[9fans] Foreign keyboard layout

[9fans] Foreign keyboard layout

[Jesus A. Mora]

Surely I'm missing the point, but is there any simple and straightforward
way to configure an alternative keyboard layout (spanish keyboard, in my
case)? I've seen no clear reference in the man pages, and found only a
hardcoded mapping in /sys/src/9/cputype/kbd.c

Hints? RTFM? man anything(x)?

TIA

JAMM

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> Hints? RTFM? man anything(x)?

iirc, there is a driver that has a mapping file which you
can read/write tuples of the form:

    shift scancode rune

trouble is that on azerty it breaks as soon as you get near
the scancode for the zero key;  high bit(s) used in the driver
become overloaded.  it's been a a while since i looked at this,
but these bits are used to indicate the shiftedness of the
scancode and the scancode.

the other issue is there are 3 shifts on azerty:

    shift
    ctrl
    alt-gr[aphic]

because of my peculiar h/w config i didn't persue it.

nobody, but nobody, gets internationali[sz]ation right.

Re: [9fans] Foreign keyboard layout

[Jesus A. Mora]

On Tue, 18 Sep 2001, Boyd Roberts wrote:

> iirc, there is a driver that has a mapping file which you
> can read/write tuples of the form:
>
>     shift scancode rune
>
Thanks. Please, do you remember what driver and/or file(s) are you talking
about?
I've found hardcoded keyboard mapping in both:
	/sys/src/9/$cputype/kbd.c       (machine-dependent kernel code)
	/sys/src/libcontrol/keyboard.c
Maybe there are other files like these.

TIA,

JAMM

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

i believe i got said code from charles forsyth.

i don't have a copy due to gravity accidents and
the inability to back my vaio installed plan 9
system to cd rom [no fault of the labs guys].

given i couldn't really type and couldn't back it up,
i gave up.  this is because of my h/w config and the
fact that ~6 people _cannot_ support every weird device
on the planet.  �sloth have their armies of typing
droids and they AND their partners can't even do it.

adaptec cd burner s/w?  it doesn't work and it was
bundled with my kodac [teac] cd burner.  i/o underuns
on a 700Mhz pentium with 64Mb of ram.  gee, makes me
wanna pick up an 11/780 on ebay...

Re: [9fans] Foreign keyboard layout

[Quinn Dunkan]

> On Tue, 18 Sep 2001, Boyd Roberts wrote:
>
> > iirc, there is a driver that has a mapping file which you
> > can read/write tuples of the form:
> >
> >     shift scancode rune
> >
> Thanks. Please, do you remember what driver and/or file(s) are you talking
> about?
> I've found hardcoded keyboard mapping in both:
> 	/sys/src/9/$cputype/kbd.c       (machine-dependent kernel code)
> 	/sys/src/libcontrol/keyboard.c
> Maybe there are other files like these.

There's a program that provides its own /dev/cons that translates from the old
/dev/cons and provides /dev/kbd to configure the translation (you write
old_unicode new_unicode pairs to it).  I use it to get a dvorak layout.  It
was posted to the list a while back--take a look at the archives.  I believe
it was called kbd.c or something.  Unfortunately I managed to lose my copy of
the src, and I can't find it in the archives or the wiki.

It doesn't work perfectly, but it's useful and small and doesn't require
messing with the kernel.

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> It doesn't work perfectly, but it's useful and small and doesn't require
> messing with the kernel.

the latter was one of my reasons not to mess with it.  i'm too old to
hack through crash dumps and broken file-systems.  well, that's not
strictly true, but i've done my share.

Re: [9fans] Foreign keyboard layout

[Jesus A. Mora]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1465 bytes --]



On Wed, 19 Sep 2001, Quinn Dunkan wrote:

> There's a program that provides its own /dev/cons that translates from the old
> /dev/cons and provides /dev/kbd to configure the translation (you write
> old_unicode new_unicode pairs to it).  I use it to get a dvorak layout.  It
> was posted to the list a while back--take a look at the archives.  I believe
> it was called kbd.c or something.  Unfortunately I managed to lose my copy of
> the src, and I can't find it in the archives or the wiki.
>
> It doesn't work perfectly, but it's useful and small and doesn't require
> messing with the kernel.
>

The link in Charles Forsyth's home page apparently leads you to
nowhere. Searching in the web, I've managed to find it anywhere (hope this
is the stuff you're talking about). Enclosed you'll find a copy of
kbd.tgz, if you are interested in it.

Also, looking back in the 9fans archive, I've realized this is a recurrent
topic. One of the messages refers to another possible solution,
contributed by Dave Presotto. You'll find it attached as kbdmap.c.gz, for
the sake of completion.

I don't know yet whether they work as expected, but will try to take a
look at them and get them working this weekend. Being a newcomer to Plan
9, this may be a pretty challenging task for me, but I think it worths to
make an effort.

Many thanks to both Boyd and Quinn, and to all the fellows in 9fans, for
your advice and your patience.

