Re: [9fans] runebase breaking libs build

2015-01-21 Thread erik quanstrom
 Whenever my mind wonders off thinking about revision control in Plan
 9, my first thought is how useful proto would be as a building block
 for that and other applications.  Yet it very seldom gets any mention,
 I really don't think it ought to be forgotten.

proto is how i build 9atom, so it's not like it's forgotten.  just more 
complicated.

- erik



Re: [9fans] runebase breaking libs build

2015-01-20 Thread Zbigniew Radosz
Hi 9fans,

I also attempted to build amd64 version some time ago, and it failed in
libc. I gave up then, but since someone brought this up, let me ask my
question here.

Bell Labs doesn't distribute amd64 binaries. It seems
 you are running a compiler built before the long rune
 change (before May 2013), but the sources are more recent.

 In this case, you need to re-bootstrap the compiler,
 as described here:

 http://9fans.net/archive/2013/05/81


I followed this one and it solves the runebase problem, but libc build
still fails as there is no amd64 subdir in libc. Looks like the labs
version does not have amd64 code for libc. Is that correct? And is libc the
only missing part?

-- 
Zbigniew Radosz


Re: [9fans] runebase breaking libs build

2015-01-20 Thread Quintile
there are a few subdirectories that need to be created when you build the amd64 
code for the first time. maybe it would be nice to have a /sys/lib/newobjtype 
script to help, but  if you just:

cd /sys/src
objtype=amd64 mk install

and create the directories as needed you should be ok

-Steve





On 20 Jan 2015, at 09:50, David du Colombier 0in...@gmail.com wrote:

 I followed this one and it solves the runebase problem, but libc build still
 fails as there is no amd64 subdir in libc. Looks like the labs version does
 not have amd64 code for libc. Is that correct? And is libc the only missing
 part?
 
 Yes, some parts of the amd64 user space are missing in
 the Bell Labs distribution. You have you apply these patches:
 
 hget http://9legacy.org/9legacy/patch/amd64.diff | ape/patch -p0
 hget http://9legacy.org/9legacy/patch/amd64-fix.diff | ape/patch -p0
 
 Also, you have to create the sub-directories in /amd64/bin.
 
 You can take a look to my notes available here:
 
 http://9legacy.org/9legacy/doc/nix/usage
 
 -- 
 David du Colombier



Re: [9fans] runebase breaking libs build

2015-01-20 Thread Zbigniew Radosz

 Yes, some parts of the amd64 user space are missing in
 the Bell Labs distribution. You have you apply these patches:

 hget http://9legacy.org/9legacy/patch/amd64.diff | ape/patch -p0
 hget http://9legacy.org/9legacy/patch/amd64-fix.diff | ape/patch -p0

 Also, you have to create the sub-directories in /amd64/bin.

 You can take a look to my notes available here:

 http://9legacy.org/9legacy/doc/nix/usage


Thanks David, that worked fine. Now I just need a 64-bit kernel :) I saw
the nix and 9k notes on 9legacy. I'll give it a shot.

-- 

Zbigniew Radosz


Re: [9fans] runebase breaking libs build

2015-01-20 Thread David du Colombier
 I followed this one and it solves the runebase problem, but libc build still
 fails as there is no amd64 subdir in libc. Looks like the labs version does
 not have amd64 code for libc. Is that correct? And is libc the only missing
 part?

Yes, some parts of the amd64 user space are missing in
the Bell Labs distribution. You have you apply these patches:

hget http://9legacy.org/9legacy/patch/amd64.diff | ape/patch -p0
hget http://9legacy.org/9legacy/patch/amd64-fix.diff | ape/patch -p0

Also, you have to create the sub-directories in /amd64/bin.

