digging around in startx

[twobob]

Originally Posted by hawhill

I mentioned the problems with nibble ordering for framebuffer output. I played a bit and what I've found is that a combination of Xvfb and x11vnc (both built with the buildroot environment) plus kindlevncviewer does a resonably good job. It's a bit awkward to use three programs in the place of one X server, but x11vnc does a good job recognizing changed parts of the screen and kindlevncviewer knows about eink updates. Together, they provide a nice bundle. Memory footprint is about ~20MByte RSS.

I did not explore this further, but x11vnc has options to start an Xvfb server automatically upon vnc client connection.

I was still using this with SSH-forwarded X11 applications (and even a window manager) from the USB host machine. It would be interesting to see if the K3 could be switched over to a Debian-ARM-based userland, running Xvnc (I think Debian has this, would probably work better than Xvfb+x11vnc) plus kindlevncviewer for the graphical UI.

but how? I have read the man page end to end several times. read the help. read guides. googled it. BAH! how :\

I thought it was -listen or -connect - or both. or something? but I just cant find a worked example nor can I find it in the docs sigh.

[hawhill]

This was the answer to your earlier post, see below for an answer to your latest...:

that's great :-) I would advise against Xfbdev in such a scenario, since both Xfbdev and kindlevncviewer would use the framebuffer in such a case and this might cause conflicts or artefacts, I think.

Using VNC, even with "copyrect", still means that data is shoveled through the kernel's networking layer and replicated onto the framebuffer. I think it would be interesting to patch x11vnc to _not_ encode recognized changes and send them via VNC, but plug kindlevncviewer's screen update method into the code at that point. Will see if I can come up with something like this...

Well, best thing to have would actually be a working Xfbdev. I didn't dig too much into the XOrg code yet, but maybe that would make even more sense. There should be some way to overcome that nibble-in-byte order problem...

---snip---

The documentation for x11vnc is in fact, erm, extensive: http://www.karlrunge.com/x11vnc/
(plus the FAQ!)

See the various mentions of Xvfb, Xdummy (need to check that one, I think buildroot can build it too with the modular Xorg!) and x11vnc's -create and -display switches. In fact, the amount of documentation makes my head hurt :-)

[twobob]

[root@kindle root]# x11vnc -noxdamage -forever -shared -noncache -shiftcmap 240 -notruecolor -vncconnect -nopw -solid white -modtweak -xkb -add_keysyms -wireframe -wcr always -scr_area 50000 -nap -threads

decent options. now just need to weld the Xvfb -fp /mnt/us/sh/fonts/X11/100dpi/ -noreset -screen 0 600x800x8 -bs -wr into that

EDIT: No - I won't bother. it would be too fiddly whilever we are using shared logins.

I will keep it separate and just script it up as /mnt/us/usr/bin/xvnc and /mnt/us/usr/bin/xvfb that's simple enough. shove a sleep 5 in there and you have a workable startup script. Chuck a few xset's at it and I think that's as good as you get without - as you rightly say - going the extra mile to fix the tty issues.

I was considering a workaround of some description as the config for the 4 way etc are actually referenced (if missing?) in the old /usr/bin/startx which makes me at least think I can get a likely xkb configuration happening. that gives me some faith

I am no "developer of repute" so if you do find the time in your probably stupidly busy life to have a further look at the current buildroot sources that would be great. your code amends backport nicely in the xorg-server arena just FYI. I applied them to an older version that I build against with no problems, so I guess that VT code hasn't been touched in a while more faith that the work may not bit rot quickly.

Thanks for your amazing works mate. Really pleased to be able to hack something REALLY workable (and flexible) together in under a few days. Brilliant. Can't wait to show my buddy to amazing improvements that have happened in just a few days. : )

[twobob]

Originally Posted by hawhill

The documentation for x11vnc is in fact, erm, extensive: http://www.karlrunge.com/x11vnc/
(plus the FAQ!)

See the various mentions of Xvfb, Xdummy (need to check that one, I think buildroot can build it too with the modular Xorg!) and x11vnc's -create and -display switches. In fact, the amount of documentation makes my head hurt :-)

Ah yes, I did try that one and egrep was invoked? I will give it another go as I have egrep built now. (yes I know it's an alias)

anyways thanks.

With a bit more massage I got quite a bit more running.
Running much nicer now with a bit of switch love.

: )

EDIT: I'll give the create thing a try again

[twobob]

[root@kindle root]# strace -o /mnt/us/x11vnc-strace-init.txt x11vnc -noxdamage -forever -shared -noncache -nopw -solid white -modtweak -xkb -add_keysyms -wireframe -wcr always -scr_area 50000 -nap -display WAIT:cmd=CREATEDISPLAY-Xvfb -create_xsrv "Xvfb -fp /mnt/us/sh/fonts/X11/100dpi/ -noreset -screen 0 600x800x8 -bs -wr"

gives us:

xauth: error in locking authority file /home/root/.Xauthority
02/11/2012 01:27:13 wait_for_client: find display cmd failed.
02/11/2012 01:27:13 wait_for_client: FINDCREATEDISPLAY cmd: /bin/sh /tmp/x11vnc-find_display.8oOpHL Xvfb
The setpgrp() function is unimplemented at -e line 1.
02/11/2012 01:27:13 wait_for_client: read failed: /bin/sh /tmp/x11vnc-find_display.8oOpHL Xvfb
02/11/2012 01:27:13 fgets: Bad file descriptor

strace show us that it is the -w flag for grep, a call to egrep, some calls to ps that fail...

so supporting structure issues. adding ENV to .bashrc didn't help, need to trace where non interactive processes include again (I'm sure I already did this once, should read my own threads )

thanks for your help. busyboxes shortcomings are to blame.

