asac | they do that kind of stuff elsewhere | 00:00 |
---|---|---|
asac | so seems thats ok | 00:00 |
asac | maybe forward that and add to package for now ;) | 00:00 |
asac | i did just put it to patches/ directory | 00:00 |
asac | err patches/ugly | 00:00 |
asac | guess you dont like it ;) | 00:00 |
fta | indeed | 00:00 |
fta | :) | 00:00 |
asac | fta: ok so the CFLAGS in rules should be added everywhere | 00:06 |
asac | according to kees its bad to have libs without fPIC | 00:06 |
asac | fta: and the patch i gave you upstreamed ;) | 00:08 |
asac | "quickly upstream" | 00:08 |
asac | hehe | 00:08 |
fta | pinged the guy, but he's not there atm | 00:09 |
asac | great | 00:09 |
asac | ok let me build on lucid without armv7 | 00:09 |
asac | lets hope that just works | 00:09 |
asac | ok kicked off chromium without armv7 | 00:12 |
asac | you said the SSE issue we had in karmic is probably fixed too now? guess that means this might work then | 00:13 |
* asac resets router .... port forward issues | 00:21 | |
asac_ | fta: for neon background see slides 7 and following https://wiki.ubuntu.com/Mobile/ARMv7AndThumb | 00:34 |
asac_ | arm folks suggest that neon optimization should at best be don in post-processing rather than at compile time | 00:34 |
asac_ | but no tool exists yet | 00:34 |
pwnguin | peephole neon? | 00:37 |
pwnguin | that sounds unlikely | 00:37 |
Sarvatt | theres a bug about ffmpeg/neon/stack protector problems if it helps any -- https://bugs.launchpad.net/ubuntu/karmic/+source/ffmpeg/+bug/383240 (comment 41 and newer) | 00:37 |
ubot4 | Launchpad bug 383240 in ffmpeg "Integrate and enable ARMv5TE/v6/VFP and NEON optimisations from ffmpeg trunk for armel" [Medium,Fix released] | 00:37 |
=== asac_ is now known as asac | ||
asac | thanks. guess i should check with lool or dmart on what to best to to optimize the ffmpeg for chromium | 00:47 |
=== powderluv_ is now known as powderluv | ||
=== ApOgEE__ is now known as ApOgEE | ||
=== ___bjf is now known as bjf-afk | ||
joshm | Anyone ever used a SmartQ 7? | 02:47 |
joshm | had one dropped in my lap today. A customer wants to use it in some kind of retail/bar scenario as an entertainment/targeted marketing doodad | 02:48 |
joshm | I've been playing with it for about an hour and it looks fun, just looking for any tips, pointers, gotchas ect | 02:49 |
persia | joshm: The SmartQ5 is an fun little bit of hardware. It is restricted to Ubuntu 9.04, due to hardware limitations (doesn't support the newer instruction sets used in newer releases). | 03:29 |
joshm | persia: cool thanks, is the 7 restricted like that as well? | 03:29 |
persia | I believe so, but I've only hearsay (I don't have one). | 03:30 |
persia | My understanding is that the only substantive difference between the 5 and the 7 is the screen size. | 03:30 |
joshm | thanks for the heads up though. I'll be doing more research before I do anything drastic to the little guy | 03:30 |
persia | One concern though is that you need a special kernel for those: they don't work with the standard 9.04 kernel. | 03:31 |
joshm | I've only been toying with it for a couple hours, but it seems like a nifty device | 03:31 |
persia | (But I think that's true for nearly everything except a couple developer boards today) | 03:31 |
joshm | persia: From my instructions I've decided I don't need to replace the OS at all | 03:31 |
persia | I wouldn't think so. I'd recommend doing an update against the latest jaunty-updates repo though. | 03:32 |
joshm | no DRM or anything so I was able to install openssh and get access from my desktop ( much easier to type ) | 03:32 |
persia | And then just install whatever apps you need to meet your use case. | 03:32 |
joshm | I do need to work on paring down the install though | 03:32 |
