Re: CML2 funkiness

Re: CML2 funkiness

[Andrew Rodland]

First off, I'd like to apologize for lack of all the
 information I'd like to have, I'm at school, and
 temporarily semidisconnected at home.

CML2 is definitely still not quite right for me
(2.4.17 + kpreempt-rml, latest CML2 as of 3ish days ago).

Menuconfig and friends seem okay, as far as I can tell (and
 they've apparently been tested pretty well), but oldconfig
 is wacky...

Basically, it seems to have random (but deterministic)
 amnesia: It forgets the answers to certain questions,
 apparently on write-out.

So, "mv config .config ; make mrproper ; mv config .config
 ; make oldconfig" does odd things to my config, but more
 in-your-face, on "make oldconfig ; make oldconfig" (ad
 inifinitum if you want), it will continue asking the same
 questions, and never remember the answer.

I'm 99% sure the problem is on the write-out, rather than
 the read in, but I'll go do some extra digging tonight.
 Python isn't really my language. Yet, at least. :)

Anyway, thanks for reading. If you need more information,
 let me know. I'm subscribed to the 100k digest.

--Andrew Rodland

Re: CML2 funkiness

[Eric S. Raymond]

David Relson <relson@osagesoftware.com>:
>  From past testing of CML2 I know it uses file config.out as its 
> "memory".  Looking in it, I didn't see any CONFIG symbols for these symbols.
> 
> There's definitely something here for Eric to fix!

I'm on it.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

Live free or die; death is not the worst of evils.
	-- General George Stark.

Re: CML2 funkiness

[David Relson]

At 09:03 AM 1/2/02, Andrew Rodland wrote:
>First off, I'd like to apologize for lack of all the
>  information I'd like to have, I'm at school, and
>  temporarily semidisconnected at home.
>
>CML2 is definitely still not quite right for me
>(2.4.17 + kpreempt-rml, latest CML2 as of 3ish days ago).
>
>Menuconfig and friends seem okay, as far as I can tell (and
>  they've apparently been tested pretty well), but oldconfig
>  is wacky...
>
>So, "mv config .config ; make mrproper ; mv config .config
>  ; make oldconfig" does odd things to my config, but more
>  in-your-face, on "make oldconfig ; make oldconfig" (ad
>  inifinitum if you want), it will continue asking the same
>  questions, and never remember the answer.

Andrew,

I have just tested this, and have reproduced your problem.  Using 
kernel-2.4.16 and cml2-1.2.20, i.e. my current kernel and the latest CML2, 
I ran "make oldconfig" three times.  The first time I answered "n" to 21 
queries.  The second and third times, I had to answer "n" to 9 
queries.  The 9 all appeared in the first run and were exactly the same in 
the second and third runs.

Here're the 9 queries from runs 2 and 3:
EXPERT: Prompt for expert choices (those with no help attached) 
(EXPERIMENTAL) [ ] (NEW)?:
DEVELOPMENT: Configure a development or 2.5 kernel? (EXPERIMENTAL) [ ] (NEW)?:
CD_NO_IDESCSI: Support CD-ROM drives that are not SCSI or IDE/ATAPI [ ] 
(NEW)?:
IP_ADVANCED_ROUTER: Advanced router [ ] (NEW)?:
NET_VENDOR_SMC: Western Digital/SMC cards [ ] (NEW)?:
NET_VENDOR_RACAL: Racal-Interlan (Micom) NI cards [ ] (NEW)?:
NET_POCKET: Pocket and portable adapters [ ] (NEW)?:
HAMRADIO: Amateur Radio support [ ] (NEW)?:
FBCON_FONTS: Select other compiled-in fonts [ ] (NEW)?:

 From past testing of CML2 I know it uses file config.out as its 
"memory".  Looking in it, I didn't see any CONFIG symbols for these symbols.

There's definitely something here for Eric to fix!

David

Re: CML2 funkiness

[Andrew Rodland]

Actually, I had 1.9.16, but it looks like the same problem
 (and it seems like it's been around for a bit).
I'm definitely seeing the same thing as David, except the
 symbols I'm seeing are:

DANGEROUS: Prompt for features that can trash data.
 (DANGEROUS) [ ] (NEW)?:
DEVELOPMENT: Configure a development or 2.5 kernel?
 (EXPERIMENTAL) [ ] (NEW)?:
ISA_CARDS: Support for ISA-bus cards. [Y] (NEW)?:


