gmane.os.inferno.general

Re: Authentication basics

Pete Elmore — 2012-05-25T23:51:18

I've spoken too soon!  Authentication between machines is working fine
once the keys are exchanged, but keyfs still seems to go away somewhat
randomly (coinciding with a request for getauthinfo from another
machine), and I'm still not quite sure why.  changelogin doesn't seem
to cause it, only a remote getauthinfo.  It happens perhaps 50 or 70%
of the time.

I'm not sure if it's intended behavior or not, but if keyfs is supposed
to continue serving keys indefinitely, then there may be a bug.  I
didn't see anything obviously[1] wrong in the source code  in
tools/libstyx/styxserver.c and appl/cmd/auth/keyfs.b, and the man page
for keyfs doesn't seem to indicate that it is supposed to stop at some
point.

My best guess is "gcc/libpthread/host OS is misbehaving" but that's sort
of like "jiggle the cables, maybe it'll work after that".  I am
currently making do with:
      while {} {or {ps | grep -s Keyfs} {auth/keyfs -n /n/nvr}; sleep 1}
...but I suppose that's non-ideal.

I'm still not certain that I understand

Re: Authentication basics

Pete Elmore — 2012-05-24T21:43:18

This is immensely helpful; I can't believe I hadn't seen it.  All this
time if I wanted to do something similar, I'd been using strace to
attach to emu and watch file access syscalls, although that only shows
what gets touched on the host OS's filesystem.  iostats(4) is a
life-saver.

This OS is so well thought-out that if anything that looks out of place,
incomplete, or incorrect, it's usually because I've missed something.
It's a great problem to have in an OS.

Re: Inferno on Android

Chris Double — 2012-05-24T10:31:55

On Wed, May 23, 2012 at 12:48 PM, Chris Double
<chris.double-AmvDL87l6zFIJsTUW+i+0g< at >public.gmane.org> wrote:

I made some more progress getting it running on my device. The
capacitive hardware keys not working issue is due to having to
manually handle 'virtual keys'. See [1] for a description of the
virtual key files and having to load and parse them. Then it looks
like it needs to detect touches in the range of this and translate to
the appropriate key code.

The problem of button presses not working when running under an ICS
based android install are due to step 14 at [2]:

----------8<-------
When a tool becomes inactive (finger goes "up"), it should stop
appearing in subsequent multi-touch sync reports. When all tools
become inactive (all fingers go "up"), the driver should send an empty
sync report packet, such as SYN_MT_REPORT followed by SYN_REPORT.

Previous versions of Android expected "up" events to be reported by
sending a pressure value of 0. The old behavior was incompatible with
the Linux input

Source for mediaserver-inferno in Android port?

Chris Double — 2012-05-23T06:56:53

In the android port there's a mediaserver-inferno binary:

<https://bitbucket.org/floren/inferno/src/75008e7031e1/android-tools/audio>

Is there source for this or instructions on how to build it?

Chris

Re: Inferno on Android

Chris Double — 2012-05-23T00:48:46

On Wed, May 23, 2012 at 10:56 AM, Chris Double
<chris.double-AmvDL87l6zFIJsTUW+i+0g< at >public.gmane.org> wrote:

I installed a Gingerbread ROM and was able to get Inferno installed
and running. The touch screen works and it's pretty snappy!

There's some issues, probably due to a mismatch between the ROM I'm
using and what the port was based on. The touchscreen is slightly out
and the capacative hardware buttons just act as touchscreen pushes at
the bottom of the screen. So I couldn't test using the keyboard in the
apps since I couldn't bring them up.

I also had to start the android 'media' service as there was a log
complaining about 'audioflinger' not running that blocked emu-g from
running. This was even though the inferno specific media server was
running.

 I'll keep tweaking though and see if I can see what's going on.

Chris.

Re: Authentication basics

Charles Forsyth — 2012-05-22T23:32:50

try iostats(4)

you still need the protected name space in the end in case the program
randomly cheats (say)
but iostats can be useful to work out what ought to be in that name space.

On 5 April 2012 21:27, Pete Elmore <pete-Obkyh9HiVfY< at >public.gmane.org> wrote:

Re: Inferno on Android

Chris Double — 2012-05-22T22:56:24

Ok thanks, I'll try with a Gingerbread ROM base.

Re: Inferno on Android

John Floren — 2012-05-22T21:59:22

We built on top of Gingerbread, no idea how it works with ICS.
I wouldn't be surprised if button problems are related.
 On May 22, 2012 3:29 PM, "Chris Double" <chris.double-AmvDL87l6zFIJsTUW+i+0g< at >public.gmane.org> wrote:

Inferno on Android

Chris Double — 2012-05-22T21:09:50

I'm trying to install the Android Inferno port on a Nexus S but I'm
having issues when the phone reboots. The 'picker' application doesn't
run - I just get stuck on the 'Google' logo screen.

If I 'adb shell' into the phone and then start the inferno service
manually it runs. I'm thinking there might be some incompatibility
with the init.rc script that the inferno install puts on the device vs
the one that's already on there. Is there a 'diff' of the changes that
init.rc script made against the original init.rc it was based on? The
one on my device is quite different.

I've tried installing over a stock 4.0.3 android ROM, CyanogenMod 7
and even Mozilla's B2G. For anyone who has it working, what version of
Android did you start with? Should I be using a Gingerbread based
Android?

When I start emu-g and wm/wm manually it runs, and a mouse cursor
moves around when I touch the screen. No button presses work and the
capacitive 'hardware' buttons on the bottom of the phone don't work.
Is this related to possibly

Re: Authentication basics

Pete Elmore — 2012-05-22T18:41:24

Fortunately (if a bit frustrating :), it seems like I just had a bad build.  

