Re: [9fans] quotefmtinstall

Re: [9fans] quotefmtinstall

[Charles Forsyth]

i'd check that you've got text and data loaded at the locations
they expect.  it can often get that far if not (thanks to pc-relative
references).  fmtinstall also invokes tas for the first time.

Re: [9fans] quotefmtinstall

[Ronald G. Minnich]

On Fri, 11 Mar 2005, Charles Forsyth wrote:

> i'd check that you've got text and data loaded at the locations
> they expect.  it can often get that far if not (thanks to pc-relative
> references).  fmtinstall also invokes tas for the first time.

I'm pretty sure text is good. Data I am not so sure about.

ron

Re: [9fans] quotefmtinstall

[Charles Forsyth]

>I'm pretty sure text is good. Data I am not so sure about.

if it uses u-boot, it's bound to get in the way at some point.
what is it about bootstrap authors?
more like elephant ears than boot-straps,
and just try tugging one of those!

usually i try to load a headerless binary image so that
i can do hexdumps of chunks of memory to ensure that
matches what the nm -n suggests.

Re: [9fans] quotefmtinstall

[Ronald G. Minnich]

On Fri, 11 Mar 2005, Charles Forsyth wrote:

> if it uses u-boot, it's bound to get in the way at some point.

here's my elf so far:

ELF Header:
  Magic:   7f 45 4c 46 01 01 01 00 00 00 00 00 00 00 00 00
  Class:                             ELF32
  Data:                              2's complement, little endian
  Version:                           1 (current)
  OS/ABI:                            UNIX - System V
  ABI Version:                       0
  Type:                              EXEC (Executable file)
  Machine:                           ARM
  Version:                           0x1
  Entry point address:               0x218010
  Start of program headers:          52 (bytes into file)
  Start of section headers:          0 (bytes into file)
  Flags:                             0x0
  Size of this header:               52 (bytes)
  Size of program headers:           32 (bytes)
  Number of program headers:         3
  Size of section headers:           0 (bytes)
  Number of section headers:         0
  Section header string table index: 0

There are no sections in this file.

Program Headers:
  Type           Offset   VirtAddr   PhysAddr   FileSiz MemSiz  Flg Align
  LOAD           0x0000a0 0x00218010 0x00218010 0x68bb8 0x68bb8 R E 0x4
  LOAD           0x068c58 0x00280bc8 0x00280bc8 0x30910 0x39460 RW  0x4
  NULL           0x099568 0x00000000 0x00000000 0x1dc7d 0x0ce0e R   0x4


Here's the requisite symbols

c0280bc8 D edfruntime
c0280bc8 D bdata
c0280bc8 T etext

So I think that second header in there (280bc8) is correct. I have not
verified, however, that when you unpack the elf the data matches. I guess
I should do that next.

ron

Re: [9fans] quotefmtinstall

[Charles Forsyth]

another thing to watch:
if you're linked against c000... base,
and the mmu isn't set to c000... virtual
on entry to your l.s, but is still running physical
addresses, you'll need to set R12 to the
physical address for setSB until you load the mmu.

Re: [9fans] quotefmtinstall

[Ronald G. Minnich]

I think that's ok, it's PIC until I load the MMU.

I'm actually getting into calling functions , returning, etc. It's having
some trouble in lock, it looks like, which I don't understand.

I think it may be splhi, actually; probably some wrong bit that doesn't
match the SA1100 (I assumed CPSR would stay the same but I guess not).


ron

Re: [9fans] quotefmtinstall

[vdharani]

> I think that's ok, it's PIC until I load the MMU.
>
> I'm actually getting into calling functions , returning, etc. It's having
> some trouble in lock, it looks like, which I don't understand.
>
> I think it may be splhi, actually; probably some wrong bit that doesn't
> match the SA1100 (I assumed CPSR would stay the same but I guess not).

what would happen if you simply mask splhi (say you replace with a
return)? that way, wont it be possible to see atleast whether it
progresses?

i am not an expert in this area but i remember i faced similar problem
when i tried with an x86 box. in my case, the board was freezing during
bootup and i found that it happened when the ethernet interrupt tried to
do locking or so which led to splhi it caused the problem. and it happened
only when i enabled the third ethernet port (all being Reaktek 8139C). it
was fine as long as i used just the first two ports. i tried this
experiment many years before and i vaguely remember i tried to mask splhi.

thanks
dharani


in my case, i fugured that it was working fine as long as i use upto two
ethernet ports.

Re: [9fans] quotefmtinstall

[Ronald G. Minnich]

oh yeah, loading is a bit tricky on this thing; it runs RedBoot.

