cc: trimmed]

Russell King wrote:
> 
[...]
> 
> Generally it seems like diff needs to produce one more line of context, and
> most of these problems will go away.  Yes, there will still be the odd
> problem, so then it becomes the "how much do you crank the setting" problem.
>
 
$ diff -6 ...
will give 6 lines of context. patch will understand the output without any
extra help.

Cheers,
Tom

-- 
The Daemons lurk and are dumb. -- Emerson

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Grant Grundler]

"Eric S. Raymond" wrote:
> Here's what I have for you guys:

...
> CONFIG_DMB_TRAP: arch/parisc/kernel/sba_iommu.c
> CONFIG_FUNC_SIZE: arch/parisc/kernel/sba_iommu.c
> 
> Would you please take these out of the CONFIG_ namespace?  Changing the 
> prefix to CONFIGURE would do nicely.

As willy noted, both mine. I'll remove or rename them rename them so
they aren't in the CONFIG_ name space. Probably s/CONFIG_/SBA_/ for
those two.

I'm going to submit a "wishlist" bug to our debian BTS
(bugs.parisc-linux.org) for "Data Memory Break Trap" support.
It's a damn good Hammer! :^)
(GDB will probably want to use this too)

I once had a working "Data Memory Break Trap" handler to catch other
parts of the kernel when they corrupted the IO Pdirs. Hooks in sba_ccio.c
helped mark which pages would trap and define which code was allowed to
touch the page. My implementation had issues and I never bothered to
re-implement as suggested by our parisc CPU god, John Marvin.

CONFIG_FUNC_SIZE is just a bad choice of name (asking for trouble).
One might consider this a bug that hasn't happened yet - thanks Eric!

#define CONFIG_FUNC_SIZE 4096   /* SBA configuration function reg set */


> CONFIG_KWDB: arch/parisc/Makefile arch/parisc/config.in arch/parisc/defconfig
>    arch/parisc/kernel/entry.S arch/parisc/kernel/traps.c arch/parisc/mm/init.
>   c

This ones actually mine too. It could be replaced with the SGI debugger
CONFIG option if/when that ever gets supported. The hooks will have to
be in the same place. I'm pretty sure now the HP KWBD team will never give me
permission to publish KWDB sources (they've had almost a year now).
I sorta almost had the damn thing working too...*sigh*.
Willy should do whatever he thinks is right in this case.

> CONFIG_PCI_LBA: arch/parisc/config.in arch/parisc/defconfig arch/parisc/kerne
>   l/Makefile
...
> Looks like these need Configure.help entries.

That's mine too.
We've been lazy about documentation since the getting the code working
has been a higher priority.  I think having them documented will be a
prerequisite to merging upstream (either to Alan Cox or Linus).

thanks,
grant

Grant Grundler
parisc-linux {PCI|IOMMU|SMP} hacker
+1.408.447.7253

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone

[rmk]

Tom Leete writes:
> $ diff -6 ...
> will give 6 lines of context. patch will understand the output without any
> extra help.

Indeed, but I can't do that to a patch that Alan or Linus produces.

--
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[David Woodhouse]

esr@thyrsus.com said:
> If it can't be mechanically verified that the symbol has a correct
> reference  pattern within the tree, then it's broken.  That's a
> definition. 

Here's an alternative definition:

If the symbol has the letters 'F', 'I', 'S' and 'H' in it, in any order, 
then it's broken.

That's also a definition. It's not a particularly useful one, but neither 
was yours.

/me looks for a way to equate the original definition with the halting 
problem :)

--
dwmw2

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Eric S. Raymond]

Grant Grundler <grundler@puffin.external.hp.com>:
> One might consider this a bug that hasn't happened yet - thanks Eric!

Thank you very much for your cooperation.  This is the third real problem that
the CONFIG_ namespace audit has turned up, and a good example of the sort of
thing I have been hoping to accomplish with it.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

(Those) who are trying to read the Second Amendment out of the Constitution by
claiming it's not an individual right (are) courting disaster by encouraging
others to use the same means to eliminate portions of the Constitution they
don't like.
	-- Alan Dershowitz, Harvard Law School

Re: [parisc-linux] Re: OK, let's try cleaning up another nit. Is anyone paying attention?

[Jes Sorensen]

>>>>> "Eric" == Eric S Raymond <esr@thyrsus.com> writes:

Eric> Alan Cox <alan@lxorguk.ukuu.org.uk>:
>> Many of your 'broken' symbols arent. We have no idea what the real
>> amount is

Eric> If it can't be mechanically verified that the symbol has a
Eric> correct reference pattern within the tree, then it's broken.
Eric> That's a definition.

It's a definition but not necessarily the best one to follow.

Eric> The fact that it might become un-broken someday, by somebody's
Eric> intention to merge in future code, is interesting but irrelevant
Eric> to the fact that symbols broken in present time *mask bugs* in
Eric> present time.

Symbols that are not referenced at all by the code does not hide any
bugs. They might make it take longer time for people to configure
their kernel but thats about it.

This does not mean that obsolete symbols should not be removed,
however running around telling people to remove symbols that they
might be using in their tree does cause unnecessary work for the
people who are writing the code.

Jes