JAMM

[-- Attachment #2: Type: APPLICATION/octet-stream, Size: 3928 bytes --]

[-- Attachment #3: Type: APPLICATION/octet-stream, Size: 4200 bytes --]

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> i think i have found the vileness.

me, being a bonehead, never having read any pc h/w doc (arch yes)
and the cryptic comment in the int handler meant more kernel
print's -- yuk.  anway i have a version for azerty.  it needs
a code review and a good thrashing to death, but i think it's ok.

/lib/azerty lets me type:

    1234567890 +
    &�"'(-�_��)=
     ~#{[|`\^@]}

there is some other junk to be mapped in, but that's easy and
they're not important like:

     ������

just one edit and a:

    cp /lib/azerty /dev/kbmap

haven't trashed a f/s yet :)  backing up onto cd-rw right now.

for those who want to suffer, read on:

    certain scan codes are 2 byte sequences starting with a
    0xE0 followed by the 7 bit scan code, the 8th bit indicates
    keyup.  this is just great when it meets and mapping of
    � (iirc).

    ghastly, just ghastly

a lot of _loud_ toto was involved.  don't ask why.
well, what they play is junk, but it's well executed.

btw: in french 'toto' is used like 'foo'.

bbtw: it's nice to use a real system again.

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

i think i have found the vileness.  it's 'cos the Latin
key is to the right of the space key and it's scan code
is _known_ by the interrupt handler.  kernel print's sorted
that out -- ugh.  then i remembered it was like that back
in '92 or so, iirc.

on azerty you have:

    alt - space - alt-gr

alt/Latin is handled in a weird way and the scan code for
alt-gr is 0xE0, which is 'known' -- not for very much longer.

i'm pretty sure in japan i've seen similar keyboards
that have kana and there's a shift on the right of the
keyboard for that, but that's 7 years ago.

alt-gr is a shift used on the 1234567890 keys and
a few others, which gives you  ~#{[|`\^@]}� (actually
there are two ^'s).

my plan is to fix the driver, without breaking it for
qwerty and then loading up /lib/azerty into /dev/kbmap.

then i attack (or is it charm? :) python.

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> hence the comment in kbd.c:
>
> KF= 0xF000, /* function key (begin Unicode private space) */

yes the unicode 3.0 spec says that it's E000-F8FF and
that the top end should used by 'corporate' stuff and
bottom half for 'user' stuff.

i changed Spec to be 0xF800 and added a 3rd map table.
i gave up, for tonight, when i ran into scancode hell.
will look at it tomorrow.

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> hence the comment in kbd.c:

i have to read the driver.  i've skimmed it, but i haven't
understood it (well, not to any degree that i'm satisfied
with).  i'll have a read of it tomorrow.

Re: [9fans] Foreign keyboard layout

[Russ Cox]

[-- Attachment #1: Type: text/plain, Size: 236 bytes --]

hence the comment in kbd.c:

	KF=		0xF000,	/* function key (begin Unicode private space) */

my comment was more that looking at the code
i think nothing really depends on Spec being 0x80
and it could just as easily be 0xF000.


[-- Attachment #2: Type: message/rfc822, Size: 1884 bytes --]

From: "Boyd Roberts" <boyd@fr.inter.net>
To: <9fans@cse.psu.edu>
Subject: Re: [9fans] Foreign keyboard layout
Date: Thu, 27 Sep 2001 00:58:41 +0200
Message-ID: <01a401c146de$c989f470$a2b9c6d4@SOMA>

> i think you can safely ,s/Spec/KF/g and the
> keyboard driver will still work, without
> any 0x80 bits.

brucee tells me:

# Runes for special purposes (0xe800-0xfdff is Private Use Area)

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> i think you can safely ,s/Spec/KF/g and the
> keyboard driver will still work, without
> any 0x80 bits.

brucee tells me:

# Runes for special purposes (0xe800-0xfdff is Private Use Area)

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

i think i'll steal some stuff from a reserved code space.

Re: [9fans] Foreign keyboard layout

[Russ Cox]

i think you can safely ,s/Spec/KF/g and the
keyboard driver will still work, without
any 0x80 bits.

Re: [9fans] Foreign keyboard layout

[forsyth]

[-- Attachment #1: Type: text/plain, Size: 145 bytes --]

you can map Shift and some of the other special keys
elsewhere in the Unicode space; it's not quite like
the use of 0200 by the old shells.


[-- Attachment #2: Type: message/rfc822, Size: 2790 bytes --]

From: "Boyd Roberts" <boyd@fr.inter.net>
To: <9fans@cse.psu.edu>
Subject: Re: [9fans] Foreign keyboard layout
Date: Wed, 26 Sep 2001 23:43:44 +0200
Message-ID: <00f401c146d4$512f7400$a2b9c6d4@SOMA>

azerty support is advancing, but i'm now at the point
of a pc/kbd.c hack or re-write to get it support certain
latin1 chars and shift/alt/alt-gr to work right.

the problem, but i'll have to study the code, is that it
uses the 8th bit to indicate a shifted char and then a series
of chars in the bottom end of the latin1 codespace to indicate
shift/ctrl/alt/etc.

this is a big problem because the runes are overloaded 'cos
an azerty keyboard 'generates' stuff with the 8th bit set.

this is reminiscent of the way the bourne shell used the
8th bit to indicate quoting -- yuk, and that's before
we get to the sigsegv memory management ...

a quick hack would be to use the 16th bit, but that may
provoke exactly the same problem further down the track,
which i most certainly want to avoid.

i hope i don't have to read the unicode spec -- gag ...

