Re; MTD (2) Block Driver broken

Re; MTD (2) Block Driver broken

[Gareth Bult]

Hi,

A little while back I re-wrote blkmtd.c effectively to provide proper
"write" access to jffs2 filesystems mounted on standard Linux block
devices.

The driver seems to have worked well (for me) and after many thousands
of downloads of FlashLinux, I've not had any reports of JFFS2 problems
or corruption.

It appears that someone has "cleaned up" this driver and it now appears
as an option in the standard kernel tree for 2.6.11.

Unfortunately it doesn't appear to work. As soon as you mount a JFFS2
filesystem on it and try to write data, corruption sets in resulting is
a fairly sharp system crash.

Copying back the source from the original driver I submitted to this
list and recompiling the kernel (with a one line change) appears to
yield a working system again.

So, two questions;
a.	Did anyone actually test the code after cleaning it up?
b.	Were there any life changing improvements that I really
	*need* to make to the original code?

tia
Gareth

-- 
Gareth Bult <gareth@linux.co.uk>

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Sat, 5 March 2005 17:26:09 +0000, Gareth Bult wrote:
> 
> A little while back I re-wrote blkmtd.c effectively to provide proper
> "write" access to jffs2 filesystems mounted on standard Linux block
> devices.
> 
> The driver seems to have worked well (for me) and after many thousands
> of downloads of FlashLinux, I've not had any reports of JFFS2 problems
> or corruption.
> 
> It appears that someone has "cleaned up" this driver and it now appears
> as an option in the standard kernel tree for 2.6.11.
> 
> Unfortunately it doesn't appear to work. As soon as you mount a JFFS2
> filesystem on it and try to write data, corruption sets in resulting is
> a fairly sharp system crash.
> 
> Copying back the source from the original driver I submitted to this
> list and recompiling the kernel (with a one line change) appears to
> yield a working system again.
> 
> So, two questions;
> a.	Did anyone actually test the code after cleaning it up?
> b.	Were there any life changing improvements that I really
> 	*need* to make to the original code?

Gareth, check your archives.

I sent the whole list of cleanup patches to you, you could have simply
tested them in order and told me which one broke.  Instead you got
pissed at dwmw2 since he didn't accept your broken mails and
explicitly told me various times, you didn't care about the driver.

The fact that you don't remember me, you don't remember being on Cc:
when I sent patches to Andrew Morton and now come running back is more
than a little disturbing to me.  Va snpg, V nz frireyl cvffrq ng lbh!

So can we pretty please both swallow our feelings and work
constructively together?  Or should I merely treat this mail as a bug
report, go fix it, remove every single line of your code from the
driver and remove your name afterwards?  Both are valid options, you
choose.

Jörn

-- 
It does not matter how slowly you go, so long as you do not stop.
-- Confucius

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Mon, 7 March 2005 12:04:13 +0100, Jörn Engel wrote:
> > 
> > It appears that someone has "cleaned up" this driver and it now appears
> > as an option in the standard kernel tree for 2.6.11.
> > 
> > Unfortunately it doesn't appear to work. As soon as you mount a JFFS2
> > filesystem on it and try to write data, corruption sets in resulting is
> > a fairly sharp system crash.

Btw, it is interesting to note that noone has sent me bug reports
before.  Apparently, 2.6.<final> is getting *much* more testing than
mtd cvs, -mm, -bk* and -rc* together.

Jörn

-- 
I can say that I spend most of my time fixing bugs even if I have lots
of new features to implement in mind, but I give bugs more priority.
-- Andrea Arcangeli, 2000

Re: Re; MTD (2) Block Driver broken

[Josh Boyer]

On Mon, 2005-03-07 at 12:13 +0100, Jörn Engel wrote:
> 
> Btw, it is interesting to note that noone has sent me bug reports
> before.  Apparently, 2.6.<final> is getting *much* more testing than
> mtd cvs, -mm, -bk* and -rc* together.

Given that -mm, and -bk* are generally broken for at least half a dozen
things when they first come out, I'm not surprised.  Plus, the driver in
question probably isn't used all that often to begin with.

josh

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Mon, 7 March 2005 06:45:17 -0600, Josh Boyer wrote:
> 
> Given that -mm, and -bk* are generally broken for at least half a dozen
> things when they first come out, I'm not surprised.

Agreed.

> Plus, the driver in
> question probably isn't used all that often to begin with.

Not yet.  Gareth spoke of a few thousand downloads and the main
application - USB keys - is worth a few million users.  Provided the
driver actually works, that is.

Jörn

-- 
Time? What's that? Time is only worth what you do with it.
-- Theo de Raadt

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

Ahh, right, that makes sense.

I re-write a driver so it works properly.

You clean up all the code I've written to the extent that it no longer
works. (was it 22 different patches?)

Code gets submitted to kernel (obviously without testing) and ends up in
the live tree as broken code.

And it it's "my" fault for not checking and fixing all YOUR changes ?!

Right.

Why am I mailing you now?

Well, firstly I have in interest in the integrity of the kernel tree -
I've been working with it and on it since 1991. Secondly, you inserted
my name into your broken code - I do not appreciate this.

What would I like, I would suggest one of;

a.    Fix your changes, TEST, and re-submit to the kernel tree
or
b.    Remove the new driver from the kernel tree

Either way, I would appreciate it greatly if you would remove my name
from your work.

Incidentally, there's nothing wrong with my memory (yet) and I'm still
not interested in working on the mtd driver for exactly the same reasons
as previously mentioned.

Many thanks,
Gareth.


On Mon, 2005-03-07 at 12:04 +0100, Jörn Engel wrote:
> On Sat, 5 March 2005 17:26:09 +0000, Gareth Bult wrote:
> > 
> > A little while back I re-wrote blkmtd.c effectively to provide proper
> > "write" access to jffs2 filesystems mounted on standard Linux block
> > devices.
> > 
> > The driver seems to have worked well (for me) and after many thousands
> > of downloads of FlashLinux, I've not had any reports of JFFS2 problems
> > or corruption.
> > 
> > It appears that someone has "cleaned up" this driver and it now appears
> > as an option in the standard kernel tree for 2.6.11.
> > 
> > Unfortunately it doesn't appear to work. As soon as you mount a JFFS2
> > filesystem on it and try to write data, corruption sets in resulting is
> > a fairly sharp system crash.
> > 
> > Copying back the source from the original driver I submitted to this
> > list and recompiling the kernel (with a one line change) appears to
> > yield a working system again.
> > 
> > So, two questions;
> > a.	Did anyone actually test the code after cleaning it up?
> > b.	Were there any life changing improvements that I really
> > 	*need* to make to the original code?
> 
> Gareth, check your archives.
> 
> I sent the whole list of cleanup patches to you, you could have simply
> tested them in order and told me which one broke.  Instead you got
> pissed at dwmw2 since he didn't accept your broken mails and
> explicitly told me various times, you didn't care about the driver.
> 
> The fact that you don't remember me, you don't remember being on Cc:
> when I sent patches to Andrew Morton and now come running back is more
> than a little disturbing to me.  Va snpg, V nz frireyl cvffrq ng lbh!
> 
> So can we pretty please both swallow our feelings and work
> constructively together?  Or should I merely treat this mail as a bug
> report, go fix it, remove every single line of your code from the
> driver and remove your name afterwards?  Both are valid options, you
> choose.
> 
> Jörn
> 
--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Mon, 7 March 2005 13:06:56 +0000, Gareth Bult wrote:
> 
> Ahh, right, that makes sense.
> 
> I re-write a driver so it works properly.
> 
> You clean up all the code I've written to the extent that it no longer
> works. (was it 22 different patches?)
> 
> Code gets submitted to kernel (obviously without testing) and ends up in
> the live tree as broken code.
> 
> And it it's "my" fault for not checking and fixing all YOUR changes ?!
> 
> Right.

Sure.  You have your reasons to be pissed and I have no problem to
acknowledge them.  The current driver is broken and a fscking
embarrassment.  And I am to blame.  No doubt.

Yet - you have proven before and continue to prove than you are
impossible to work with:

> Well, firstly I have in interest in the integrity of the kernel tree -
> I've been working with it and on it since 1991. Secondly, you inserted
> my name into your broken code - I do not appreciate this.
> 
> What would I like, I would suggest one of;
> 
> a.    Fix your changes, TEST, and re-submit to the kernel tree
> or
> b.    Remove the new driver from the kernel tree
> 
> Either way, I would appreciate it greatly if you would remove my name
> from your work.