You can take a look to my notes available here:

http://9legacy.org/9legacy/doc/nix/usage

-- 
David du Colombier



Re: [9fans] runebase breaking libs build

2015-01-20 Thread erik quanstrom
On Tue Jan 20 04:10:48 PST 2015, st...@quintile.net wrote:
 there are a few subdirectories that need to be created when you build the 
 amd64 code for the first time. maybe it would be nice to have a 
 /sys/lib/newobjtype script to help, but  if you just:
 
 cd /sys/src
 objtype=amd64 mk install
 
 and create the directories as needed you should be ok

objtype=amd64 cd $objtype/bin  mkdir -p ape auth aux bitsy dial disk fossil 
fs games ip ip/httpd ndb rune opc replica upas venti usb

we don't add arches very often, so i've just created the full set of 
directories in 9atom.

by the way, the default 9atom install is amd64.  so there's no need to convert. 
 it will
also build 386 and arm exes at install time, too.

- erik



Re: [9fans] runebase breaking libs build

2015-01-20 Thread lucio
 mkdir -p ape auth aux bitsy dial disk fossil fs games ip ip/httpd ndb rune 
 opc replica upas venti usb

Shouldn't this be in a proto file?

bin
ape
auth
aux
bitsy
dial
disk
fossil
fs
games
ip
httpd
ndb
rune
opc
replica
upas
venti
usb

Whenever my mind wonders off thinking about revision control in Plan
9, my first thought is how useful proto would be as a building block
for that and other applications.  Yet it very seldom gets any mention,
I really don't think it ought to be forgotten.

Lucio.

PS: Off topic, the last posting to comp.os.plan9 in Google seems to
date back to September 2013.  Am I the first to notice?




Re: [9fans] runebase breaking libs build

2015-01-18 Thread erik quanstrom
On Sun Jan 18 20:10:38 PST 2015, don.bai...@gmail.com wrote:
 The build scripts should probably address this so people that haven't updated 
 in a while aren't spinning wheels. 
 

maybe, but libc was changed in 2013, so, even gentoo doesn't handle this case.

- erik



Re: [9fans] runebase breaking libs build

2015-01-18 Thread Bakul Shah
On Sun, 18 Jan 2015 22:09:02 MST Don Bailey don.bai...@gmail.com wrote:
 So this issue is clearly documented somewhere that people can catch up 
 quickly?

From an old message by David du Colombier:

1. remove runebase from /sys/src/libc/port/mkfile
2. cd /sys/src/libc  mk install  mk clean
3. for(i in /sys/src/cmd/8?) {cd $i  mk install  mk clean}
4. add runebase in /sys/src/libc/port/mkfile
5. cd /sys/src/libc  mk install  mk clean
6. cd /sys/src  mk install  mk clean



Re: [9fans] runebase breaking libs build

2015-01-18 Thread Don Bailey
This will be the last email I send regarding this issue as I have no
interest in getting into a long drawn-out 9fans-style discussion.

I upgraded via pull, then `{cd /sys/src/; mk objtype='amd64' install};
which broke at libc. If the issue is the compilers not being built first,
maybe it could be documented on the wiki that they need to be built first,
otherwise odd errors would occur during a world-build.

Thanks.

On Sun, Jan 18, 2015 at 10:16 PM, erik quanstrom quans...@quanstro.net
wrote:

 On Sun Jan 18 21:05:02 PST 2015, don.bai...@gmail.com wrote:
  So this issue is clearly documented somewhere that people can catch up
 quickly?
 
 
 
   On Jan 18, 2015, at 9:14 PM, erik quanstrom quans...@quanstro.net
 wrote:
  
   On Sun Jan 18 20:10:38 PST 2015, don.bai...@gmail.com wrote:
   The build scripts should probably address this so people that haven't
 updated in a while aren't spinning wheels.
  
   maybe, but libc was changed in 2013, so, even gentoo doesn't handle
 this case.

 i don't follow.  pull from sources doesn't compile anything.  did you
 update by hand?
 if so, why would hand-update be well-documented.  by definition, one would
 think that
 it couldn't be.  (since every hand update is different.)

 - erik




Re: [9fans] runebase breaking libs build

2015-01-18 Thread Don Bailey
So this issue is clearly documented somewhere that people can catch up quickly?