persia | Oh, almost forgot. Mono support is weak, so f-spot and tomboy might not be best choices for a 9.04-based solution (although I'm not sure how much you need that for retail/bar use cases) | 03:33 |
joshm | from what I gather ( I have very limited info atm ) my customer wants to use it as a kind of table trick/entertainment/ordering platform/targeted marketing thing | 03:33 |
persia | With a 5" screen! | 03:33 |
joshm | so I don't think an SD card will be wise | 03:33 |
persia | I'd recommend the 7 for such a scenario. I carry a 3" and a 4.5" all day every day, but I do bring those closer to my face. | 03:34 |
joshm | unless we epoxy the bastard in | 03:34 |
persia | Most of the at-table electronic ordering systems I use seem to be in the 7-9" range, at not very high DPI (100 or less). This seems more than sufficient for most entertainment/menu display, and avoids narrow touch points for drunken fingers. | 03:35 |
joshm | basicly this got dumped on me ( I don't mind! it's fun! ) because I just started at this company and I have linux experience | 03:35 |
persia | The Q5 is probably a good dev platform, and just tell the customer to order the Q7s when it comes to installation. Last I checked (about 9 months back), they were about the same price, and had nearly the same specs (including screen resolution) | 03:36 |
joshm | it's a small windows shop. The owner thought this would be fun for me ( he was right ) | 03:36 |
joshm | I have no idea what the customer has bought yet, I have a 7 sitting next to me that was dropped off this afternoon | 03:37 |
joshm | one of the requirements byt he customer was firefox. after a little research I've seen alot of posts about firefox being dog slow on arm | 03:38 |
persia | You oughtn't need that much storage. My Netwalker has 4G onboard, and I mostly use the microSD for convenience, rather than from need. The Q7 is a little tighter, but if you're running it in kiosk mode, you shouldn't need that many apps. | 03:38 |
joshm | is that correct? should I be looking at another browswer? The customer specificly requested firefox. I've emailed him asking why firefox is a must. But haven't heard back yet | 03:39 |
persia | Well, it's slow, but it's usable. I find firefox on the Netwalker to have significantly faster performance than the default Zaurus browser (just because the HW is faster). | 03:39 |
joshm | I saw a post from the ubuntu devs saying they would look at chromium for next release | 03:40 |
joshm | so firefox is useable? | 03:40 |
joshm | from the way the post read, firefox was a bear | 03:40 |
persia | At the recent UDS, we had a session to talk about it, and while I didn't attend, my memory of the results was that webkit-based browsers were considered preferable. That said, I don't think any of the webkit-based browsers in 9.04 are in good enough shape to use out of the box. | 03:40 |
persia | Firefox is definitely usable, but the performance is much slower than on a desktop. It depends on your expectations. | 03:41 |
joshm | --->> link https://wiki.ubuntu.com/Specs/ArmLightweightBrowser | 03:41 |
persia | Right. Thats the spec for the session I mentioned. | 03:41 |
joshm | great thanks | 03:42 |
joshm | I'm about done futzing with it tonight, Thanks for the info, you have been awesome. | 03:43 |
joshm | I'm wondering where my customer is getting them now | 03:43 |
joshm | I think he is getting them pretty cheap, and my wife has been wanting an e-book reader | 03:44 |
persia | No problem. Good luck with your project. One other note: 9.04 is only supported until around October 2010, so once you have a PoC, it might be worth investigating alternate hardware solutions for the final instalaltion (that could take advantage of a newer release) | 03:44 |