On Wed, 02 Jan 2002 10:10:42 -0500
 David Relson <relson@osagesoftware.com> wrote:
> At 09:03 AM 1/2/02, Andrew Rodland wrote:
> >First off, I'd like to apologize for lack of all the
> > information I'd like to have, I'm at school, and
> > temporarily semidisconnected at home.
> >
> >CML2 is definitely still not quite right for me
> >(2.4.17 + kpreempt-rml, latest CML2 as of 3ish days
>  ago).
> >
> >Menuconfig and friends seem okay, as far as I can tell
>  (and
> > they've apparently been tested pretty well), but
>  oldconfig
> > is wacky...
> >
> >So, "mv config .config ; make mrproper ; mv config
>  .config
> > ; make oldconfig" does odd things to my config, but
>  more
> > in-your-face, on "make oldconfig ; make oldconfig" (ad
> > inifinitum if you want), it will continue asking the
>  same
> > questions, and never remember the answer.
> 
> Andrew,
> 
> I have just tested this, and have reproduced your
>  problem.  Using kernel-2.4.16 and cml2-1.2.20, i.e. my
>  current kernel and the latest CML2, I ran "make
>  oldconfig" three times.  The first time I answered "n"
>  to 21 queries.  The second and third times, I had to
>  answer "n" to 9 queries.  The 9 all appeared in the
>  first run and were exactly the same in the second and
>  third runs.
> 
> Here're the 9 queries from runs 2 and 3:
> EXPERT: Prompt for expert choices (those with no help
>  attached) (EXPERIMENTAL) [ ] (NEW)?:
> DEVELOPMENT: Configure a development or 2.5 kernel?
>  (EXPERIMENTAL) [ ] (NEW)?:
> CD_NO_IDESCSI: Support CD-ROM drives that are not SCSI or
>  IDE/ATAPI [ ] (NEW)?:
> IP_ADVANCED_ROUTER: Advanced router [ ] (NEW)?:
> NET_VENDOR_SMC: Western Digital/SMC cards [ ] (NEW)?:
> NET_VENDOR_RACAL: Racal-Interlan (Micom) NI cards [ ]
>  (NEW)?:
> NET_POCKET: Pocket and portable adapters [ ] (NEW)?:
> HAMRADIO: Amateur Radio support [ ] (NEW)?:
> FBCON_FONTS: Select other compiled-in fonts [ ] (NEW)?:
> 
> From past testing of CML2 I know it uses file config.out
>  as its 
> "memory".  Looking in it, I didn't see any CONFIG symbols
>  for these symbols.
> 
> There's definitely something here for Eric to fix!
> 
> David
> 
> 

Re: CML2 funkiness

[Andrew Rodland]

Actually, I had 1.9.16, but it looks like the same problem
 (and it seems like it's been around for a bit).
I'm definitely seeing the same thing as David, except the
 symbols I'm seeing are:

DANGEROUS: Prompt for features that can trash data.
 (DANGEROUS) [ ] (NEW)?:
DEVELOPMENT: Configure a development or 2.5 kernel?
 (EXPERIMENTAL) [ ] (NEW)?:
ISA_CARDS: Support for ISA-bus cards. [Y] (NEW)?:


On Wed, 02 Jan 2002 10:10:42 -0500
 David Relson <relson@osagesoftware.com> wrote:
> At 09:03 AM 1/2/02, Andrew Rodland wrote:
> >First off, I'd like to apologize for lack of all the
> > information I'd like to have, I'm at school, and
> > temporarily semidisconnected at home.
> >
> >CML2 is definitely still not quite right for me
> >(2.4.17 + kpreempt-rml, latest CML2 as of 3ish days
>  ago).
> >
> >Menuconfig and friends seem okay, as far as I can tell
>  (and
> > they've apparently been tested pretty well), but
>  oldconfig
> > is wacky...
> >
> >So, "mv config .config ; make mrproper ; mv config
>  .config
> > ; make oldconfig" does odd things to my config, but
>  more
> > in-your-face, on "make oldconfig ; make oldconfig" (ad
> > inifinitum if you want), it will continue asking the
>  same
> > questions, and never remember the answer.
> 
> Andrew,
> 
> I have just tested this, and have reproduced your
>  problem.  Using kernel-2.4.16 and cml2-1.2.20, i.e. my
>  current kernel and the latest CML2, I ran "make
>  oldconfig" three times.  The first time I answered "n"
>  to 21 queries.  The second and third times, I had to
>  answer "n" to 9 queries.  The 9 all appeared in the
>  first run and were exactly the same in the second and
>  third runs.
> 
> Here're the 9 queries from runs 2 and 3:
> EXPERT: Prompt for expert choices (those with no help
>  attached) (EXPERIMENTAL) [ ] (NEW)?:
> DEVELOPMENT: Configure a development or 2.5 kernel?
>  (EXPERIMENTAL) [ ] (NEW)?:
> CD_NO_IDESCSI: Support CD-ROM drives that are not SCSI or
>  IDE/ATAPI [ ] (NEW)?:
> IP_ADVANCED_ROUTER: Advanced router [ ] (NEW)?:
> NET_VENDOR_SMC: Western Digital/SMC cards [ ] (NEW)?:
> NET_VENDOR_RACAL: Racal-Interlan (Micom) NI cards [ ]
>  (NEW)?:
> NET_POCKET: Pocket and portable adapters [ ] (NEW)?:
> HAMRADIO: Amateur Radio support [ ] (NEW)?:
> FBCON_FONTS: Select other compiled-in fonts [ ] (NEW)?:
> 
> From past testing of CML2 I know it uses file config.out
>  as its 
> "memory".  Looking in it, I didn't see any CONFIG symbols
>  for these symbols.
> 
> There's definitely something here for Eric to fix!
> 
> David
> 
> 

