**** BEGIN LOGGING AT Wed Dec 23 02:59:57 2009
Dec 23 03:08:19 <Abyssul> why
Dec 23 03:12:41 <zsoc> Abyssul: electroinduction magnet + harddrive?
Dec 23 03:26:16 <oil> rwhitby: with zenity, are you attempting a direct port?
Dec 23 03:26:41 <oil> like, what about adding other dialog types
Dec 23 03:26:59 <oil> like a "slider" type, that returns a number between 0 and 100
Dec 23 03:32:28 <jacques> I'm surprised zenity doesn't lready have one of those.
Dec 23 03:52:07 <PuffTheMagic> zenity again?
Dec 23 03:53:52 <PuffTheMagic> i cant wait for 1.3.5 to officially come out and see if anything is fixed/changed
Dec 23 03:54:03 <PuffTheMagic> compared to beta
Dec 23 04:02:57 <zsoc> PuffTheMagic: are you anti zenity? lol
Dec 23 04:03:32 <PuffTheMagic> well, unless I am mistaken, zenity is a cli tool that makes gnome/gtk guis
Dec 23 04:03:43 <PuffTheMagic> so i cant see how that is of any use on the pre
Dec 23 04:12:13 <sdodson> PuffTheMagic: when did the beta go out?
Dec 23 04:12:46 <bpadalino> i don't understand the zenity stuff either
Dec 23 04:15:08 <oil> PuffTheMagic
Dec 23 04:15:12 <oil> for patches
Dec 23 04:15:17 <oil> and optware
Dec 23 04:15:28 <oil> like a customize carrier string patch
Dec 23 04:15:45 <oil> that prompts you using mojo to enter the text you wish to have as your carrier string
Dec 23 04:16:22 <oil> when you install it
Dec 23 04:17:31 <zsoc> hmm... interactive user-input install
Dec 23 04:17:45 <oil> thats what zenity would allow the patch to do
Dec 23 04:17:52 <oil> or optware
Dec 23 04:19:24 <bpadalino> i am still confused how it draws to the screen or grabs the input i suppose .. i was also under the impression of the gnome/gtk stuff :(
Dec 23 04:19:35 <oil> like
Dec 23 04:19:40 <oil> for the question message
Dec 23 04:19:49 <oil> it comes up in the confirmation/dash area saying "yes, no?"
Dec 23 04:19:58 <oil> or whatever the params its called with tell it to
Dec 23 04:20:48 <bpadalino> so it's not really zenity per se, but the same idea as zenity allowing the commandline to send notifications for commandline input ?
Dec 23 04:21:04 <oil> right
Dec 23 04:21:06 <bpadalino> gotcha
Dec 23 04:58:23 <rwhitby> bpadalino: zenity-compatible front-end, displaying the user prompting on webos
Dec 23 04:58:47 <bpadalino> gotcha
Dec 23 04:58:48 <rwhitby> oil: yeah, we could do a superset if we want, but I do want complete zenity compatibility on the command line
Dec 23 04:59:19 <rwhitby> (i.e a superset is no problem, as long as the full zenity is there and fully compatible)
Dec 23 04:59:33 <rwhitby> the rationale is cause then we don't need to document anything - we just point to zenity
Dec 23 05:01:10 <jacques> rwhitby, I built preware completely on 32-bit gentoo, but on fedora am having a problem with their ant package
Dec 23 05:02:19 <rwhitby> jacques: ok.
Dec 23 05:03:16 <jacques> rwhitby, this mean anything to you: "The import java.net.InterfaceAddress cannot be resolved"
Dec 23 05:03:26 <rwhitby> jacques: nope
Dec 23 05:03:40 <jacques> I think it's because fedora's and package doesn't use sun's jdf
Dec 23 05:03:51 <jacques> s/jdf/jdk/
Dec 23 05:03:51 <infobot> jacques meant: I think it's because fedora's and package doesn't use sun's jdk
Dec 23 05:13:18 <rwhitby> jacques: that's good news that you could get the service stuff working fine from the repo on gentoo, cause gentoo is what PuffTheMagic uses
Dec 23 05:15:51 <PuffTheMagic> well i have an overlay of ebuilds for some older versions of stuff
Dec 23 05:17:09 <PuffTheMagic> jacques: yeah fedora switched to icedtea i believe
Dec 23 05:17:54 <rwhitby> PuffTheMagic: what I mean is that you should be able to reproduce exactly what the autobuilder does on a gentoo checkout of preware/build.git
Dec 23 05:17:57 <jacques> it's strange; they have like 4 different java engines
Dec 23 05:18:08 <jacques> openjdk is the default install
Dec 23 05:18:15 <jacques> but their ant package pulls in java-1.5.0-gcj.x86_64
Dec 23 05:18:37 <jacques> I had practically no trouble at all on gentoo
Dec 23 05:19:10 <jacques> I just had to install the sun jdk in addition to the jre which I already had
Dec 23 05:19:18 <jacques> (on gentoo)
Dec 23 05:19:43 <jacques> I have the sun jre on fedora too - it's the only thing that worked well with webosquickinstall
Dec 23 05:20:09 <jacques> I'm going to look into installing sun jdk on fedora and then maybe building my own ant
Dec 23 05:20:13 <PuffTheMagic> jacques: it works on gentoo cause i use gentoo ;)
Dec 23 05:20:13 <rwhitby> so much for "write once, run anywhere", hey?
Dec 23 05:20:38 <PuffTheMagic> it build with all javac's
Dec 23 05:20:43 <PuffTheMagic> gcj does not count
Dec 23 05:21:29 <jacques> I would probbly never look at fedora except it's what we use at work and I have to keep my hand in it
Dec 23 05:21:52 <jacques> since one of my responsibilities is making custom versions of it
Dec 23 05:24:49 <tmzt> rpm database recovery?
Dec 23 05:25:17 <tmzt> I so prefer the debian available list with the weird text corruption, fix it and it's fine
Dec 23 05:41:37 <jacques> I installed sun jdk on F12 and then used alternatives to set its javac as the default, and then my wolservice completed :-)
Dec 23 05:43:19 <jacques> (wolservice build that is)
Dec 23 05:57:32 <rwhitby> PuffTheMagic: when you call LSMessageGetPayloadJSON(message); in client.c in wIRCd, do you think the resulting json_t *object needs to be freed by the caller or not?
Dec 23 05:57:46 <rwhitby> (can't tell from Palm's header files)
Dec 23 05:58:13 <PuffTheMagic> no its a local var
Dec 23 05:58:54 <rwhitby> so you don't think that LSMessageGetPayloadJSON allocates memory?
Dec 23 05:59:20 <rwhitby> (since it could be any size hierarchy of JSON objects as a tree)
Dec 23 06:00:24 <rwhitby> I was thinking that maybe we need to call json_free_value(&object) on it after we've finished using it
Dec 23 06:00:46 <rwhitby> (I'm adding an "echo" method to the example service)
Dec 23 06:01:38 <atlanta> Hey anyone up
Dec 23 06:02:48 <jacques> gitorious is slow :-\
Dec 23 06:03:29 <atlanta> It would be a nice patch to turn off the vibrate and/or turn off sound on incoming instant messages .. Will make it more of an I'm
Dec 23 06:03:32 <rwhitby> jacques: yeah, we need to change the way we check out patches from there, so we only pull the tree once for each tag rather than once for each patch
Dec 23 06:03:33 <jacques> I'm getting ~20kiB/s on checkouts
Dec 23 06:04:04 <atlanta> Now it takes so long when it vibrates and trys to update the screen
Dec 23 06:04:09 <jacques> rwhitby, yeah I noticed that it was seemingly pulling the same thing over and over
Dec 23 06:04:29 <rwhitby> jacques: yep, I've got a very naieve implementation in there
Dec 23 06:04:58 <rwhitby> it should pull it once, tarball it into downloads, and use it from downloads if it's already there.
Dec 23 06:05:05 <jacques> agreed
Dec 23 06:06:41 <jacques> well, it's midnight here and I gotta let my friend get to sleep
Dec 23 06:06:43 <rwhitby> hmm - it's written to do that actually
Dec 23 06:06:52 <rwhitby> do there's a bug
Dec 23 06:06:52 <jacques> maybe some check is failing
Dec 23 06:06:56 <jacques> yeah
Dec 23 06:06:58 <PuffTheMagic> rwhitby: we should move all the patches to the main git server
Dec 23 06:07:04 <jacques> I'll be on from home in a little while
Dec 23 06:07:04 <PuffTheMagic> then make gitorious mirror our repo
Dec 23 06:07:10 <PuffTheMagic> untill all urls are fixed
Dec 23 06:14:08 <rwhitby> PuffTheMagic: answered my own question.  object should not be freed by the caller, just like you said (and like you have correctly coded in wIRCd)
Dec 23 06:15:11 <rwhitby> exampleservice now has an echo method
Dec 23 06:15:40 <rwhitby> next to add a launch method
Dec 23 06:16:27 <PuffTheMagic> launch method?