 On Jan 18, 2015, at 9:14 PM, erik quanstrom quans...@quanstro.net wrote:
 
 On Sun Jan 18 20:10:38 PST 2015, don.bai...@gmail.com wrote:
 The build scripts should probably address this so people that haven't 
 updated in a while aren't spinning wheels.
 
 maybe, but libc was changed in 2013, so, even gentoo doesn't handle this case.
 
 - erik
 



Re: [9fans] runebase breaking libs build

2015-01-18 Thread David du Colombier
 I upgraded via pull, then `{cd /sys/src/; mk objtype='amd64' install};
 which broke at libc. If the issue is the compilers not being built
 first, maybe it could be documented on the wiki that they need to be
 built first, otherwise odd errors would occur during a world-build.

Bell Labs doesn't distribute amd64 binaries. It seems
you are running a compiler built before the long rune
change (before May 2013), but the sources are more recent.

In this case, you need to re-bootstrap the compiler,
as described here:

http://9fans.net/archive/2013/05/81

Alternatively, you can simply recompile your amd64
binaries from a 386 machine, since the 386 binaries
on /n/sources are already up-to-date.

-- 
David du Colombier



Re: [9fans] runebase breaking libs build

2015-01-18 Thread erik quanstrom
On Sun Jan 18 21:36:26 PST 2015, don.bai...@gmail.com wrote:

 This will be the last email I send regarding this issue as I have no
 interest in getting into a long drawn-out 9fans-style discussion.
 
 I upgraded via pull, then `{cd /sys/src/; mk objtype='amd64' install};
 which broke at libc. If the issue is the compilers not being built first,
 maybe it could be documented on the wiki that they need to be built first,
 otherwise odd errors would occur during a world-build.

sorry if that came off snarky.  i am however thinking that it's more likely
a build error on sources, or other snafu rather than something that is
a systematic issue.

and assuming that you were running on 386
so there are executables to pull, it sounds like the error
was that the correct executables were not built.
do your 386/6[acl] dates and sums match sources?

do you think that could be it?  if so, it sounds like 6[acl] need to rebuilt on
sources.  

- erik



Re: [9fans] runebase breaking libs build

2015-01-18 Thread Don Bailey
The build scripts should probably address this so people that haven't updated 
in a while aren't spinning wheels. 

Thanks.



 On Jan 18, 2015, at 9:01 PM, erik quanstrom quans...@quanstro.net wrote:
 
 On Sun Jan 18 15:19:46 PST 2015, don.bai...@gmail.com wrote:
 
 Hi All,
 
 Looks like a change in libc/port/runebase.c is breaking builds. Does anyone
 have a quick fix? Looks like the issue isn't specific to a particular
 compiler, but illegal runes in the runebase.c file.
 
 runebase.c:1255 illegal rune in string
 
 the compiler needs to be updated before recompiling libc
 
 - erik
 



Re: [9fans] runebase breaking libs build

2015-01-18 Thread erik quanstrom
On Sun Jan 18 21:05:02 PST 2015, don.bai...@gmail.com wrote:
 So this issue is clearly documented somewhere that people can catch up 
 quickly?
 
 
 
  On Jan 18, 2015, at 9:14 PM, erik quanstrom quans...@quanstro.net wrote:
  
  On Sun Jan 18 20:10:38 PST 2015, don.bai...@gmail.com wrote:
  The build scripts should probably address this so people that haven't 
  updated in a while aren't spinning wheels.
  
  maybe, but libc was changed in 2013, so, even gentoo doesn't handle this 
  case.

i don't follow.  pull from sources doesn't compile anything.  did you update by 
hand?
if so, why would hand-update be well-documented.  by definition, one would 
think that
it couldn't be.  (since every hand update is different.)

- erik



[9fans] runebase breaking libs build

2015-01-18 Thread Don Bailey
Hi All,

Looks like a change in libc/port/runebase.c is breaking builds. Does anyone
have a quick fix? Looks like the issue isn't specific to a particular
compiler, but illegal runes in the runebase.c file.

runebase.c:1255 illegal rune in string
...

Thanks,
D