It all seems to work fine (no processes disappearing, getauthinfo/changelogin/cpu/mount/dial all doing what I thought they would) after I did mk nuke, updated via hg again, reset the mkfiles, wiped all the keys I had, and started over.

So I'm not quite sure what was causing any of the problems, but they have disappeared.  Thanks to everyone for their help, and sorry it turned out to be all on my end.

I'll probably have this written up on the web before long, as promised.  The summary for anyone who would like an introduction is:

== on server ==
# If you don't run the setup functions that wm/wm runs, this fixes one bit:
if {~ $#user 0} { user = `{cat /dev/user} }
auth/createsignerkey `{cat /dev/sysname}
svc/auth # A shell script, mostly simple stuff.
auth/changelogin $user # To create/change the user's password
getauthinfo default # Using localhost as signer, same username/password

== on client ==
svc/net
# Using the server as

Re: Inferno on Mac OS X 10.7

Charles Forsyth — 2012-05-13T16:55:45

That's not a bad idea anyway, but it's a bug in the Inferno-specific code
which is likely to be wrong in the newer port.
Specifically, it transgressed an unwritten law.

On 13 May 2012 17:28, Oleksandr Iakovliev <yshurik-9wnLQIiZAfpWk0Htik3J/w< at >public.gmane.org> wrote:

Re: Inferno on Mac OS X 10.7

Oleksandr Iakovliev — 2012-05-13T16:38:59

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Forgot to add a link:

https://bitbucket.org/yshurik/inferno-os/overview

On 2012-05-13 18:28 , Oleksandr Iakovliev wrote:

-----BEGIN PGP SIGNATURE-----
Version: GnuPG/MacGPG2 v2.0.17 (Darwin)
Comment: GPGTools - http://gpgtools.org
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iF4EAREIAAYFAk+v458ACgkQeH8C/33k0Z26bQEAkK9oYI+zkbahl0nyrFQLS1sH
9sm6kwNizNxh0fpKOoABAK37+M6+Y0YpB+cGTw1EXpKKNcPv4xXS/7ZD88ihUvyL
=0Zv0
-----END PGP SIGNATURE-----

Re: Inferno on Mac OS X 10.7

Oleksandr Iakovliev — 2012-05-13T16:28:07

Garry,

You can give a try to my inferno-os branch with updated Freetype up to
2.4.6.
If it is really some freetype-specific bug, then the version may help you.

Best regards

On 2012-05-13 17:22 , Garry Taylor wrote:

Re: Inferno on Mac OS X 10.7

Charles Forsyth — 2012-05-13T15:59:02

Fixing the bug is the most likely one. If you send me the details I'll have
a look at it.
Reproducible ones are usually not that hard, although Freetype is a
significant chunk of code.
It looks as though a pointer is being used after free.

On 13 May 2012 16:22, Garry Taylor <taylor.garry-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:

Inferno on Mac OS X 10.7

Garry Taylor — 2012-05-13T15:22:11

Hello all,
I am trying to do a bit of development on Inferno, using my Mac. I've
written a small program using Freetype, but every time I "kill" my program,
it takes the whole of Inferno down with it. On my Terminal, it reports:

imac:Inferno gman$ sh run.sh
; wm/wm
alloc:D2B: 77ce8c8 in heap: in free blk at 7751020'7d920
area at 77ce8b8:  742d2065 7751020 0 80 0 1 263890 77ce8c0 0
    77ce8dc:  0 64 1e 0 0 64 1e 20 68281808 0 410c18
panic: alloc:D2B 77ce8c8 (from 42b64/4950e)

This is now getting to be fairly easy to replicate.

Any thoughts on what I can do to prevent?

Cheers

Garry

Re: Inferno on Mac OS X 10.7

Charles Forsyth — 2012-05-07T12:08:35

I've pushed those changes out, which should continue to work on older MacOS
X (with loss of drawcursor).

On 4 May 2012 09:22, Eldar Tsraev <sarelius-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:

Re: Inferno on Mac OS X 10.7

Eldar Tsraev — 2012-05-04T08:22:38

Thanks a lot for your help.

changing cc to gcc helps.
And I am already found issue about drawcursor :-) commenting body of
the function works fine.

2012/5/4 Noah Evans <noah.evans-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org>:

Re: Inferno on Mac OS X 10.7

Noah Evans — 2012-05-03T22:42:07

You can get around this by commenting out win. You may need to add a
line to the code section of your emu config file thought.

Noah


On Thu, May 3, 2012 at 11:53 PM, Joseph Stewart
<joseph.stewart-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:

Re: Inferno on Mac OS X 10.7

Serge Gagnon — 2012-05-03T22:34:23

Hello

If you think that it can help :

http://sergegagnon.com/workaround-to-build-vita-nuova-inferno-on-osx


Serge Gagnon

Le 2012-05-03 à 17:08, Eldar Tsraev <sarelius-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> a écrit :

Re: Inferno on Mac OS X 10.7

Charles Forsyth — 2012-05-03T21:57:29

I believe if you set the compiler to gcc instead of cc, in the CC and LD
values in
mkfiles/mkfile-MacOSX-386 it will work. cc is apparently now clang/llvm,
not gcc,
and it moans about that.

Re: Inferno on Mac OS X 10.7

Joseph Stewart — 2012-05-03T21:53:11

Beware:

I believe there are deprecated graphics/mouse API's used in the OSX
graphics part of emu that will cause a build fail later in the chain...

-joe

On Thu, May 3, 2012 at 2:28 PM, Noah Evans <noah.evans-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote: