<kristianpaul> Fallenou: u were right about comma, sorry
<kristianpaul> undo the troubles caused by binfmt for lm32...
<kristianpaul> and got freetype to compile..
<kristianpaul> ok, back to normal, well mupdf dint compile.. but my flicernise with no pdf support is okay
<kristianpaul> Fallenou: http://paste.debian.net/107783/, when you able, tell your results with mupdf
<kristianpaul> Zzz
<Fallenou> lekernel: kristianpaul  : cannot get mupdf to link
<Fallenou> I git weird messages
<Fallenou> got*
<Fallenou> the lm32 toolchain cannot find -lfreetype -ljbig2dec -lopenjpeg -ljpeg and -lz
<kristianpaul> yeah !
<kristianpaul> same here
<Fallenou> even if it is in $RTEMS_MAKEFILE_PATH/lib
<kristianpaul> lekernel: ^^^^
<kristianpaul> yes thats weird..
<Fallenou> and the variable correctly set and so on
<kristianpaul> exactly !
<lekernel> there's nothing to link with mupdf, it's a static library only
<Fallenou> I am just following what's in the wiki
<kristianpaul> me too
<Fallenou> and actually it is trying to link cmapdump
<Fallenou> the line before the errors is
<Fallenou> LD build/release/cmapdump
<lekernel> iirc cmapdump should be compiled with your *native* toolchain
<Fallenou> yes I thought I did
<Fallenou> I will check why it is not done
<lekernel> so the Makefile should not touch it again
<lekernel> mupdf isn't made for cross compiling, so you need a bit of hand-tweaking
<kristianpaul> harcode Makefile !
<lekernel> for example compile the code generation tools manually before... then the makefile should think they're up to date and not touch them anymore
<lekernel> if you can't get that to work,remove the rules in the makefile
<Fallenou> ok it works now
<Fallenou> there seem to be a command order issue
<Fallenou> since I am sure I did the gcc -o stuff before
<Fallenou> somehow I had to do it after compilation fails
<Fallenou> and then retry to compile
<Fallenou> it's fixed for me
<Fallenou> I will try to find out the real command order and update the wiki
<kristianpaul> let me check
<Fallenou> kristianpaul: when the make build=release fails, just do again the two gcc -o commands
<Fallenou> and then again make build=release .........
<Fallenou> worked for me
<kristianpaul> let see
<Fallenou> ok flickernoise compiles and links without any error
<Fallenou> let's try it with qemu
<kristianpaul> cmapdump: could not open output file 'build/generated/cmap_unicode.c'
<kristianpaul> arg
<kristianpaul> trying agin
<Fallenou> :o
<lekernel> Fallenou: so, works?
<Fallenou> yes but not if you just follow exactly the tutorial
<Fallenou> the tutorial needs a little modification
<lekernel> ah?
<Fallenou> I think
<Fallenou> I will try again later
<lekernel> Fallenou: btw, have you used svn bisect already?
<lekernel> we need to fix that damn gcc 4.6 bug
<Fallenou> lekernel: never used svn bisect
<Fallenou> but used git bisect
<Fallenou> to spot a problem in rtems
<lekernel> ok, that's what I feared
<Fallenou> and it works damn well
<lekernel> yeah, git tools usually work well
<lekernel> svn tools don't
<Fallenou> I try as much as I can to stay away from svn and cvs
<lekernel> iirc there's an unofficial git mirror of the gcc svn, maybe i'll try with that instead of messing with svn
<Fallenou> hum there is a missing line too in qemu wiki
<Fallenou> should add git pull origin milkymist:milkymist
<Fallenou> right after the clone
<Fallenou> to get the milkymist branch
<Fallenou> by default a simple git clone just pulls the master branch
<Fallenou> on which there is no milkymist file
<lekernel> yeah or just git checkout milkymist
<Fallenou> lekernel: when starting flickernoise under qemu
<Fallenou> I have infinite loop with "milkymist_memcard: read more cmd bytes than available.
<Fallenou> Clipping
<Fallenou> and nothing shows up, the screen stays black
<lekernel> obviously, it's a problem with the memory card emulation
<lekernel> try disabling it
<lekernel> or fixing the driver (mwalle posted about the problem on the mailing list a while ago)
<Fallenou> ok
<kristianpaul> Fallenou: whicj version of mupdf r u using?
<kristianpaul> wich*
<Fallenou> the one in milkymist wiki
<kristianpaul> the file build/generated/cmap_unicode.c is in your side?
<Fallenou> i'm sorry what ?
<kristianpaul> s/side/mupdf folder
<kristianpaul> thats my build log http://paste.debian.net/107833/
<Fallenou> I closed my virtual machine :x
<kristianpaul> arg !
<kristianpaul> np
<Fallenou> but I used the one in the wiki
<kristianpaul> sure
<Fallenou> rm -rf your folder and try again
<kristianpaul> hehe i did that alredy
<Fallenou> oh :(
<kristianpaul> lets said is gnu tools fault for now ;-)
<kristianpaul> but i noticed i got same error with make build=release ... even if i dont do the gcc -o ... before
<kristianpaul> Fallenou: my faultagain :( i forgot created the build/generated folder..
<kristianpaul> It is okay now
<Fallenou> ah yes I had to create it too
<rejon> lekernel where did we put that gsoc app?
<rejon> from last year
<rejon> did they announce that yet?
<rejon> should get that underway
<lekernel> well it's pretty outdated now
<rejon> was it on a wiki?
<lekernel> yes, second
<rejon> i'm more focused now on this...b4 was juggling statusnet massively
<rejon> what are top three things milkymist needs?
<lekernel> since then, the software situation has changed a lot though
<rejon> but just off the top of your head?
<rejon> (i'm writing an email)
<lekernel> first, you can add the linux port (drivers and so)... since I'm not sure about how this will work, let's put a non critical task :)
<rejon> yeah, agree...gsoc is for nice features, but not critical tasks
<lekernel> then research on multitouch interfaces... but I'm not sure if and how this could really be part of gsoc
<lekernel> probably not a good idea (needs expensive hardware at the student side)
<rejon> yeah, seems like a separate project
<rejon> a cool controller would be nice
<lekernel> ah, got an idea
<lekernel> implement on-video controls into flickernoise
<lekernel> no, I don't read makezine, too much crap gets posted there
<rejon> yeah, but that is cool controller
<lekernel> then implement upgrade over the internet in flickernoise
<lekernel> as well as a "patch store" i.e. a simple online interface, accessible from the M1, where people can browse and upload patches
<rejon> that is cool
<rejon> i wonder how we could get that patch store done?
<lekernel> well it's pretty easy... one part of web and one part of C :)
<lekernel> there's already unix-like tcp/ip networking available on the board
<lekernel> with the BSD stack
<lekernel> the patch store should give the option to the user to try the proposed patches with a single click
<lekernel> and then download them with another click
<lekernel> ah, we'll need a filemanager too... unfortunately we can't re-use one since they all need X11 or other crazy libraries
<lekernel> though it's a rather critical task (atm there's no possibility to delete or rename files in the GUI for example) so I'm not really sure it's fit for gsoc
<lekernel> but you can still add in the task list, we'll see
<rejon> yes, totally
<rejon> what about fixes on the board
<lekernel> everything should be fine with the board now
<lekernel> the fpga design still has a few issues with USB, but it's a super crappy job and also needs some hardware
<lekernel> oh, you can still also add one task to implement OHCI into the USB controller
<lekernel> same with linux, it would be nice to have (enables a lot of devices and stuff to be supported) but non-critical
<lekernel> so what I see:
<lekernel> 1. Linux port (device drivers)
<lekernel> 2. patch store + online upgrade
<lekernel> 3. on-video controls
<lekernel> 4. file manager
<lekernel> 5. ohci
<lekernel> also, maybe LM32 MMU, but I doubt there are many motivated and competent people for that
<rejon> hahaha
<lekernel> what? :)
<lekernel> every geek is asking about it, but no one fucking does it
<kristianpaul> ask != need
<lekernel> oh, totally, it's a non-critical task for me :)
<lekernel> that's why it might belong in gsoc
<rejon> ok cool
<kristianpaul> ah,so gcc is critical for you now :-)
<kristianpaul> eventought you should consider it for gsoc i think.
<rejon> lekernel can you update http://www.milkymist.org/wiki/ to use cc by-sa 3.0
<rejon> :)
<lekernel> it's already cc by sa
<rejon> really
<rejon> i only see GFDL at the bottom of the wiki
<lekernel> Content is available under GNU Free Documentation License 1.3 and CC-BY-SA 3.0 Unported.
<rejon> oh, the button isn't on the page
<rejon> aha
<rejon> weird
<rejon> lekernel you have moderation on on that wiki
<lekernel> huh?
<lekernel> what's that
<rejon> i register no remail, no login
<rejon> i registered (i can't believe i asn't b4)
<lekernel> I have simply disabled account creation, because spam bots and mediawiki's complete suckiness in dealing with them makes it impossible
<lekernel> iirc you already have an account
<lekernel> seriously, mediawiki doesn't even let you delete an account and revoke all its edits...