If I had linuxbios, life would be simple as I could make it do whatever I
want.

ron

[9fans] quotefmtinstall

[Charles Forsyth]

In these days of 3 Gbyte memory requirements for browsers, and 1 Gbyte
gnome-panels, should we still insist on quotefmtinstall to ensure %q?

Here's the before & after:
h% size 8.out
19196t + 2388d + 788b = 22372 8.out

h% size 8.out
21915t + 2396d + 788b = 25099 8.out

Re: [9fans] quotefmtinstall

[lucio]

> In these days of 3 Gbyte memory requirements for browsers, and 1 Gbyte
> gnome-panels, should we still insist on quotefmtinstall to ensure %q?

I'm with you, but it's also a slippery slope...

I do wish we could spend effort in the opposite direction, myself.
Maybe it could be an optional patch, for those of us who get bitten
and have more important things in life than remembering that 8c is not
8g :-)

++L

Re: [9fans] quotefmtinstall

[erik quanstrom]

On Sat Dec  8 11:27:32 EST 2012, lucio@proxima.alt.za wrote:
> > In these days of 3 Gbyte memory requirements for browsers, and 1 Gbyte
> > gnome-panels, should we still insist on quotefmtinstall to ensure %q?
>
> I'm with you, but it's also a slippery slope...
>
> I do wish we could spend effort in the opposite direction, myself.
> Maybe it could be an optional patch, for those of us who get bitten
> and have more important things in life than remembering that 8c is not
> 8g :-)

some points.

0.  the essence of charles' point is that persistent memory is cheep, and little
fiddly bits are expensive, from a developer's point of view.  and by
cheep, we're talking silly cheep.  1kb of disk costs 100 nanodollars.  it's the
order of magnitude that's important here, a small integer factor won't
make disks expensive.

1.  it can't be optional.  this is the sort of thing that got
latter-day unix in so much trouble.  if you want your program to
work everywhere (as most people do) you have to program for the
least common denominator, so every program will need
quotefmtinstall().  so there will be a net savings of zero.  and
since the problem will now appear on some systems, it will probablly
be harder to remember that you've forgotten to quotefmtinstall.

2.  if you want to save overall storage, the print library should ditch
standard support for the rune*print() functions.  only 4 programs use
them.

- erik

Re: [9fans] quotefmtinstall

[lucio]

> 0.  the essence of charles' point is that persistent memory is cheep, and little
> fiddly bits are expensive, from a developer's point of view.  and by
> cheep, we're talking silly cheep.  1kb of disk costs 100 nanodollars.  it's the
> order of magnitude that's important here, a small integer factor won't
> make disks expensive.
>
The digital divide, what can I say?

Sadly, I can't buy nanodollars' worth of technology, even though it is
being discarded daily in the western world by the megadollar.  Even
the RaspberryPI is beyond reach (we can approach western sponsors for
funding, but that creates a form of slavery that africans are
understandably reluctant to depend on - our leaders don't mind, but
some of us have a social conscience).

I guess I'm asking you and Charles and others to keep that in mind
when you contribute to the conspiracy to make the digital divide
bigger, no matter how small your contribution.

> 1.  it can't be optional.  this is the sort of thing that got
> latter-day unix in so much trouble.  if you want your program to
> work everywhere (as most people do) you have to program for the
> least common denominator, so every program will need
> quotefmtinstall().  so there will be a net savings of zero.  and
> since the problem will now appear on some systems, it will probablly
> be harder to remember that you've forgotten to quotefmtinstall.
>
> 2.  if you want to save overall storage, the print library should ditch
> standard support for the rune*print() functions.  only 4 programs use
> them.

I understand the technological issues more or less well enough.  But
where I stand, I can't ignore the social implications of consuming
resources as soon as they become available.  I can think of a penalty
that could be paid by inefficient consumers, but in a world where
survival of the fittest is the guiding civilising principle, I am
geographically an endangered species.

++L

PS: Your "if you want your program to work everywhere (as most people
do)" is spurious.  Why are we programming for anything but MS Windows,
if that is the case?

Re: [9fans] quotefmtinstall

[erik quanstrom]

> Sadly, I can't buy nanodollars' worth of technology, even though it is
> being discarded daily in the western world by the megadollar.  Even
> the RaspberryPI is beyond reach

if you don't have the money for a raspberry pi, then you can't afford
anything that will boot plan 9.  so arguing that adding 1.7% (maybe)
to the size of /bin is an argument for the moot court.  and in that
moot court, one could argue that stripping the executables would save
much more than 1.7%.

> (we can approach western sponsors for
> funding, but that creates a form of slavery that africans are
> understandably reluctant to depend on - our leaders don't mind, but
> some of us have a social conscience).

i don't know what slavery has got to do with anything.
and i find the comparison to such a horrible crime distasteful.