Deal.  Thank you for the original code.  It is a shame, but apparently
working without you is the only possible solution.

> Incidentally, there's nothing wrong with my memory (yet) and I'm still
> not interested in working on the mtd driver for exactly the same reasons
> as previously mentioned.

In that case, I shall consider it a an act of extreme politeness that
you directed your constructive criticism personally to me.  Very
efficient on so many levels.

Jörn, trying his best not to flame

-- 
"Error protection by error detection and correction."
-- from a university class

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

> Sure.  You have your reasons to be pissed and I have no problem to
> acknowledge them.  The current driver is broken and a fscking
> embarrassment.  And I am to blame.  No doubt.

You said;
>I sent the whole list of cleanup patches to you, you could have
>simply tested them in order and told me which one broke.

Hence my comments.

I'm not responsible for the testing your patches.

> The current driver is broken and a fscking embarrassment.

Indeed it is, yet I would have left well alone if it wasn't for the the
fact you put my name on it. I don't recall my name being on the code I
passed to the list .. (!) 

My code works, yet I now feel embarrassed.
- Do you understand my frustration ?

> Yet - you have proven before and continue to prove than you are
> impossible to work with:

Sure, I wasn't trying to work with you, but then it wasn't my intention
to have a go at you.

You might like (or not) to re-read my email of 5:25pm (5th Mar). I've
just re-read it. Maybe it's me, but it does seem like an entirely
reasonable email that doesn't mention or have a go at anybody ???

I trying to point out the fsck-up here where it could be fixed by
whoever submitted it, rather than on the kernel mailing lists where
people might look at the mtd double approval and ask wtf is going on.

>The fact that you don't remember me, you don't remember being on Cc:
>when I sent patches to Andrew Morton and now come running back is more
>than a little disturbing to me.

I left the list. I told you why I left the list. I told you I wasn't
going to do any more work on the driver. I'm not running back to you.

I'm not even asking you to fix any code, bottom line, all I'm really
asking is that you take my name off the code. (which I'm guessing you
put there in the first place!!)

I'm still using the code I initially submitted to the list.

> Jörn, trying his best not to flame

Mmm, flame me for reporting a fsck up ? Great. Really useful.

Flame everyone who reports a problem and slowly but surely, all your
problems will disappear!

--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

Net summary:
o You want your name removed.
o You will have your name removed.
o I'm left wondering why you submitted the code in the first place.

Jörn

-- 
Beware of bugs in the above code; I have only proved it correct, but
not tried it.
-- Donald Knuth

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Mon, 7 March 2005 19:12:50 +0000, Gareth Bult wrote:
> 
> Because I thought others might find it useful.

For sure.

> I didn't realise at the time that this list has 'other' priorities.

We are mortals.  We make mistakes.  And we have different priorities,
depending on which individual you happen to deal with.

> I will retire from the list once again.

Ack.

Jörn

-- 
Premature optimization is the root of all evil.
-- Donald Knuth

Re: Re; MTD (2) Block Driver broken

[David Woodhouse]

On Mon, 2005-03-07 at 20:00 +0100, Jörn Engel wrote:
> o You want your name removed.
> o You will have your name removed.
> o I'm left wondering why you submitted the code in the first place.

Er, because he'd done the work and figured it might be useful to others.
For which we are grateful, Gareth. Please don't be discouraged. You
aren't obliged to maintain the code for ever just because you posted it
once.

It is common for people to want to take credit for their work -- and
thankfully it's not so common for us to screw it up for them immediately
after they submit it either :)

However, if you really want your name removed from the driver, that's
perfectly OK with us. I would normally be inclined to follow the lead
set in lib/vsprintf.c and leave your name in it in such a way that it's
clear you're not to blame if it doesn't work (grep for 'portably') but
it's your call.

-- 
dwmw2

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

> On Mon, 7 March 2005 19:12:50 +0000, Gareth Bult wrote:
> 
> > I will retire from the list once again.

and be pruned from the Cc: list.

Bug is fixed in cvs.