<rejon> ok, yep
<rejon> rejon
<mwalle> Fallenou: you have to supply a sd card image, eg -sd <img filename>
<lekernel> and doesn't even ship with a captcha by default, which you have to painfully patch in manually
<lekernel> how ridiculous...
<rejon> fuckin ridiculus
<rejon> ;)
<mwalle> Fallenou: at least as a workaround until i'm pushing the fix ;)
<lekernel> mwalle: larsc: any task you'd like to see in gsoc? (if the gsoc thing happens at all)
<Fallenou> hi
<Fallenou> ok thanks mwalle :)
<mwalle> lekernel: mh no good idea atm, beside the already mentioned linux port
<mwalle> Fallenou: of course you have to create that image first :) just use dd and your desired size
<Fallenou> yes sure I gues whatever empty file created by dd if=/dev/zero will be good ?
<Fallenou> or do I have to format it ?
<mwalle> Fallenou: empty file is enough
<Fallenou> ok thanks
<Fallenou> just transfered 30 MB file without rx fifo overflow on qemu
<Fallenou> will try tcp ping as kristianpaul suggested
<lekernel> Fallenou: the rx fifo overflow problem can't happen on QEMU
<lekernel> there's simply no RX fifo on QEMU :)
<lekernel> (as far as I know)
<Fallenou> yes that's what I guess too :/
<lekernel> it could also be a hardware issue, though it shouldn't cause any software crash when it happens and only cause a few packets lost
<Fallenou> anyway I will try to do the tcp ping anyway
<Fallenou> err ttcp*
<Fallenou> to see what happens :)
<lekernel> btw the rx buffer size problems could explain a lot of the troubles we've had
<Fallenou> yes
<Fallenou> I think
<Fallenou> but kristianpaul said the buffer overflow keeps happening
<Fallenou> even with this fix
<Fallenou> so there is another problem
<Fallenou> I will try to do what you said
<Fallenou> a pool of buffer
<larsc> maybe i'm missing something, but isn't the overflow the result of data being send faster then it can be processed?
<lekernel> it can be either
<lekernel> 1. data within a packet received faster than the DMA master can write them to memory
<lekernel> which would have to be fixed in hardware
<lekernel> 2. any hardware bug, e.g. not taking the DMA address of another loaded slot
<lekernel> 3. software not refilling slots fast enough
<lekernel> in either case, the board should not crash, only lose packets
<larsc> but it crashes?
<lekernel> unless the DMA engine goes berserk and corrupts memory, but I don't think that's very likely
<lekernel> it used to
<lekernel> but Fallenou recently spotted a RX buffer that wasn't large enough to hold complete packets, which is a nice candidates for the crashes
<lekernel> I have not tested since then
<Fallenou> but you do NFS and network boot don't you ?
<lekernel> yeah, but I have not done so since you patched the buffer size
<Fallenou> ok
<Fallenou> and it used to crash sometime ? just by doing NFS and network boot ?
<lekernel> it crashed all the time... sometimes only the broadcasts when on a shared ethernet network locked it up
<Fallenou> oh ok
<Fallenou> so you will clearly be able to see if there is an amelioration or not
<Fallenou> that's good to know :)
<Fallenou> just did a ttcp test between his ubuntu VM and milkymist bsp
<Fallenou> I got 2,111 Mo/sec
<Fallenou> not very fast :o
<rejon> lekernel
<rejon> ok updated
<rejon> and I emailed my google contacts to let them know we are applying
<rejon> canvassing this shit
<rejon> can you please update
<rejon> emailed the list
<roh> nice
<lekernel> ah, new feature: i18n
<lekernel> can translating software (and implementing language switching code) count as a gsoc project?
<Fallenou> rejon: maybe rename the page
<Fallenou> (and put an auto forward)
<lekernel> it's done
<Fallenou> oh ok
<rejon> i did
<rejon> hahaha
<rejon> :)
<rejon> lekernel i think that is too easy
<rejon> if you can obfuscate that a bit more, yet!
<rejon> yes!
<lekernel> btw anyone get what the heck this is about? https://github.com/opencpi/opencpi
<Fallenou> another RTOS ?
<Fallenou> or RT middleware
<lekernel> nope
<Fallenou> CORBA
<Fallenou> pukes
<lekernel> sounds like Java for electronics
<Fallenou> ahahah
<lekernel> with a touch of cloud computing
<Fallenou> stop it stop it
<Fallenou> pleaase stop