joshm | I know it isn't the liquid paper type stuff like kindle, but from what I gather it has a good uptime on battery | 03:44 |
persia | I've seen them advertised for export including shipping to Europe for ~ 140 euros from a few sources. | 03:45 |
joshm | she hasn't gotten off work yet tonight | 03:45 |
persia | I didn't end up ordering one, so I can't recommend any specific vendor (as none of the places I saw it advertised were familiar to me) | 03:45 |
joshm | I'm gonna toss a book on it and have her play with it | 03:46 |
persia | If it's not installed, you might want to try installing fbreader first. | 03:46 |
joshm | yeah I've already "rooted" it | 03:46 |
persia | It's a reasonably nice interface for reading (unless your books are PDF or HTML) | 03:46 |
joshm | installed openssh so I didn't have to deal with it's interface | 03:47 |
persia | Dont the SmartQ series come "rooted"? | 03:47 |
joshm | yeah it was superficial | 03:47 |
joshm | hence the "" | 03:47 |
persia | It's nice to see more consumer hardware coming out that way, so we can mess with it properly :) | 03:47 |
joshm | there was no obstical, though trying to add a user and give it sudo rights while the onscreen keyboard covered the shell was fun | 03:48 |
persia | There's no USB port? | 03:48 |
joshm | yeah there is, but I don't have a usb keyboard at my house | 03:48 |
persia | Ah, it all makes sense now :) | 03:49 |
joshm | all mine are PS2 | 03:49 |
joshm | actually I did find one after I was done, but it's almost as bad | 03:49 |
persia | Worse in some ways, because one has all the pain of not having one and then gets the fun of slapping one's forehead | 03:50 |
joshm | my keyboard my sister gave me, it's dish washer washable | 03:50 |
joshm | it's all plastic | 03:50 |
joshm | hard to type on | 03:51 |
joshm | the onscreen one was almost as good | 03:51 |
joshm | except with the usb one I could see what I was typing | 03:51 |
persia | Oh my. That's an interesting choice. | 03:52 |
joshm | I typed enough to add openssh-server, and add a user onscreen | 03:54 |
joshm | then I logged in with ssh and created a script to add me to sudoers, script named /tmp/a | 03:55 |
joshm | then ran that | 03:55 |
persia | And once you have an ssh connection, you can use the interface of your choice to hack it. | 03:55 |
joshm | wasn't so much bypassing security as bypassing inconveinence | 03:55 |
joshm | exactly | 03:55 |
persia | Just be careful with that: one issue I've seen a lot with using ssh to hack the little machines is that one forgets to make sure the interface works without it in the final solution. | 03:56 |
joshm | well right now I'm just playing | 03:56 |
joshm | per my boss this is low priority | 03:56 |
joshm | think it might be by bosses friend or something | 03:57 |
joshm | I don't know | 03:57 |
joshm | don't care really. free hardware to play with is payment enough for me :P | 03:57 |
persia | Yeah. Toys always win :) | 03:58 |
=== ogra_ is now known as ogra | ||
asac | lool: i saw you did some efforts on supporting neon for ffmpeg ... whats the idea? | 12:10 |
asac | like: how is that supposed to work only on hardare that supports neon etc. | 12:11 |
asac | fta: chromium seems not to open pdf files etc. is that known? | 12:16 |
asac | e.g. lack of mime integration with desktop et al | 12:16 |
ogra | hmm, db4.2 seems to need a swp fix | 12:29 |
* ogra gives back a bunch of packages that might be fixed with -mimplicit-it=thumb | 12:32 | |
armin76 | asac is now an arm expert! | 12:38 |
ogra | armin76, we dont leave him a choice ;) | 12:39 |
asac | haha | 12:39 |
asac | i know armin76 ... my personal super-expert | 12:39 |
asac | ;) | 12:39 |
asac | ogra: is implicit-it in toolchain now? | 12:40 |