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

azerty support is advancing, but i'm now at the point
of a pc/kbd.c hack or re-write to get it support certain
latin1 chars and shift/alt/alt-gr to work right.

the problem, but i'll have to study the code, is that it
uses the 8th bit to indicate a shifted char and then a series
of chars in the bottom end of the latin1 codespace to indicate
shift/ctrl/alt/etc.

this is a big problem because the runes are overloaded 'cos
an azerty keyboard 'generates' stuff with the 8th bit set.

this is reminiscent of the way the bourne shell used the
8th bit to indicate quoting -- yuk, and that's before
we get to the sigsegv memory management ...

a quick hack would be to use the 16th bit, but that may
provoke exactly the same problem further down the track,
which i most certainly want to avoid.

i hope i don't have to read the unicode spec -- gag ...

Re: [9fans] Foreign keyboard layout

[rog]

>   http://home.fr.inter.net/boyd/office/MVC-329S.JPG

and the gun is just in case that "weird" vaio gets up
and starts havin' a go at you?

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> and the gun is just in case that "weird" vaio gets up
> and starts havin' a go at you?

you just need the right tools to get the job done.

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

i've hacked up an rc script to create a tar backup on c:
from there it gets copied to cd-rw (4 generations) from
windoze.

this way i can hack away and if i smash something i can
get it back.

now if i could just get a kbd.c that works with the
kbdmap boddle i'll be able to type a lot faster.

i'm going to tame python's configure by auto-generating
'mkdeps' which will be included in mkfile's so the
dependencies will be quasi-automatic and the magic
in the mkfile's won't get smashed.

coma has a new home and a logitec 3 button mouse that
works perfectly on the external vaio peripheral block
(shock, horror -- those vaio's are weird):

   http://home.fr.inter.net/boyd/office/MVC-329S.JPG

another lousy 'battleship grey' day hence the grainyness
and/or overexposure.

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> so i should be able to update the package on the
> web site.

cool, thanks.

Re: [9fans] Foreign keyboard layout

[forsyth]

[-- Attachment #1: Type: text/plain, Size: 218 bytes --]

i think part of the problem is that the package
that's on the net isn't up to date.  i've got the
keyboard map stuff in one of our 3rd edition machines
so i should be able to update the package on the
web site.


[-- Attachment #2: Type: message/rfc822, Size: 2145 bytes --]

From: "Boyd Roberts" <boyd@fr.inter.net>
To: <9fans@cse.psu.edu>
Subject: Re: [9fans] Foreign keyboard layout
Date: Mon, 24 Sep 2001 19:21:00 +0200
Message-ID: <001001c1451d$47e431f0$a2b9c6d4@SOMA>

> a package that also changes kbd.c to allow its rune tables to
> be read and written.

trouble is i need a kbd.c with a sum of:

    35efc618 11837

what i have is:

    f7763504 8358

i've snooped around in the updates but can't find what i need.

Re: [9fans] Foreign keyboard layout

[Boyd Roberts]

> a package that also changes kbd.c to allow its rune tables to
> be read and written.

trouble is i need a kbd.c with a sum of:

    35efc618 11837

what i have is:

    f7763504 8358

i've snooped around in the updates but can't find what i need.

Re: [9fans] Foreign keyboard layout

[forsyth]

[-- Attachment #1: Type: text/plain, Size: 180 bytes --]

the keyboard map driver boyd mentions provides a keyboard map device, but
is distributed in a package that also changes kbd.c to allow its rune tables to
be read and written.


[-- Attachment #2: Type: message/rfc822, Size: 2282 bytes --]

To: Boyd Roberts <boyd@fr.inter.net>
Cc: "9 fans (lista correo)" <9fans@cse.psu.edu>
Subject: Re: [9fans] Foreign keyboard layout
Date: Wed, 19 Sep 2001 12:52:51 +0200 (CEST)
Message-ID: <20010919122949.H1071-100000@groucho.hvm.sas.junta-andalucia.es>



On Tue, 18 Sep 2001, Boyd Roberts wrote:

> iirc, there is a driver that has a mapping file which you
> can read/write tuples of the form:
>
>     shift scancode rune
>
Thanks. Please, do you remember what driver and/or file(s) are you talking
about?
I've found hardcoded keyboard mapping in both:
	/sys/src/9/$cputype/kbd.c       (machine-dependent kernel code)
	/sys/src/libcontrol/keyboard.c
Maybe there are other files like these.

TIA,

JAMM