Jörn

-- 
He who knows others is wise.
He who knows himself is enlightened.
-- Lao Tsu

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

On Mon, 2005-03-07 at 20:21 +0000, David Woodhouse wrote:
> On Mon, 2005-03-07 at 20:00 +0100, Jörn Engel wrote:
> > o You want your name removed.
> > o You will have your name removed.
> > o I'm left wondering why you submitted the code in the first place.
> 
> Er, because he'd done the work and figured it might be useful to others.
> For which we are grateful, Gareth. Please don't be discouraged. You
> aren't obliged to maintain the code for ever just because you posted it
> once.

Oh, I'm not discouraged. What I have works for now.

I did a fair bit of work while writing the driver and had every
intention of tidying it up, commenting it etc etc.

Then I got an email from the list telling me that what I'd posted was
unacceptable and had been rejected. Furthermore it was telling me what I
could and couldn't post and how I should format my emails and
effectively, that I should read and follow a set of posting instructions
specific to the linux-mtd list.

All things considered. I was offended.

My choice was to leave the list and spend my time elsewhere.

I'm an unpaid / full time Linux "developer", I have no problem
(generally) maintaining the code I write.

> It is common for people to want to take credit for their work -- and
> thankfully it's not so common for us to screw it up for them immediately
> after they submit it either :)

Sure, generally I have no problem taking credit for my work. However I
made it fairly clear (I thought) to Jorn at the time that I wanted
nothing more to do with it.

> However, if you really want your name removed from the driver, that's
> perfectly OK with us. I would normally be inclined to follow the lead
> set in lib/vsprintf.c and leave your name in it in such a way that it's
> clear you're not to blame if it doesn't work (grep for 'portably') but
> it's your call.

Well, my inclination is to go with "please remove it".

My priority is working code, working documentation and working Linux.

To reject posts from a developer because he's formatted an email in such
a way that it doesn't conform to someone's idea of a perfect posting,
just beggars belief.

Gareth.

--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Mon, 7 March 2005 20:57:43 +0000, Gareth Bult wrote:
> 
> Well, my inclination is to go with "please remove it".
> 
> My priority is working code, working documentation and working Linux.

Fwiw, I've found and fixed the bug.  Patch, along with the removal of
your name, is on the way to Linus.

If you insist, I can remove the driver completely and rewrite it from
scratch, with none of your code remaining.  Tell me to do so and I
will.

BTW: Thank you for the bug report.  I've noticed this earlier but
suspected jffs2 or my cowlink work to be at fault.  Despite our
ability to get mad at each other, both your initial code and the bug
report were very helpful (and worth any grief on my side).

Jörn

-- 
If you're willing to restrict the flexibility of your approach,
you can almost always do something better.
-- John Carmack

Re: Re; MTD (2) Block Driver broken

[David Woodhouse]

On Mon, 2005-03-07 at 20:57 +0000, Gareth Bult wrote:
> On Mon, 2005-03-07 at 20:21 +0000, David Woodhouse wrote:
> > On Mon, 2005-03-07 at 20:00 +0100, Jörn Engel wrote:
> > > o You want your name removed.
> > > o You will have your name removed.
> > > o I'm left wondering why you submitted the code in the first place.
> > 
> > Er, because he'd done the work and figured it might be useful to others.
> > For which we are grateful, Gareth. Please don't be discouraged. You
> > aren't obliged to maintain the code for ever just because you posted it
> > once.
> 
> Oh, I'm not discouraged. What I have works for now.
> 
> I did a fair bit of work while writing the driver and had every
> intention of tidying it up, commenting it etc etc.
> 
> Then I got an email from the list telling me that what I'd posted was
> unacceptable and had been rejected. Furthermore it was telling me what I
> could and couldn't post and how I should format my emails and
> effectively, that I should read and follow a set of posting instructions
> specific to the linux-mtd list.
> 
> All things considered. I was offended.

I apologise for that. I don't mean to offend anyone -- and any messages
'from the list' come from me, originally. 

If there was criticism of the code, it was meant constructively -- we
can be very picky at times, but it's certainly not meant to offend;
we're just perfectionists.

I think we have to agree to disagree on the rules for posting. The page
to which you were referred (http://david.woodhou.se/email.html) wasn't
specific to this list -- it gives guidelines for polite use of email
under _all_ circumstances, with only the last point being specific to
Linux kernel related lists. By being strict about them, we _do_ keep the
signal:noise ratio down, and I think it's worth it overall.

To see the advantages, you only have to look at the noise on certain
other lists where people do get away with posting whatever crap they
like. It's not as if you were asked to do anything particularly onerous,
surely? But again, if you were offended by being asked to conform to
what we consider to be common netiquette, then I apologise. I can only
point out that others would be offended if we let it happen -- if not
directly by the posts in question, then by the fact that the developers
who can be most useful in answering questions and fixing bugs start to
pay less attention because of the noise.

> > It is common for people to want to take credit for their work -- and
> > thankfully it's not so common for us to screw it up for them immediately
> > after they submit it either :)
> 
> Sure, generally I have no problem taking credit for my work. However I
> made it fairly clear (I thought) to Jorn at the time that I wanted
> nothing more to do with it.

That's OK -- but still I suspect it was just such an alien attitude that
we didn't quite believe you :)

Jörn screwed up your code, but seems to have fixed it now. He's removed
your name at your request, but if you want it put back now that the code
appears to be working, just say the word. Personally, I'm much happier
giving credit where it's due.

-- 
dwmw2

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

> If you insist, I can remove the driver completely and rewrite it from
> scratch, with none of your code remaining.  Tell me to do so and I
> will.

Insist? I didn't either mention or suggest it (!)

Given that very few people use it, spending a week rewriting and testing
would seem a little inefficient (?)

fyi; there are at least 10,000 copies in circulation and no outstanding
bug or problem reports (that I know of) .. that said, most users are
probably unaware they're using it.

> BTW: Thank you for the bug report.  I've noticed this earlier but
> suspected jffs2 or my cowlink work to be at fault.  Despite our
> ability to get mad at each other, both your initial code and the bug
> report were very helpful (and worth any grief on my side).

No problem.

One issue you might like to be aware of ..

To use blkmtd as a module, you do;

modprobe blkmtd device=/dev/sda1 (for example)

For block2mtd, you do;

modprobe block2mtd block2mtd=/dev/sda1 (I think..)

Anyone who's coded up an initrd for example would have to recode for
both a different driver name and a different parameter name. (which
incidentally isn't immediately obvious)

It might (?) make sense to have the new driver replace the old (as an
option, rather than have a new name) and use the same parameter name ..
just as a general principle so people don't have to go and recode to
match the driver upgrade (?)

.. the "write" speed on the new driver is > 100x faster than the old ..
I was unable to think of a practical use for the old driver .. (!)

Gareth.

--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Mon, 7 March 2005 21:46:20 +0000, Gareth Bult wrote:
> 
> > If you insist, I can remove the driver completely and rewrite it from
> > scratch, with none of your code remaining.  Tell me to do so and I
> > will.
> 
> Insist? I didn't either mention or suggest it (!)

Good.

> One issue you might like to be aware of ..
> 
> To use blkmtd as a module, you do;
> 
> modprobe blkmtd device=/dev/sda1 (for example)
> 
> For block2mtd, you do;
> 
> modprobe block2mtd block2mtd=/dev/sda1 (I think..)

Correct.

> Anyone who's coded up an initrd for example would have to recode for
> both a different driver name and a different parameter name. (which
> incidentally isn't immediately obvious)
> 
> It might (?) make sense to have the new driver replace the old (as an
> option, rather than have a new name) and use the same parameter name ..
> just as a general principle so people don't have to go and recode to
> match the driver upgrade (?)
> 
> .. the "write" speed on the new driver is > 100x faster than the old ..
> I was unable to think of a practical use for the old driver .. (!)

The old driver also got the interface wrong.  In your example, with
just a single device to use, it doesn't matter.  But once you deal
with several devices, the old variant really sucks.  That is why I
changed it (and why I renamed the driver).

That said, I would like to agree with you.  If the new driver could
mimikri the old one, that would be a clear advantage.  The question
is, how can both advantages - better interface and compatibility - be
combined.  At the moment, I don't have an answer, but I'll keep
thinking about it.

Jörn

-- 
The grand essentials of happiness are: something to do, something to
love, and something to hope for.
-- Allan K. Chalmers

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

> To see the advantages, you only have to look at the noise on certain
> other lists where people do get away with posting whatever crap they
> like. It's not as if you were asked to do anything particularly onerous,
> surely? But again, if you were offended by being asked to conform to
> what we consider to be common netiquette, then I apologise. I can only
> point out that others would be offended if we let it happen -- if not
> directly by the posts in question, then by the fact that the developers
> who can be most useful in answering questions and fixing bugs start to
> pay less attention because of the noise.

Mmm, sure. I've only been an Internet email user / Unix developer for 18
years, what would I know about netiquette .. 

> That's OK -- but still I suspect it was just such an alien attitude that
> we didn't quite believe you :)

If I type an email, I do it for a reason.

If an automaton mails me back telling me that it's not proper
netiquette, my time is being wasted. Spending time is fine, wasting it
is unacceptable and offencive.

I get a LOT of junk mail. I have a "Junk" button in my mailer. I don't
waste people's time telling them how to format emails, it's not
reasonable. I rely on common sense. If I get an unacceptable email, I
junk it.

> Jörn screwed up your code, but seems to have fixed it now. He's removed
> your name at your request, but if you want it put back now that the code
> appears to be working, just say the word. Personally, I'm much happier
> giving credit where it's due.

When I get time I'll test it. If it works, I may even use it. I'm not
however desperate to have my name on it. Everyone screws up code - it's
not a problem .. submitting changed code however without testing it is a
little .. overconfident (?)

I don't work on code for credit, I work on it because I enjoy it.

Gareth.

--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[David Woodhouse]

On Mon, 2005-03-07 at 22:10 +0000, Gareth Bult wrote:
> I get a LOT of junk mail. I have a "Junk" button in my mailer. I don't
> waste people's time telling them how to format emails, it's not
> reasonable. I rely on common sense. If I get an unacceptable email, I
> junk it.

I, on the other hand, get very little junk mail -- mostly because of the
way I run my mail servers. I happen to like it that way.

Out of interest, what in particular was it that got your mail rejected?
I don't recall.

-- 
dwmw2

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

> The old driver also got the interface wrong.  In your example, with
> just a single device to use, it doesn't matter.  But once you deal
> with several devices, the old variant really sucks.  That is why I
> changed it (and why I renamed the driver).

If you were to change the parameter name back to "device", it would be
compatible with the old driver. (pretty much) Hence you'd not need to
change the name (?)

.. initially all my system worked with the pre-existing driver. The new
code was a drop-in replacement and I didn't need to make changes to my
system.

Can you not make the parameter handling "sane" and still keep the syntax
such that it's backwards compatible ?

--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

I don't recall the specifics .. it was part of a conversation .. nothing
startling different from the emails I've been posting here today.


On Mon, 2005-03-07 at 22:20 +0000, David Woodhouse wrote:
> On Mon, 2005-03-07 at 22:10 +0000, Gareth Bult wrote:
> > I get a LOT of junk mail. I have a "Junk" button in my mailer. I don't
> > waste people's time telling them how to format emails, it's not
> > reasonable. I rely on common sense. If I get an unacceptable email, I
> > junk it.
> 
> I, on the other hand, get very little junk mail -- mostly because of the
> way I run my mail servers. I happen to like it that way.
> 
> Out of interest, what in particular was it that got your mail rejected?
> I don't recall.
> 
--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[David Woodhouse]

On Mon, 2005-03-07 at 22:26 +0000, Gareth Bult wrote:
> I don't recall the specifics .. it was part of a conversation ..
> nothing startling different from the emails I've been posting here
> today

Which were all accepted on the list apart from the one which was HTML.

Of the HTML mail we reject, most of it is spam -- I'm perfectly content
with rejecting HTML mail to the list, and it's not a particularly
uncommon occurrence. If you have content which can only sanely be
represented in HTML form, then you can always send it as an attachment.

-- 
dwmw2

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Mon, 7 March 2005 22:10:17 +0000, Gareth Bult wrote:
> 
> When I get time I'll test it. If it works, I may even use it. I'm not
> however desperate to have my name on it. Everyone screws up code - it's
> not a problem .. submitting changed code however without testing it is a
> little .. overconfident (?)

When I wrote it, I lacked the hardware to test it.  Then I tested it
with mtdblock, and it worked.  Then I tested it with jffs2 and that
didn't work.  I suspected jffs2, spent a week (about 50-60 hours) on
that one, wrote 34 seperate patches for jffs2 and still couldn't find
it.

Bug was actually in block2mtd, mtdblock just didn't trigger it.  Darn!