ogra | since dec. 4th | 12:40 |
asac | odd. thoughti talked to doko after that | 12:40 |
asac | he said it was blocked on eglibc fixes | 12:40 |
ogra | i gave back everything that was failing before that date with assembler messages that didnt contain swp errors | 12:40 |
asac | good | 12:40 |
ogra | well, gcc at least sets the option | 12:41 |
asac | ogra: do you have buildd powers or did you use launchpad web ui? | 12:41 |
ogra | * Pass -mimplicit-it=thumb to as by default on ARM. LP: #488302. | 12:41 |
ogra | LP ui | 12:41 |
ogra | the pages are linked directly from the ftbfs list | 12:41 |
asac | we need a buildd admin in the team | 12:41 |
ogra | we have lamont :) | 12:41 |
asac | so we can prioritize armel builds | 12:41 |
asac | well | 12:41 |
lool | asac: It's done using hwcaps | 12:41 |
ogra | NCommander can priorize | 12:41 |
lool | asac: glibc is patched to look at /lib/neon, /lib/vfp, /lib/vfp/neon and the like | 12:42 |
ogra | asac, archive admins can do that iirc, no need for a buildd admin | 12:42 |
lool | asac: So we have two ffmpeg builds, one with neon enabled and one without; the neon enabled one goes into the neon-vfp dir | 12:42 |
asac | ogra: afaik buildd admin (like doko) ... i dont think there is more power attached to "buildd" admin | 12:42 |
asac | lool: ok. so the package just does two runs? | 12:43 |
lool | I think ffmpeg should be revised for thumb2 / armv7 though, the flags are hardcoded to the lowest supported ABI | 12:43 |
asac | lool: where are the dirs located it looks at? | 12:43 |
ogra | asac, there surely is, but i know that StevenK as well as NCommander can bump priorities for us | 12:43 |
lool | asac: glibc? | 12:43 |
asac | lool: i mean: where are they coded. ok | 12:43 |
asac | thx | 12:43 |
asac | ogra: ncommander is archive admin? | 12:43 |
ogra | no | 12:43 |
asac | good ;) | 12:43 |
ogra | but he is in the porters team | 12:44 |
lool | He is porter | 12:44 |
ogra | so he can bump packages | 12:44 |
asac | ok | 12:44 |
asac | both are not really there though ;) | 12:44 |
ogra | right, in that case ping in #ubuntu-devel | 12:44 |
ogra | usually takes not more than ten mins until someone helps | 12:45 |
ogra | hrm | 12:45 |
ogra | The following packages have unmet dependencies: | 12:45 |
ogra | libmetacity-private0: Conflicts: libmetacity0 but 1:2.28.0-0ubuntu1 is to be installed | 12:45 |
ogra | E: Broken packages | 12:45 |
ogra | still :/ | 12:45 |
ogra | root@babbage2:/# apt-cache rdepends compiz | 12:46 |
ogra | compiz | 12:46 |
ogra | Reverse Depends: | 12:46 |
ogra | ubuntu-netbook-remix | 12:46 |
ogra | oh, intresting | 12:46 |
asac | ogra: https://edge.launchpad.net/~launchpad-buildd-admins | 12:46 |
ogra | why is that seeded ? | 12:46 |
asac | thats the team ncommander is in | 12:46 |
ogra | ubuntu-porters i think | 12:46 |
asac | no | 12:46 |
asac | https://edge.launchpad.net/~mcasadevall/+participation | 12:46 |
asac | Build Daemon Maintainers | 12:47 |
ogra | ah | 12:47 |
asac | so yes. thats what i ment with buildd admins | 12:47 |
ogra | for sparc and ia64 i think | 12:47 |
asac | ok | 12:47 |
ogra | hmm, there is no rdep for libmetacity0 anymore | 12:47 |
lool | ogra: There is a workaround in the seed to avoid pulling compiz | 12:48 |
lool | It's due to the x-window-manager provides | 12:48 |
* ogra doesnt get why his build working | 12:48 | |
ogra | lool, ah | 12:48 |
ogra | *is not working (indeed) | 12:49 |
lool | we dropped it at tsome point because maximus got fixed | 12:50 |
lool | but apparently it's broken again | 12:50 |
ogra | right | 12:50 |
ogra | well, i dont really care atm | 12:51 |
ogra | my armel builds fail ... thats more fatal | 12:51 |
ogra | nothing in ubuntu-desktop should pull in libmetacity0 anymore ... but the image still moans | 12:52 |