Dec 23 06:20:31 <PuffTheMagic> rwhitby: what is a launch method and why is it needed?
Dec 23 06:26:11 <PuffTheMagic> cant imagine why a launch method would be needed
Dec 23 06:29:32 <rwhitby> that's what you said about zenity too
Dec 23 06:31:02 <PuffTheMagic> rwhitby: i dont even get how zenity would run in luna, and if you rewrote it for luna then its no longer zenity
Dec 23 06:31:30 <PuffTheMagic> but it sounds like you are making a "is-running" method just so you can use luna-send to start it
Dec 23 06:31:38 <PuffTheMagic> whch make no sense with a C service
Dec 23 06:31:41 <rwhitby> zenity is a command line application
Dec 23 06:31:51 <rwhitby> it takes arguments on the command line
Dec 23 06:31:53 <PuffTheMagic> yeah i know what zenity is
Dec 23 06:31:56 <rwhitby> it displays a gui to the user
Dec 23 06:31:56 <PuffTheMagic> and make a gui from cli
Dec 23 06:32:09 <rwhitby> it takes a gui reponse from the suer
Dec 23 06:32:09 <PuffTheMagic> i dont see how that can be used in luna
Dec 23 06:32:21 <rwhitby> it returns the gui response on the command line
Dec 23 06:32:32 <PuffTheMagic> yeah i exactly what it is
Dec 23 06:32:52 <rwhitby> in this case the gui is not a linked in GTK library, it is a launch of a webOS card, and a service call back with the response from the user.
Dec 23 06:33:22 <PuffTheMagic> so u dont want zenity
Dec 23 06:33:23 <rwhitby> using exactly the same method that ipkgservice uses to ask the user a question
Dec 23 06:33:29 <PuffTheMagic> you want a webos/port/fork of it
Dec 23 06:33:53 <PuffTheMagic> thats a little different than just saying zenity
Dec 23 06:33:58 <rwhitby> sigh
Dec 23 06:34:28 <rwhitby> you should read the backlog more ...
Dec 23 06:34:48 <joshua> reading is boring
Dec 23 06:34:54 <PuffTheMagic> cause i have time for that
Dec 23 06:35:09 <rwhitby> yeah, and I have time to repeat myself just for your benefit too ...
Dec 23 06:36:18 <PuffTheMagic> you were the one that just brought it up....  but it would have taken 2 seconds to say a port of zenity to webos
Dec 23 06:36:20 <PuffTheMagic> or something like that
Dec 23 06:36:37 <rwhitby> someone considerate to other's time would say "oh, I've come into the middle of a discussion about zenity on webOS, perhaps I should go back and read what had been said in the last hour, rather than just coming in and saying that it's not needed without any other context for the conversation"
Dec 23 06:37:57 <PuffTheMagic> for all i knew you guys were gonna leave luna to run some directfb gtk version of zenity.... idk
Dec 23 06:38:47 <PuffTheMagic> and you brought this up not me... we were talking about the example service
Dec 23 06:38:59 <rwhitby> I'm talking about the other day, not now.
Dec 23 06:38:59 <PuffTheMagic> which should not need a "is-running" method
Dec 23 06:39:23 <rwhitby> I have no idea what you are talking about with an "is-running" method
Dec 23 06:39:44 <rwhitby> a launch method is what ipkgservice does to launch the webOS card.
Dec 23 06:40:15 <rwhitby> http://developer.palm.com/index.php?option=com_content&view=article&id=1590#ApplicationManager-launch
Dec 23 06:41:40 <rwhitby> and it what a zenity front-end will use to launch the webOS card which implements the UI to the user.
Dec 23 06:41:56 <rwhitby> and therefore needs to be supported by the C service which will be part of the zenity front end
Dec 23 06:42:39 <rwhitby> (since the service needs to recieve the response sent back from the user by the webOS card)
Dec 23 06:43:13 <PuffTheMagic> yeah i get how it works
Dec 23 06:56:01 <rwhitby> PuffTheMagic: awesome work you've done working out how all this C service stuff works, BTW.
Dec 23 06:56:26 <rwhitby> especially the subscription stuff
Dec 23 06:56:50 <PuffTheMagic> i know how to use the iterators too
Dec 23 06:57:03 <PuffTheMagic> for when you use LSSubscriptionProcess
Dec 23 06:57:07 <PuffTheMagic> or what ever that method is
Dec 23 07:03:54 <rwhitby> was there any other doco you found other than the header files, or was it all working it out from no information/
Dec 23 07:03:56 <rwhitby> ?
Dec 23 07:05:09 <jacques2> howdy
Dec 23 07:05:49 <PuffTheMagic> rwhitby: just playing with the code
Dec 23 07:05:54 <PuffTheMagic> no other docs
Dec 23 07:06:19 <PuffTheMagic> oh well
Dec 23 07:06:28 <PuffTheMagic> libpurpleadapter did give some cluses
Dec 23 07:06:29 <PuffTheMagic> clues
Dec 23 07:11:34 <tmzt> no java needed?
Dec 23 07:11:46 <rwhitby> tmzt: correct
Dec 23 07:12:17 <rwhitby> which (as PuffTheMagic correctly continually reminds me) makes it much easier to start and stop services without impacting the rest of the phone
Dec 23 07:12:26 <PuffTheMagic> i might make python bindings for it ;)
Dec 23 07:12:40 <PuffTheMagic> since i have to learn python
Dec 23 07:13:14 <tmzt> great, java slowdown seems to be one of the issues with pre
Dec 23 07:13:29 <tmzt> walling that /proc/cpuinfo another
Dec 23 07:13:34 <tmzt> who does that?
Dec 23 07:13:48 <rwhitby> PuffTheMagic: please do - I put mjson and dbus-python into optware recently, cause I had the same idea of python bindings to the service API for such things as python scriptable profile triggers
Dec 23 07:14:14 <tmzt> look at ologicd very cool tricks
Dec 23 07:14:15 <PuffTheMagic> did u put liblunaservice in it?
Dec 23 07:14:22 <rwhitby> PuffTheMagic: yep
Dec 23 07:14:28 <tmzt> for that kind of situation aware thing
Dec 23 07:15:00 <rwhitby> tmzt: the lua stuff from openmoko?
Dec 23 07:15:21 <rwhitby> tmzt: that's exactly what I'm aiming at
Dec 23 07:15:45 <rwhitby> hmm - I wonder if a port of that might be more appropriate ...
Dec 23 07:15:47 <tmzt> yes I think it's lua but the scripts themselves are another language
Dec 23 07:16:13 <tmzt> talk to the developer
Dec 23 07:16:21 <tmzt> forget who it is right now though
Dec 23 07:16:24 * jacques2 really wants to learn lua
Dec 23 07:16:41 <rwhitby> tmzt: this one, right: http://build.shr-project.org/fsoshrudcon/FSOSHRUDCON09-Ologicd-Didier-Ptitjes.pdf
Dec 23 07:17:01 <tmzt> yes
Dec 23 07:17:02 <jacques2> but now I'm thinking about ecmascript
Dec 23 07:17:03 <tmzt> didjier
Dec 23 07:17:20 <tmzt> uh, ptitjes
Dec 23 07:17:30 <tmzt> or that, simple and easy towrite
Dec 23 07:17:42 <tmzt> not easy to parse as you need some overhead
Dec 23 07:17:48 <tmzt> unless you go with gjs
Dec 23 07:18:19 <tmzt> hmm, I've played with that to write a gnome3 phone ui but it will be too slow for my device (not cortex)
Dec 23 07:19:25 <rwhitby> tmzt: ah, I was actually thinking of org.freesmartphone.events: http://www.freesmartphone.org/index.php/Architecture#Events
Dec 23 07:19:50 <rwhitby> http://git.freesmartphone.org/?p=specs.git;a=blob_plain;f=html/org.freesmartphone.Events.html
Dec 23 07:19:59 <tmzt> then just use upstart?
Dec 23 07:20:19 <tmzt> it has events already
Dec 23 07:20:22 <rwhitby> it uses yaml
Dec 23 07:20:35 <tmzt> yeah, it's for audio and things
Dec 23 07:20:44 <tmzt> not advanced enough for ptijles stuff
Dec 23 07:20:53 <tmzt> I think he's replacing it
Dec 23 07:21:10 <rwhitby> http://www.freesmartphone.org/index.php/Brainstorm/fsoeventsd
Dec 23 07:21:11 <tmzt> I need a standalone version for my project
Dec 23 07:39:02 * rwhitby bbl
Dec 23 09:32:21 <Guest3033> hello?
Dec 23 09:33:24 <Guest3033> I'm a error wit preware on launch I get onfeed error
Dec 23 09:33:34 <Guest3033> having a error*
Dec 23 10:21:55 <rwhitby-pre> anyone know how to fix a touchstone that keeps cycling through charging on and charging off?