Only after testing did I submit the code to cvs, so I was not
overconfident.  Stupid is more to the point.

> I don't work on code for credit, I work on it because I enjoy it.

Well, the credit is free advertisement.  On the next job interview, a
"just grep the kernel sources" comment explains my qualifications and
may keep me from unemployment here or raise my salary there.

Credit is never the reason to write code in the first place.  My
block2mtd patches were written for fun and because I would like to
replace my hard driver with a USB key.  But the free extra doesn't
hurt, at least until I win the lottery (and I never play).

Jörn

-- 
Don't patch bad code, rewrite it.
-- Kernigham and Pike, according to Rusty

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Mon, 7 March 2005 22:20:38 +0000, Gareth Bult wrote:
> 
> > The old driver also got the interface wrong.  In your example, with
> > just a single device to use, it doesn't matter.  But once you deal
> > with several devices, the old variant really sucks.  That is why I
> > changed it (and why I renamed the driver).
> 
> If you were to change the parameter name back to "device", it would be
> compatible with the old driver. (pretty much) Hence you'd not need to
> change the name (?)

That would work for 99% of all users - those that only use a single
device and no other parameters like erase size.  For the remaining 1%,
I don't want to break anything, so the different name should remain.

What I could do is introduce a new config option, like
CONFIG_RENAME_BLOCK2MTD_TO_BLKMTD.  With the option set, the driver
also accepts the "device" parameter and the kernel module is renamed.

Would this work for you?

Jörn

-- 
Eighty percent of success is showing up.
-- Woody Allen

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

> Which were all accepted on the list apart from the one which was HTML.

Mmm, as I say, I don't recall the detail.

Like many mailers, mine prefers to send HTML by default. Occasionally
one will forget one is posting to a list where people have some sort of
religious objection to html and forget to uncheck the appropriate
buttons to format as pure text.

> Of the HTML mail we reject, most of it is spam -- I'm perfectly content
> with rejecting HTML mail to the list, and it's not a particularly
> uncommon occurrence. If you have content which can only sanely be
> represented in HTML form, then you can always send it as an attachment.

Well, it's your list.

--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

> Only after testing did I submit the code to cvs, so I was not
> overconfident.  Stupid is more to the point.

Stupid is a term that's often misused and misunderstood , I try to avoid
it. Just because someone makes a mistake doesn't make them stupid ..
possible prone to making errors, possibly overconfident, possibly
lacking in experience .. possibly just unlucky.

> Well, the credit is free advertisement.  On the next job interview, a
> "just grep the kernel sources" comment explains my qualifications and
> may keep me from unemployment here or raise my salary there.

Mmm, not something I tend to worry about.

> Credit is never the reason to write code in the first place.  My
> block2mtd patches were written for fun and because I would like to
> replace my hard driver with a USB key.  But the free extra doesn't
> hurt, at least until I win the lottery (and I never play).

I guess it depends on the sort of work you're after. If you're talking
short contracts - sure.

If you're talking "real" work, it doesn't take more than a couple of
days to see what someone can "really" do. Having one's name on a Linux
device driver is hardly indicative on one's ability.

(!)

Gareth.

--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

Mmm, 

would make a lot of sense.

At the end of the day and at the moment, I'm probably the only one using
the new driver so it's really not an issue .. I was really just talking
about "good form" when releasing updates ..

Gareth.

On Mon, 2005-03-07 at 23:42 +0100, Jörn Engel wrote:
> On Mon, 7 March 2005 22:20:38 +0000, Gareth Bult wrote:
> > 
> > > The old driver also got the interface wrong.  In your example, with
> > > just a single device to use, it doesn't matter.  But once you deal
> > > with several devices, the old variant really sucks.  That is why I
> > > changed it (and why I renamed the driver).
> > 
> > If you were to change the parameter name back to "device", it would be
> > compatible with the old driver. (pretty much) Hence you'd not need to
> > change the name (?)
> 
> That would work for 99% of all users - those that only use a single
> device and no other parameters like erase size.  For the remaining 1%,
> I don't want to break anything, so the different name should remain.
> 
> What I could do is introduce a new config option, like
> CONFIG_RENAME_BLOCK2MTD_TO_BLKMTD.  With the option set, the driver
> also accepts the "device" parameter and the kernel module is renamed.
> 
> Would this work for you?
> 
> Jörn
> 
--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw

Re: Re; MTD (2) Block Driver broken

[Jörn Engel]

On Tue, 8 March 2005 10:33:31 +0000, Gareth Bult wrote:
> 
> would make a lot of sense.
> 
> At the end of the day and at the moment, I'm probably the only one using
> the new driver so it's really not an issue .. I was really just talking
> about "good form" when releasing updates ..