cooloney_ | ogra: just interrupt for a while. we still don't have the patch drop from fsl? | 12:53 |
lool | StevenK borke maximus | 12:53 |
ogra | cooloney_, i'll ping the world as soon as they release something | 12:53 |
cooloney_ | ogra: thanks, i just ordered a sata cable and try to test my patch on the platform tomorrow | 12:54 |
ogra | cooloney_, they said there should be an untested unstable drop by end of this week | 12:54 |
ogra | The following packages have unmet dependencies: | 12:54 |
ogra | compiz-gnome: Depends: libmetacity0 (>= 1:2.25.8) but it is not installable | 12:54 |
ogra | E: Broken packages | 12:54 |
ogra | ARGH ! | 12:54 |
lool | Was maximus imported to Bzr yet? | 12:56 |
lool | I thikn StevenK has a spec on this | 12:56 |
ogra | no idea, i thought he dropped all UNR work | 12:57 |
asac | he is still doing this for the time being | 12:59 |
ogra | ah | 12:59 |
asac | not sure which spec he really puts work in though | 12:59 |
* ogra cries | 12:59 | |
ogra | i have no idea why compiz-gnome is still tried to install | 12:59 |
lool | I uploaded a fixed maximus | 13:01 |
ogra | sigh | 13:02 |
lool | ogra: Is still tried to install? | 13:02 |
lool | EPARSE | 13:02 |
ogra | lool, i dropped compipz from ubuntu-desktop | 13:02 |
lool | For armel only I guess? | 13:02 |
ogra | but yes, something still pulls it in | 13:03 |
ogra | yep, armel only | 13:03 |
lool | Did you seed metacity on armel? | 13:03 |
ogra | no, its in the default seed | 13:03 |
lool | The default seed? | 13:03 |
lool | What's that? | 13:03 |
ogra | ubuntu-desktop | 13:03 |
lool | Oh it's always seeded already ok | 13:03 |
ogra | both WMs are in there | 13:03 |
ogra | i just dropped compipz | 13:03 |
ogra | -p | 13:04 |
ogra | root@babbage2:/# apt-cache rdepends compiz-gnome | 13:04 |
ogra | compiz-gnome | 13:04 |
ogra | Reverse Depends: | 13:04 |
ogra | |gnome-session | 13:04 |
ogra | |compiz | 13:04 |
lool | Did you wait long enough that the tasks get updated? | 13:04 |
ogra | but gnome session has a "metacity | compiz-gnome | sawfish" | 13:04 |
ogra | so it shouldnt be pulled | 13:04 |
ogra | no idea, isnt the task updated with the seed commit ? | 13:05 |
ogra | which happens long before the -meta is promoted | 13:05 |
lool | Apparently compiz-gnome is still in the seed | 13:06 |
lool | ogra: no, the tasks are updated two publisher runs after the seed is pushed | 13:06 |
lool | *If* at least one package changes | 13:07 |
ogra | compiz-gnome was never in the seed | 13:07 |
ogra | only compiz | 13:07 |
lool | chdist -a armel apt-cache lucid-armel show compiz-gnome | grep Task | 13:07 |
lool | Task: ubuntu-desktop, edubuntu-desktop, ubuntu-netbook-remix | 13:07 |
lool | It's definitely seeded on armel; it might be because of some dep | 13:07 |
lool | compiz depends on compiz-gnome | 13:08 |
lool | compiz is still seeded on armel ATM | 13:08 |
ogra | well, rdepends shows gnome-session but as i said above ... we have metacity to fulfill that | 13:08 |
lool | Again, that might be due to some dep | 13:08 |
lool | Note that gnome-session is listed *first* in the seed | 13:09 |
lool | That said, metacity is listed first in gnome-session's deps | 13:09 |
shenki_ | asac: ping | 13:10 |
shenki_ | asac: im the person who emailed you a few hours back about chromium | 13:11 |
lool | ogra: it might be xinit pulling that | 13:12 |
ogra | lool, thats fine ... i was apparently missing the second publisher run | 13:12 |
ogra | colin said its gone now in germinate and will be gone after the next publisher | 13:12 |
shenki_ | asac: if you try the link again the PDF should download correctly. it's just slides from a talk I gave a few weeks ago, i'd like to write up my results before i publish it broadly | 13:14 |
=== shenki_ is now known as shenki | ||
asac | shenki: ok thanks. | 13:17 |
asac | and welcome ;) | 13:19 |
shenki | thanks | 13:20 |
ogra | everyone cross your fingers ! | 13:56 |
* ogra just fired off an armel live build | 13:56 | |
* JamieBennett crosses everything | 13:57 | |
ogra | heh, ppc and ia64 are not happy ... | 14:00 |
ogra | armel still runs though | 14:00 |
asac | me crosses fingers too | 14:47 |
ogra | another 30-45min to go until it should spit out the first image | 14:48 |
* asac waits for that | 14:49 | |
asac | though i cannot test while this chromium build is going ;) | 14:49 |
asac | maybe i could try "hibernate" ;) | 14:49 |
* ogra glares at the clutter FTBFS log | 14:51 | |
ogra | -g -O2 -g -Wall -O2 -march=armv5t | 14:51 |
ogra | err | 14:51 |
ogra | WTF | 14:52 |
ogra | CPU := $(shell dpkg-architecture -qDEB_HOST_ARCH_CPU) | 14:53 |
ogra | ifeq ($(CPU),arm) | 14:53 |
ogra | CFLAGS += -march=armv5t | 14:53 |
ogra | endif | 14:53 |
ogra | GAR ! | 14:53 |
ogra | how silly is that ! | 14:53 |
asac | ogra: in rules? | 14:58 |
ogra | yeah | 14:58 |
ogra | rules is about ten lines | 14:58 |
ogra | thats one half | 14:58 |
asac | isnt ifeq (armel,$(DEB_BUILD_ARCH)) better? | 14:59 |
asac | i mean for the test (not for the flag ;)) | 14:59 |
ogra | yes, that too | 15:00 |
ogra | the whole chunk of code needs to go anyway though | 15:00 |
* ogra will take care for it tomorrow after the freeze | 15:01 | |
asac | who packaged that? | 15:01 |
asac | clutter upstream? | 15:01 |
asac | or debian? | 15:01 |
ogra | debian i think | 15:01 |
* ogra checks | 15:01 | |
ogra | well, upstream are DDs | 15:02 |
ogra | Ross Burton <ross@debian.org> worked for openhand last i checked | 15:02 |
asac | tell him that march is toolchain business ;) | 15:02 |
asac | and CPU is bad ;) | 15:03 |
ogra | funnily i cant find anything that mentions why the arch was hardcoded | 15:03 |
ogra | * Use LDFLAGS/CFLAGS vars directly to avoid overriding them completely; bdep | 15:03 |
ogra | on cdbs >= 0.4.41 | 15:03 |
asac | armin76 said that debian made hacks to support armv4 or something | 15:03 |
ogra | thats all i find even mentioning CFLAGS | 15:04 |
asac | maybe it didnt work for clutter so they forced armv5 | 15:04 |
ogra | and thats a patch from lool | 15:04 |
ogra | who i cant imagine doing something like above | 15:04 |
asac | no git/bzr for packaging? | 15:04 |
asac | dont we have debian bzr imports already? | 15:04 |
asac | or still "just" ubuntu? | 15:04 |
ogra | we should have debian as well as upstream imports | 15:05 |
asac | right. so checking there when that line was modified might give a clue | 15:05 |
asac | when this was added at least | 15:05 |
asac | unless we didnt import at that point yet | 15:06 |
shenki | http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=478152 | 15:06 |
ubot4 | Debian bug 478152 in clutter "clutter: needs armv5 on arm/armel" [Normal,Fixed] | 15:06 |
ogra | ugh | 15:07 |
suihkulokki | sorry :p | 15:08 |
ogra | heh | 15:08 |
ogra | we could probably make that conditional or some such ... so it doesnt executed when building on ubuntu | 15:09 |
ogra | *doesnt get | 15:09 |
shenki | the code that the bug references is no longer there. there are two places that are ifdef __arm__, and they contain instructions that are common to all variants | 15:09 |
shenki | so you can drop the arch check, imo | 15:10 |
ogra | though is debian building for anything smaller than v5 ? | 15:10 |
ogra | i thought it defaults to v5 too now | 15:10 |
ogra | ah, even better | 15:10 |
lool | asac: the CPU test covers any ARM distro, e.g. arm, armel, and perhaps future stuff like armelhardfloat | 15:19 |
lool | I don't think it's necessarily better to use the deb arch instead of the gnu cpu | 15:19 |