Dec 23 10:32:48 <joshua> pray to the touchstone gods
Dec 23 16:36:46 <oc80w> what's good
Dec 23 16:42:28 <DarkSpecter> rwhitby: ping
Dec 23 17:54:54 <linuxjacques> infobot: run full diagnostic sweep
Dec 23 18:03:07 <dsevil> infobot: make me a sandwich
Dec 23 18:03:08 <infobot> make: *** No rule to make target `me a sandwich'.  Stop.
Dec 23 18:04:07 <dsevil> haha
Dec 23 18:05:14 <ird> infobot: make me the perfect woman
Dec 23 18:05:16 <infobot> make: *** No rule to make target `me the perfect woman'.  Stop.
Dec 23 18:05:30 <ird> That sounds wrong...
Dec 23 19:01:50 <linuxjacques> the uptime on my Pre is still bugging me. it's just wrong.
Dec 23 19:04:07 <linuxjacques> has there been any work on adding new audio codecs? I would like ogg-vorbis and FLAC. I'm sure lots of other people would too.
Dec 23 19:10:26 <linuxjacques> infobot: tell rwhitby I'm onto him.
Dec 23 19:12:32 <DarkSpecter> http://www.webos-internals.org/wiki/Gstreamer
Dec 23 19:14:57 <linuxjacques> thanks DarkSpecter
Dec 23 20:13:15 <jacques> are the arm6 and arm7 binaries supposed to be different? I can't find where they are built with different compiler options, and readelf does not indicate any differences
Dec 23 20:14:59 <Rick_work> geist said they were different.
Dec 23 20:15:08 <en0x> one is for pixi other for pre
Dec 23 20:17:41 <roxfan> you probably mean armv6 and armv7
Dec 23 20:17:55 <roxfan> (the v is vvvery important)
Dec 23 20:17:57 <jacques> I'm looking at the output of the preware build system. I cannot detect any difference between the arm6 and arm7 binaries. nor can I detect any reason they should be from looking at the Makefiles
Dec 23 20:18:15 <jacques> roxfan: yes, you are correct. sorry :-)
Dec 23 20:18:40 <roxfan> armv7 includes some extra instructions which are probably not used in common programs
Dec 23 20:19:00 <roxfan> for example NEON simd instructions
Dec 23 20:19:22 <jacques> and am I correctly interpreting that the binaries are using the Thumb instruction set?
Dec 23 20:19:55 <roxfan> they might
Dec 23 20:20:03 <jacques> roxfan: so, for most cases (binaries not using the special armv7 instructions) the same binary would run on Pre and Pixi ?
Dec 23 20:20:11 <roxfan> both v6 and v7 support thumb and arm modes
Dec 23 20:20:30 <roxfan> not sure what's pixi but probably yes
Dec 23 20:20:40 <roxfan> lemme check the changes again...
Dec 23 20:20:45 <jacques> my understanding is Thumb gets you smaller binaries but trades off execution speed
Dec 23 20:21:08 <jacques> anyone here have a Pixi ?
Dec 23 20:23:44 <jacques> if I am correctly interpreting readelf output, the Pre distro and Preware are built only with v4 (thumb) instructions: Tag_CPU_arch: v4T
Dec 23 20:23:52 <roxfan> hmm it seems there are a couple of thumb instructions in v7 which are not supported in v6
Dec 23 20:24:44 <roxfan> v4t would be weird... v6 has a lot of useful instructions which reduce code size
Dec 23 20:25:02 <jacques> does the Pixi proc have a VFP (hardware floating point) unit?
Dec 23 20:25:45 <jacques> Pre does, and I *think* it's being used: Tag_ABI_FP_exceptions: Needed
Dec 23 20:25:55 <roxfan> ARM1136EJ-S
Dec 23 20:26:03 <roxfan> hmm, no VFP it seems
Dec 23 20:28:11 <jacques> maybe I need to disassemble some binaries to see which floating point instructions are used
Dec 23 20:28:38 <roxfan> it's possible that pixi uses softfloat
Dec 23 20:29:53 <jacques> roxfan: if that's true, and my understanding of the above is correct, I would expect Preware binaries using floating point instructions to fail on the Pixi
Dec 23 20:30:19 <chrisa> msm7x27 has a floating point unit
Dec 23 20:30:32 <jacques> aha.
Dec 23 20:30:42 <roxfan> http://pdadb.net/index.php?m=cpu&id=a7627&c=qualcomm_msm7627 this is wrong then?
Dec 23 20:30:51 <roxfan> or the fpu is not vfp?