Re: CML2 funkiness

[Andrew Rodland]

Actually, I had 1.9.16, but it looks like the same problem
 (and it seems like it's been around for a bit).
I'm definitely seeing the same thing as David, except the
 symbols I'm seeing are:

DANGEROUS: Prompt for features that can trash data.
 (DANGEROUS) [ ] (NEW)?:
DEVELOPMENT: Configure a development or 2.5 kernel?
 (EXPERIMENTAL) [ ] (NEW)?:
ISA_CARDS: Support for ISA-bus cards. [Y] (NEW)?:


On Wed, 02 Jan 2002 10:10:42 -0500
 David Relson <relson@osagesoftware.com> wrote:
> At 09:03 AM 1/2/02, Andrew Rodland wrote:
> >First off, I'd like to apologize for lack of all the
> > information I'd like to have, I'm at school, and
> > temporarily semidisconnected at home.
> >
> >CML2 is definitely still not quite right for me
> >(2.4.17 + kpreempt-rml, latest CML2 as of 3ish days
>  ago).
> >
> >Menuconfig and friends seem okay, as far as I can tell
>  (and
> > they've apparently been tested pretty well), but
>  oldconfig
> > is wacky...
> >
> >So, "mv config .config ; make mrproper ; mv config
>  .config
> > ; make oldconfig" does odd things to my config, but
>  more
> > in-your-face, on "make oldconfig ; make oldconfig" (ad
> > inifinitum if you want), it will continue asking the
>  same
> > questions, and never remember the answer.
> 
> Andrew,
> 
> I have just tested this, and have reproduced your
>  problem.  Using kernel-2.4.16 and cml2-1.2.20, i.e. my
>  current kernel and the latest CML2, I ran "make
>  oldconfig" three times.  The first time I answered "n"
>  to 21 queries.  The second and third times, I had to
>  answer "n" to 9 queries.  The 9 all appeared in the
>  first run and were exactly the same in the second and
>  third runs.
> 
> Here're the 9 queries from runs 2 and 3:
> EXPERT: Prompt for expert choices (those with no help
>  attached) (EXPERIMENTAL) [ ] (NEW)?:
> DEVELOPMENT: Configure a development or 2.5 kernel?
>  (EXPERIMENTAL) [ ] (NEW)?:
> CD_NO_IDESCSI: Support CD-ROM drives that are not SCSI or
>  IDE/ATAPI [ ] (NEW)?:
> IP_ADVANCED_ROUTER: Advanced router [ ] (NEW)?:
> NET_VENDOR_SMC: Western Digital/SMC cards [ ] (NEW)?:
> NET_VENDOR_RACAL: Racal-Interlan (Micom) NI cards [ ]
>  (NEW)?:
> NET_POCKET: Pocket and portable adapters [ ] (NEW)?:
> HAMRADIO: Amateur Radio support [ ] (NEW)?:
> FBCON_FONTS: Select other compiled-in fonts [ ] (NEW)?:
> 
> From past testing of CML2 I know it uses file config.out
>  as its 
> "memory".  Looking in it, I didn't see any CONFIG symbols
>  for these symbols.
> 
> There's definitely something here for Eric to fix!
> 
> David
> 
> 

Re: CML2 funkiness

[Andrew Rodland]

Actually, it turns out I had CML2-1.9.16, but:
It looks like I'm definitely seeing the same thing as
 David, except in my case the symbols are:

DANGEROUS DEVELOPMENT ISA_CARDS CD_NO_IDESCSI
 SERIAL_NONSTANDARD SCSI_PCMCIA IP_ADVANCED_ROUTER
 NET_VENDOR_3COM NET_VENDOR_SMC NET_VENDOR_RACAL NET_POCKET
 HAMRADIO FBCON_FONTS DONGLE
.
Not sure if it's a cause or an effect, but all of these
 symbols get marked as (NEW).
Looks like it's certain fixed symbols, as most/all of
 David's are in mine too.