I largely agree with the "good form" aspect.  Trouble with mtd is that
some of the interfaces are really broken.  And for user-visible
interfaces, there is no sane way to change them - ever.  So I
implemented a new interface, trying to make that one as good as
possible.

So I'll add the rename option and code.

Jörn

-- 
When I am working on a problem I never think about beauty.  I think
only how to solve the problem.  But when I have finished, if the
solution is not beautiful, I know it is wrong.
-- R. Buckminster Fuller

Re: Re; MTD (2) Block Driver broken

[David Woodhouse]

On Tue, 2005-03-08 at 10:24 +0000, Gareth Bult wrote:
> Like many mailers, mine prefers to send HTML by default.

I think you'll find it's configurable. 

>  Occasionally one will forget one is posting to a list where people
> have some sort of religious objection to html and forget to uncheck
> the appropriate buttons to format as pure text.

I don't have a religious objection to HTML. Like I said, there's little
cause to prefer HTML over plain text for list traffic (and indeed for
most email), and most of the HTML we reject is spam -- being HTML is a
fairly good indicator of spam.

I spend a _lot_ of time reading email each day, and almost none of it is
HTML -- it's plain text, in a typeface I have carefully selected to be
easily readable. HTML mail renders differently and is thus harder to
read. Not a _lot_ harder, mind you, but it's unfamiliar enough that it
causes me to stop and look rather than just taking in what it says at a
glance. Mostly I don't actually stop and look; mostly I just delete it. 

I _could_ spend some of my time on decoding gratuitously HTML-encoded
mail, or I could spend all of my available time on the people who can be
bothered to be polite. My time is limited enough at the moment anyway
that I really don't want to waste it.

-- 
dwmw2

Re: Re; MTD (2) Block Driver broken

[Gareth Bult]

> I think you'll find it's configurable. 

Bit like the list's spam detection then ?

> I don't have a religious objection to HTML. Like I said, there's little
> cause to prefer HTML over plain text for list traffic (and indeed for
> most email), and most of the HTML we reject is spam -- being HTML is a
> fairly good indicator of spam.

Rubbish. All of my mail is by default HTML and has been for > 10 years,
does that mean that the fact the email is from me is a good indicator
that it's spam ?

It is true that a lot of spammers use HTML, however a many don't. It's
AN indicator, it's also a very unreliable indicator. From where I'm
sitting at the moment, the "big" guys (from whom you can un-subscribe)
are using HTML, everyone else is trying every trick in the book to avoid
detection, one popular trick is not to use HTML.

After checking, 23 emails got through my filters so far this week, 4 of
them were HTML. One was Dell, one was Google .. 

> I spend a _lot_ of time reading email each day, and almost none of it is
> HTML -- it's plain text, in a typeface I have carefully selected to be
> easily readable. HTML mail renders differently and is thus harder to
> read. 

Yeah, right, in 1995 maybe ..

> Not a _lot_ harder, mind you, but it's unfamiliar enough that it
> causes me to stop and look rather than just taking in what it says at a
> glance. Mostly I don't actually stop and look; mostly I just delete it. 
> 
> I _could_ spend some of my time on decoding gratuitously HTML-encoded
> mail, or I could spend all of my available time on the people who can be
> bothered to be polite. My time is limited enough at the moment anyway
> that I really don't want to waste it.

Mmm, fortunately my talents extend to reading text in different
typefaces without undue delay. Although I don't like wasting time I am
prepared to spend an additional 1% of my reading time on less than
perfectly formed email messages.

Basically what you're saying is that it's your list and you'll set the
rules as you like because it's your time that's being spent reading the
emails.

This is fair enough and it's why I left the list.


--
Linux.co.uk - promoting open source software in the UK
"If you have an apple and I have an apple and we exchange apples then
you and I will still each have one apple. But if you have an idea and I
have an idea and we exchange these ideas, then each of us will have two
ideas." - George Bernard Shaw