Dec 23 20:31:19 <chrisa> I'm not sure about vfp, but I know we switched 7x25->7x27 midway partly due to wanting a FPU
Dec 23 20:31:36 <chrisa> webkit sure does love floats
Dec 23 20:31:52 <roxfan> jacques: i _think_ that it's possible to compile binary with softfloat so that it will use vfp if it's present
Dec 23 20:32:31 <jacques> roxfan: maybe, but I have never encountered such a thing
Dec 23 20:33:26 <jacques> the ways I understand in order of low performance to high are 1) kernel FP emulation 2) softfloat 3) real FP unit
Dec 23 20:33:37 <Rick_work> since javascript is all floats all the time,  i would think that would be important to a web-browser based os.
Dec 23 20:33:49 <jacques> in 1) the kernel traps FP instructions and emulates them - very slow
Dec 23 20:34:16 <jacques> in 2) the toolchain replaces FP instructions with integer code
Dec 23 20:34:39 <jacques> in 3) the hardware FPU handles the FP instructions
Dec 23 20:35:32 <roxfan> i think softfloat just uses libc functions for all fp manipulations
Dec 23 20:35:48 <roxfan> and so libc can choose to use hw or sw
Dec 23 20:36:08 <jacques> but what about FP instructions which are not the result of a libc call
Dec 23 20:36:20 <jacques> those would not be handles
Dec 23 20:36:24 <jacques> handled
Dec 23 20:36:30 <roxfan> they're not supposed to be in binary if it's compiled with softfloat...
Dec 23 20:37:47 <jacques> my point is softfloat is a pervasive toolchain feature. all FP instructions everywhere (libc and apps) are replaced with integer code
Dec 23 20:38:22 <jacques> you cannot generally swap a hard-float and soft-float libc without rebuilding everything
Dec 23 20:38:41 <roxfan> hmm, http://www.codesourcery.com/sgpp/lite/arm/portal/kbentry27
Dec 23 20:39:14 <geist> movw/movt is the biggest thing you'll see different
Dec 23 20:39:22 <geist> it's an armv7 thing and the compiler uses the hell out of it
Dec 23 20:39:56 <roxfan> maybe i should read the eabi specs...
Dec 23 20:40:04 <roxfan> (again)
Dec 23 20:41:01 <geist> but the use of it does not change anything about the abi, so it's not marked as such in the ELF bits
Dec 23 20:42:37 <jacques> roxfan: I read that page you linked, and basically is says if you build binaries that way, they will not use the hardware floating point unit, but they can interoperate with code which does
Dec 23 20:42:40 <jacques> not the other way around
Dec 23 20:43:06 <jacques> so you can get compatibility by giving up speed
Dec 23 20:43:53 <jacques> geist: yes that's bringing back memories. I recall disassembling code on the Nokia N800 (OMAP2) to prove that hardware float was / was not being used
Dec 23 20:44:33 <jacques> when the N800 came out, it was not using the hardware FPU, nokia used the same toolchain as their previous device, which did not have one
Dec 23 20:45:21 <jacques> we got significant speed increases for some workloads by rebuilding the libc and specific apps
Dec 23 20:45:42 <jacques> I dunno if they ever changed it officially
Dec 23 20:46:53 <oil> oh, its the 23rd!
Dec 23 20:47:02 <oil> its festivus!
Dec 23 20:47:14 <jacques> geist: when I look at a readelf dump and see "Tag_CPU_arch: v4T" am I correct that means only armv4 thumb instructions are used?
Dec 23 20:49:03 <roxfan> looks like it
Dec 23 20:49:09 <roxfan> what was gcc commandline?
Dec 23 20:49:50 <jacques> roxfan: I re-read my statement above and I think it's wrong; the CS page talks about emitting code which uses VFP instructions but can interoperate with code which doesn't
Dec 23 20:50:48 <jacques> which means it's not a compatibility thing, it's for compiling certain sensitive code for extra speed when you know it will be used on a device with a VFP
Dec 23 20:51:08 <jeremy9174> hi all
Dec 23 20:51:09 <jacques> but some of the libs the code will use were not built with VFP instructions
Dec 23 20:51:31 <kesne> Been a while.
Dec 23 20:51:32 <roxfan> so softloat then just means that fp parameters are passed in GP registers
Dec 23 20:51:36 <kesne> I havent been in here in forever
Dec 23 20:51:48 <jeremy9174> question, does the open vpn work currently? and if so where can i find some instructions?
Dec 23 20:53:00 <bpadalino|work> what are you doing with the vfp ?
Dec 23 20:53:47 <jeremy9174> anybody know?
Dec 23 20:53:54 <bpadalino|work> i know nothing about open vpn
Dec 23 20:54:13 <jeremy9174> or any vpn i guess
Dec 23 20:55:04 <bpadalino|work> unfortunately i don't do anything vpn related on my pre .. sorry :(
Dec 23 20:55:43 <jacques> roxfan: I'm looking at the Preware build and there are no tuning options passed to the cross-compiler. and the options are identical for the armv6 and armv7 versions of the packages. and the CS xcompiler specs do no tuning AFAICT
Dec 23 20:56:07 <bpadalino|work> oh, you want more speed
Dec 23 20:56:10 <jacques> heck v4t might just be the default output
Dec 23 20:57:01 <jacques> disassembly will answer this. well that along with an understanding of the intruction sets :-\
Dec 23 20:57:38 <bpadalino|work> jacques, the overall goal being optimized compile flags for the pre and/or pixi targets ?
Dec 23 20:58:21 <jacques> bpadalino|work: yes, optimal for each device
Dec 23 20:58:45 <bpadalino|work> neat
Dec 23 20:58:48 <jacques> bpadalino|work: but also needs to interoperate with the shipped libs :-) which are showing v4t on the Pre
Dec 23 20:59:07 <bpadalino|work> as geist mentions, does that change the abi ?
Dec 23 20:59:25 <jacques> roxfan: -mfloat-abi=softfp means passing of args in GP regs, soft-float is a superset of that
Dec 23 20:59:56 <jacques> bpadalino|work: at this point I do not know.
Dec 23 21:00:06 <bpadalino|work> http://wiki.debian.org/ArmEabiPort - did you read that page or are you ware of everything it says ?
Dec 23 21:00:51 <jacques> bpadalino|work: I've read that page many times in the last few years, not recently though. I will re-read it
Dec 23 21:02:03 <bpadalino|work> do the elfs provided use the eabi or traditional gcc abi ?
Dec 23 21:02:05 <jacques> bpadalino|work: thanks for pointing it out again, I am definitely seeing some relevant things
Dec 23 21:02:27 <jacques> bpadalino|work: my interpretation of readelf is EABI
Dec 23 21:02:43 <bpadalino|work> oh, very nice
Dec 23 21:03:24 <jacques> http://pastebin.com/m1543de6b
Dec 23 21:03:41 <jacques> Flags:                             0x5000002, has entry point, Version5 EABI
Dec 23 21:05:30 <bpadalino|work> interesting .. so does this mean it is using the v4t instruction set for the processor, or just that it's lowest common denominator is v4t compatibility for the library ?
Dec 23 21:05:59 <jacques> bpadalino|work: good question
Dec 23 21:07:02 <bpadalino|work> i can imagine it would be bad if it's using the v4t instruction set ..
Dec 23 21:07:36 <jacques> wow, check this out: http://pastebin.com/m5ff62c05
Dec 23 21:08:14 <bpadalino|work> excellent - so busybox is built for armv7
Dec 23 21:08:27 <bpadalino|work> that' good
Dec 23 21:08:30 <bpadalino|work> that's rather
Dec 23 21:08:35 <jacques> that's the readelf of busybox: Tag_CPU_name: "CORTEX-A8", Tag_CPU_arch: v7, Tag_VFP_arch: VFPv3
Dec 23 21:08:55 <jacques> I want the compiler flags which generated *that* binary :-D
Dec 23 21:11:24 <bpadalino|work> you're not able to make a binary which has that ?
Dec 23 21:11:34 <bpadalino|work> when your machine type is armv7 it doesn't do it, huh ?
Dec 23 21:12:18 <bpadalino|work> i am full of questions :(
Dec 23 21:14:47 <jacques> bpadalino|work: I haven't tried yet :-)
Dec 23 21:14:50 <bpadalino|work> :)
Dec 23 21:15:00 <bpadalino|work> you should feel empowered and go for it
Dec 23 21:15:48 <jacques> oh don't worry. I will.
Dec 23 21:16:18 <bpadalino|work> *phew* i was a little worried for a little bit
Dec 23 21:16:23 <jacques> lol
Dec 23 21:24:48 <dBsooner> any of you use flickr?
Dec 23 21:33:43 <jacques> bpadalino|work: adding "-mcpu=cortex-a8" got me "Tag_CPU_name: "CORTEX-A8"" and "Tag_CPU_arch: v7" but not (yet) "Tag_VFP_arch: VFPv3"
Dec 23 21:41:08 <jacques> though I do get "Tag_ABI_HardFP_use: SP and DP"
Dec 23 21:41:26 <jacques> I'm assuming that's single and double precision
Dec 23 21:42:30 <bpadalino|work> sounds like a good assumption
Dec 23 21:43:05 <bpadalino|work> what about -mfloat .. what are the valid options ?
Dec 23 21:44:20 <bpadalino|work> oh, i see .. so that probably sets the -mhard-float
Dec 23 21:44:22 <jacques> I gave it "-mcpu=cortex-a8 -mfpu=vfp3 -mfloat-abi=softfp"
Dec 23 21:44:24 <bpadalino|work> since you get hardfp
Dec 23 21:44:42 <jacques> uses hardware floating point unit, but software floating point ABI
Dec 23 21:44:58 <bpadalino|work> ah
Dec 23 21:45:00 <jauderho_>  -mnative should work, no?
Dec 23 21:45:07 <jacques> so it can interoperate wich code which uses the same ABI and software floating point
Dec 23 21:45:47 <jacques> jauderho_: I'm cross-compiling...
Dec 23 21:45:55 <jauderho_> ah
Dec 23 21:47:25 <jauderho_> I wonder if anyone is interested in building a 1password app for the pre...
Dec 23 21:53:24 <jacques> hmm, maybe it should be "-mfpu=neon"
Dec 23 21:54:01 * jacques feels the urge to do some benchmarking.
Dec 23 21:56:37 <jacques>  /usr/lib/jvm/java-1.5-palm/bin/java is also built with Tag_CPU_name: "CORTEX-A8"
Dec 23 21:59:32 <geist>   
Dec 23 22:00:08 <bpadalino|work> sup geist
Dec 23 22:00:27 <geist> lets see
Dec 23 22:00:48 <geist> we build with -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=softfp
Dec 23 22:01:03 <bpadalino|work> sounds like jacques has his answer
Dec 23 22:01:16 <geist> and
Dec 23 22:01:23 <geist> -mcpu=arm1136jf-s -mfpu=vfp -mfloat-abi=softfp
Dec 23 22:01:37 <jacques> geist: thank you!
Dec 23 22:02:08 <geist> the latter is a complete subset of the former, so it would run fine on a cortex
Dec 23 22:02:54 <bpadalino|work> geist, any idea what type of impact compiling libc and other supporting libs for a8 would do performance wise versus their current armv4t ?
Dec 23 22:09:22 <tmzt> why build for arm11 though?
Dec 23 22:09:46 <tmzt> trying to support pixi as well? but is has no neon right?
Dec 23 22:10:21 <bpadalino|work> right, so you can either support the whole set with one build, or do 2 specific builds
Dec 23 22:11:13 <jacques> we are already doing two (three including x86) builds
Dec 23 22:11:28 <jacques> but the two arm ones are identical (currently)
Dec 23 22:11:37 * jacques does "arm-none-linux-gnueabi-readelf -r java | grep aeabi"
Dec 23 22:13:56 <jacques> so the java binary calls many (31) functions in libm
Dec 23 22:14:55 <jacques>  http://pastebin.com/m60a168ac for the curious
Dec 23 22:32:09 <bpadalino|work> oh, libm is armv4t ?
Dec 23 22:34:16 <jacques> bpadalino|work: yes, libm looks identical to libc
Dec 23 22:36:11 <DarkSpecter> ... -mcpu=brain -mfpu=left_hemispere -mtune=common_sense
Dec 23 22:36:47 * DarkSpecter needs sleep it seems
Dec 23 22:39:10 <bpadalino|work> jacques, interesting - can you do the readelf on LunaSysMgr to see how that is compiled and how many calls to eabi stuff it has ?
Dec 23 22:42:22 <jacques> bpadalino|work: http://pastebin.com/m43eacbc1  this is the first time I have seen "Tag_NEON_arch: NEONv1"
Dec 23 22:42:51 <jacques> maybe those tags only show up if the binary actually uses those instructions
Dec 23 22:43:26 <bpadalino|work> quite possibly
Dec 23 22:43:38 <jacques> so if a binary was built with -mfpu=neon but no neon instrs were generated, the tag would not show up
Dec 23 22:44:21 <bpadalino|work> sounds reasonable
Dec 23 22:45:14 <jacques> my grep shows which functions are called, but (naturally) not how many times.
Dec 23 22:48:03 <bpadalino|work> right
Dec 23 22:48:06 <bpadalino|work> time for me to head home
Dec 23 22:48:08 <bpadalino|work> bbl
Dec 24 00:01:45 * jacques heads home.
Dec 24 00:57:57 <rwhitby> morning
**** ENDING LOGGING AT Thu Dec 24 02:59:57 2009