Sorry for the lame web-mailer, and thanks
--Andrew Rodland

On Wed, 02 Jan 2002 10:10:42 -0500
 David Relson <relson@osagesoftware.com> wrote:
> At 09:03 AM 1/2/02, Andrew Rodland wrote:
> >First off, I'd like to apologize for lack of all the
> > information I'd like to have, I'm at school, and
> > temporarily semidisconnected at home.
> >
> >CML2 is definitely still not quite right for me
> >(2.4.17 + kpreempt-rml, latest CML2 as of 3ish days
>  ago).
> >
> >Menuconfig and friends seem okay, as far as I can tell
>  (and
> > they've apparently been tested pretty well), but
>  oldconfig
> > is wacky...
> >
> >So, "mv config .config ; make mrproper ; mv config
>  .config
> > ; make oldconfig" does odd things to my config, but
>  more
> > in-your-face, on "make oldconfig ; make oldconfig" (ad
> > inifinitum if you want), it will continue asking the
>  same
> > questions, and never remember the answer.
> 
> Andrew,
> 
> I have just tested this, and have reproduced your
>  problem.  Using kernel-2.4.16 and cml2-1.2.20, i.e. my
>  current kernel and the latest CML2, I ran "make
>  oldconfig" three times.  The first time I answered "n"
>  to 21 queries.  The second and third times, I had to
>  answer "n" to 9 queries.  The 9 all appeared in the
>  first run and were exactly the same in the second and
>  third runs.
> 
> Here're the 9 queries from runs 2 and 3:
> EXPERT: Prompt for expert choices (those with no help
>  attached) (EXPERIMENTAL) [ ] (NEW)?:
> DEVELOPMENT: Configure a development or 2.5 kernel?
>  (EXPERIMENTAL) [ ] (NEW)?:
> CD_NO_IDESCSI: Support CD-ROM drives that are not SCSI or
>  IDE/ATAPI [ ] (NEW)?:
> IP_ADVANCED_ROUTER: Advanced router [ ] (NEW)?:
> NET_VENDOR_SMC: Western Digital/SMC cards [ ] (NEW)?:
> NET_VENDOR_RACAL: Racal-Interlan (Micom) NI cards [ ]
>  (NEW)?:
> NET_POCKET: Pocket and portable adapters [ ] (NEW)?:
> HAMRADIO: Amateur Radio support [ ] (NEW)?:
> FBCON_FONTS: Select other compiled-in fonts [ ] (NEW)?:
> 
> From past testing of CML2 I know it uses file config.out
>  as its 
> "memory".  Looking in it, I didn't see any CONFIG symbols
>  for these symbols.
> 
> There's definitely something here for Eric to fix!
> 
> David
> 
> 

Re: CML2 funkiness

[Miles Lane]

I am seeing other problems.  Using CML2 with a 2.4.18-pre1
tree, I was unable to create a kernel that would boot.
It kept crashing with messages stating that the rivafb driver
did not support 8-bit color depth.  I tried tweaking my
configuration for a while, but finally reverted to CML2
and was then able to get a working kernel.  I'll investigate
further and send along a diff of the working and broken
configuration files.

Another thing I notice is that when I create a configuration
using CML2, then switch to CML1 and run "make oldconfig"
using the same kernel tree, it appears there are configuration
options that never got set in the CML2 .config file.
I suppose this may simply be due to CML2 writing out the 
options in a different order.

	Miles

Re: CML2 funkiness

[Andrew Rodland]

That reminds me, I use vesafb, and I get a just plain black
 screen on boot, if I use CML2 to configure the kernel. I'd
 assume it's just because there are one or two FB-related
 symbols that seem to be getting dropped.