> I understand the technological issues more or less well enough.  But
> where I stand, I can't ignore the social implications of consuming
> resources as soon as they become available.  I can think of a penalty

since this isn't a social justice list, i'll just address the techncal
assertion.

it is untrue that consuming an extra 3121 bytes
of storage per executable is consuming resources at the leading edge.
sd storage is sold in big chunks.  the smallest sd card i can find is
256mb.  a 1gb sd card is roughly the same price.  (and please no quibbling.
we've already established that the pi itself is too expensive)
you will probablly need a 1gb card, as the image alone is nearly 500mb.

- erik

Re: [9fans] quotefmtinstall

[hiro]

hu? you can get x86s thrown after you for free. including the crt
monitor and dirty model m keyboard. even a three button mouse might be
there if you search the trash properly.
no comparison with pi plus power adapter plus usb mouse and keyboard
plus case plus hdmi plus tv
also thinkpads are dead cheap and you dont even have to look for a 3
button mouse

On 12/10/12, erik quanstrom <quanstro@quanstro.net> wrote:
>> Sadly, I can't buy nanodollars' worth of technology, even though it is
>> being discarded daily in the western world by the megadollar.  Even
>> the RaspberryPI is beyond reach
>
> if you don't have the money for a raspberry pi, then you can't afford
> anything that will boot plan 9.  so arguing that adding 1.7% (maybe)
> to the size of /bin is an argument for the moot court.  and in that
> moot court, one could argue that stripping the executables would save
> much more than 1.7%.
>
>> (we can approach western sponsors for
>> funding, but that creates a form of slavery that africans are
>> understandably reluctant to depend on - our leaders don't mind, but
>> some of us have a social conscience).
>
> i don't know what slavery has got to do with anything.
> and i find the comparison to such a horrible crime distasteful.
>
>> I understand the technological issues more or less well enough.  But
>> where I stand, I can't ignore the social implications of consuming
>> resources as soon as they become available.  I can think of a penalty
>
> since this isn't a social justice list, i'll just address the techncal
> assertion.
>
> it is untrue that consuming an extra 3121 bytes
> of storage per executable is consuming resources at the leading edge.
> sd storage is sold in big chunks.  the smallest sd card i can find is
> 256mb.  a 1gb sd card is roughly the same price.  (and please no quibbling.
> we've already established that the pi itself is too expensive)
> you will probablly need a 1gb card, as the image alone is nearly 500mb.
>
> - erik
>
>

[9fans] quotefmtinstall

[Ronald G. Minnich]

now why would it die in quotefmtinstall ... Man, it even zero'ed bss,
which means function calls work, etc. Anyway, that's where it is.

strongarm folks, this card is $139, low power, etc. If there are others
interested in helping to bring this card up, let me know. It's quite a
nice little toy.

ron

Re: [9fans] quotefmtinstall

[Wes Kussmaul]

For those of us who missed it, *what* card?

Ronald G. Minnich wrote:

>now why would it die in quotefmtinstall ... Man, it even zero'ed bss,
>which means function calls work, etc. Anyway, that's where it is.
>
>strongarm folks, this card is $139, low power, etc. If there are others
>interested in helping to bring this card up, let me know. It's quite a
>nice little toy.
>
>ron
>
>
>

Re: [9fans] quotefmtinstall

[Derek Fawcus]

On Fri, Mar 11, 2005 at 11:01:47AM -0500, Wes Kussmaul wrote:
> For those of us who missed it, *what* card?

Maybe Ron is referring to the ts7200
  http://www.embeddedx86.com/epc/ts7200-spec-h.html

DF

Re: [9fans] quotefmtinstall

[David Leimbach]

On Fri, 11 Mar 2005 16:24:29 +0000, Derek Fawcus <dfawcus@cisco.com> wrote:
> On Fri, Mar 11, 2005 at 11:01:47AM -0500, Wes Kussmaul wrote:
> > For those of us who missed it, *what* card?
>
> Maybe Ron is referring to the ts7200
>   http://www.embeddedx86.com/epc/ts7200-spec-h.html
>
> DF
>
That's a pretty awesome board.  Sure it's only 140 bucks? :)

I'll take 2  :)

Dave

Re: [9fans] quotefmtinstall

[Russ Cox]

See also http://gumstix.com.

Russ

Re: [9fans] quotefmtinstall

[Ronald G. Minnich]

On Fri, 11 Mar 2005, Wes Kussmaul wrote:

> For those of us who missed it, *what* card?

sorry.

The technologic ts7200, from www.embeddedx86.com

ron