[twobob]

[root@kindle root]# x11vnc -display WAIT:cmd=FINDCREATEDISPLAY-print > /mnt/us/find_create_display.txt

gives us... attached.
TODO: fixup the script to be busybox 1.7.2 happy or "hardwire" it to 1.20.x calls

at least I found it

[hawhill]

Oh dear, a bit to general and thus too complicated for our purposes, I guess.

[twobob]

Originally Posted by hawhill

Oh dear, a bit to general and thus too complicated for our purposes, I guess.

Indeed. I have actually rewritten it to be much more simple - to reference /usr/bin/egrep etc. rather than the busybox versions and so on.

still produced odd output so it would be a debug session to get it going.

I was thinking of the overall picture of what the script actually does.
Not entirely sure it is helpful anyway in many respects.

a simple call to Xvfb via a script is probably much more helpful on known hw.

: ) So I'm ditching it. simple.

[twobob]

Originally Posted by hawhill

Oh dear, a bit to general and thus too complicated for our purposes, I guess.

Okay so call me bloody minded but I wanted to see what the script actually did.

essentially it ends up making this call

[root@kindle root]# /mnt/us/usr/bin/xinit /home/root/.xinitrc -- /mnt/us/usr/bin/Xvfb :1 -screen 0 1280x1024x24 -cc 4 -nolisten tcp -auth /tmp/.xas54195700.jHQkYI

[dix] Could not init font path element /usr/share/fonts/X11/TTF/, removing from list!
[dix] Could not init font path element /usr/share/fonts/X11/OTF/, removing from list!
[dix] Could not init font path element /usr/share/fonts/X11/Type1/, removing from list!
/mnt/us/usr/bin/xinit: XFree86_VT property unexpectedly has 0 items instead of 1
/mnt/us/usr/bin/xinit: connection to X server lost

waiting for X server to shut down

so my patched version of the script did actually run to it's conclusion. Looks like xinit would need some love to make this work on the kindle.

Anyways, I've lost interest, it's nothing we can make ourselves in a basic script. no magic here. It gets the size. colours and preferred method of startup wrong... Err. no.

but - for those that really enjoy hurting their brains - like me...

[root@kindle root]# XAUTHLOCALHOSTNAME=localhost CREATE_DISPLAY_OUTPUT=/mnt/us/x11vncdebug.txt x11vnc -display WAIT:cmd=/mnt/us/x11display.txt

would be enough to get you debugging at least. (updated script with explicit calls to grep and friends attached.)

[twobob]

So a little bit more love and we have matchbox-window-manager running on the 3

Happy with that.

[twobob]

completed the mono theme for blackbox. no more visual artifacts created through dither.

Into the bargain the eink runs a bit faster

[geekmaster]

Originally Posted by twobob

completed the mono theme for blackbox. no more visual artifacts created through dither.

Into the bargain the eink runs a bit faster

That is a great accomplishment! Apps that use "old-fashioned" bitmapped fonts that are hand-tuned for a monochrome bi-level screen would look a lot nicer, if such things can still be found...

[knc1]

http://www.freedownload3.com/softwar...trix_font.html

http://www.macusergroup.com/fonts/index.html

I know I have a 24-pin Epson in the storage shed, do not have a clue if (or where) I have the downloadable fonts package for it.

[geekmaster]

Originally Posted by knc1

http://www.freedownload3.com/softwar...trix_font.html

http://www.macusergroup.com/fonts/index.html

I know I have a 24-pin Epson in the storage shed, do not have a clue if (or where) I have the downloadable fonts package for it.

I have about 20 or so font cartridges for my old HP printer. Of course, some of them are PostScript and HPGL (for emulator a pen plotter), but most of them are bitmap fonts. But then some of those fonts (BitStream, etc.) were horribly expensive as I recall...

But there is this:
http://en.wikipedia.org/wiki/Computer_font#Bitmap_fonts

There may be some public domain bitmap font files floating around too...

http://www.1001freefonts.com/bitmap-pixel-fonts.php

But of course, the app needs to support them. I used a tiny bitmap font in my "titty" eink algorithmic art script:
https://www.mobileread.com/forums/sho....php?p=2026074

[twobob]

well I managed to get a WHOLE pile of new stuff in Buildroot - whether or not it actually works yet I am not sure. might have slightly piled a bunch of bins in the wrong dir while mnt'ed rw. oops

A quick (3 hour) wipe should fix it right up though ready to test.

I may stick some more apps in the cooker while it fixes. I HOPEFULLY have gtk2 libglib2 midori, a whole pile of supporting libs and webkit built for the 3 now.

: ) yeah I'll do some more, why not