On 02 Jan 2002 17:42:47 -0800
 Miles Lane <miles@megapathdsl.net> wrote:
> I am seeing other problems.  Using CML2 with a
>  2.4.18-pre1
> tree, I was unable to create a kernel that would boot.
> It kept crashing with messages stating that the rivafb
>  driver
> did not support 8-bit color depth.  I tried tweaking my
> configuration for a while, but finally reverted to CML2
> and was then able to get a working kernel.  I'll
>  investigate
> further and send along a diff of the working and broken
> configuration files.
> 
> Another thing I notice is that when I create a
>  configuration
> using CML2, then switch to CML1 and run "make oldconfig"
> using the same kernel tree, it appears there are
>  configuration
> options that never got set in the CML2 .config file.
> I suppose this may simply be due to CML2 writing out the 
> options in a different order.
> 
> 	Miles
> 

Re: CML2 funkiness

[Andrew Rodland]

I found it!
The culprit: a bit of confusion over 'private'.
All of the not-saved symbols were just guards for 'do we
 want to display question X'... so they were marked
 private, so as not to clutter up the kernel (I assume).
 However, this prevents them from getting written to
 .config/config.out as well! Easy fix is to un-private
 them, long-term is (as I see it) either to create a new
 equivalent to private that somehow lets the symbol get
 written to defconfig, but prevents it from becoming a
 kernel define, or just to blow it off and don't worry
 about it, and leave them normal symbols. However, it's
 1:30AM and I might be missing something.

--Andrew

On Wed, 2 Jan 2002 10:03:11 -0500
 "Eric S. Raymond" <esr@thyrsus.com> wrote:
> David Relson <relson@osagesoftware.com>:
> >  From past testing of CML2 I know it uses file
>  config.out as its 
> > "memory".  Looking in it, I didn't see any CONFIG
>  symbols for these symbols.
> > 
> > There's definitely something here for Eric to fix!
> 
> I'm on it.
> -- 
> 		<a href="http://www.tuxedo.org/~esr/">Eric S.
>  Raymond</a>
> 
> Live free or die; death is not the worst of evils.
> 	-- General George Stark.

Re: CML2 funkiness

[Eric S. Raymond]

Andrew Rodland <arodland@noln.com>:
> I found it!
> The culprit: a bit of confusion over 'private'.
> All of the not-saved symbols were just guards for 'do we
>  want to display question X'... so they were marked
>  private, so as not to clutter up the kernel (I assume).
>  However, this prevents them from getting written to
>  .config/config.out as well! Easy fix is to un-private
>  them, long-term is (as I see it) either to create a new
>  equivalent to private that somehow lets the symbol get
>  written to defconfig, but prevents it from becoming a
>  kernel define, or just to blow it off and don't worry
>  about it, and leave them normal symbols. However, it's
>  1:30AM and I might be missing something.

I think the right thing will be to write private symbols into config.out,
but with an attached PRIVATE label that stops configtrans.py from 
translating these into defines for the autoconf.h file.

That's what I've done for the upcoming CML 2.0.0 release.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

Question with boldness even the existence of a God; because, if there
be one, he must more approve the homage of reason, than that of
blindfolded fear.... Do not be frightened from this inquiry from any
fear of its consequences. If it ends in the belief that there is no
God, you will find incitements to virtue in the comfort and
pleasantness you feel in its exercise...
	-- Thomas Jefferson, in a 1787 letter to his nephew

Re: CML2 funkiness

[Eric S. Raymond]

Miles Lane <miles@megapathdsl.net>:
> I am seeing other problems.  Using CML2 with a 2.4.18-pre1
> tree, I was unable to create a kernel that would boot.
> It kept crashing with messages stating that the rivafb driver
> did not support 8-bit color depth.  I tried tweaking my
> configuration for a while, but finally reverted to CML2
> and was then able to get a working kernel.  I'll investigate
> further and send along a diff of the working and broken
> configuration files.

This sounds like a rulebase bug.  That diff should pin it down.
-- 
		<a href="http://www.tuxedo.org/~esr/">Eric S. Raymond</a>

The spirit of resistance to government is so valuable on certain occasions, 
that I wish it always to be kept alive.  It will often be exercised when 
wrong, but better so than not to be exercised at all. I like a little 
rebellion now and then.	-- Thomas Jefferson, letter to Abigail Adams, 1787

Re: CML2 funkiness

[Andrew Rodland]

After sleep and further thought, that makes perfect sense.

On Thu, 3 Jan 2002 02:31:05 -0500
 "Eric S. Raymond" <esr@thyrsus.com> wrote:
> I think the right thing will be to write private symbols
>  into config.out,
> but with an attached PRIVATE label that stops
>  configtrans.py from 
> translating these into defines for the autoconf.h file.
> 
> That's what I've done for the upcoming CML 2.0.0 release.
> --