lool | build is wrong though, should be host | 15:19 |
ogra | well, its apparently obsolete anyway | 15:20 |
lool | The current rules say CPU := $(shell dpkg-architecture -qDEB_HOST_ARCH_CPU) | 15:20 |
lool | So these are correct | 15:21 |
ogra | yes, thats what i pasted above | 15:21 |
lool | asac wrote DEB_BUILD_ARCH though | 15:21 |
asac | yeah at least host | 15:21 |
ogra | yes, why is host arch better here ? | 15:21 |
lool | For cross-compiling | 15:21 |
asac | x-compile | 15:21 |
asac | yes | 15:21 |
lool | host is where it will run | 15:21 |
ogra | ah | 15:22 |
ogra | the stuff i never do :P | 15:22 |
asac | ah ... at least didnt fail yet | 16:16 |
ogra | well, it usually takes 1.5h per image | 16:17 |
=== bjf-afk is now known as bjf | ||
armin76 | ogra: what CFLAGS you guys use? | 16:51 |
armin76 | ok, found it at https://wiki.edubuntu.org/ARM/Thumb2 | 16:53 |
asac | fta: fixed the skia build probs on arm wo armv7 | 19:17 |
fta | which pbs? | 19:17 |
asac | fta: it failed a) trying to build SSE and b) missing some opts files when SSE was excluded for arm | 19:18 |
fta | hm | 19:19 |
asac | one second | 19:19 |
asac | i cannot do a diff because i dont have the original source unpacked anymore | 19:19 |
asac | let me give you the full file ;) | 19:19 |
asac | http://pastebin.com/f482b739d | 19:19 |
asac | so what i added as an "arm" only section | 19:20 |
asac | well all-"arm" section | 19:20 |
asac | that exluces the SSE2.cpp and adds the other files that are also added for armv7 | 19:20 |
asac | actually we can remove the files from the armv7 block too | 19:20 |
asac | one second | 19:20 |
asac | fta: http://paste.ubuntu.com/338224/ | 19:21 |
asac | and remove those two blocks from the armv7 section | 19:22 |
asac | so http://paste.ubuntu.com/338225/ | 19:22 |
armin76 | asac broke launchpad | 19:24 |
asac | how? | 19:25 |
asac | still works for me ;) | 19:25 |
=== bjf is now known as bjf-afk | ||
asac | ok restarted build with WANT_TESTS=0 ;) | 19:27 |
asac | fta: ok so WANT_TEST=0 fails to biuld as there are differences (or maybe just because i aborted a half built test build before) | 19:49 |
asac | chromium-codecs-ffmpeg depends on chromium-browser (>= 4.0.203.0~); however: | 19:49 |
asac | thats bad | 19:49 |
asac | two sided depends ;) | 19:50 |
asac | one of those should be dropped | 19:50 |
asac | i guess this one | 19:50 |
fta | it was for a transition | 19:51 |
asac | hmm. yeah. but pleaes drop it ;) | 19:53 |
playya_ | is it now possible to compile my code for arm on lp? | 21:30 |
playya_ | or should i squeeze everything out of my n800? | 21:36 |
=== bjf-afk is now known as bjf | ||
asac | playya_: squeeze | 21:52 |
playya_ | :/ | 21:52 |
playya_ | i should use the 2 GB card for /var/cache | 21:53 |
playya_ | but i can upload it to my ppa? | 21:54 |
asac | playya_: ppas dont build for arm | 21:59 |
asac | and you can only upload sources there | 21:59 |
playya_ | ok | 21:59 |
playya_ | bye bye n800 | 21:59 |
asac | hehe | 21:59 |
playya_ | see you in the next decade | 21:59 |
playya_ | afair there's a bug report for it | 22:00 |
playya_ | and the arm machines are quite idle most of the time | 22:00 |
playya_ | i just want to build the fso stuff :( | 22:00 |
asac | playya_: they are usually busy | 22:01 |
asac | and break regularly atm | 22:02 |
asac | now is freeze time | 22:02 |
asac | thats why they are (luckily) idle | 22:02 |
playya_ | just saw the load on a info page | 22:02 |
asac | https://edge.launchpad.net/builders | 22:02 |
playya_ | maybe that was on the end of a release cycle | 22:02 |
playya_ | s/on/at/ | 22:03 |
fta | asac, still there? | 22:52 |
asac | fta: yes | 23:00 |
asac | bb in 5 min | 23:00 |
fta | asac, want me to bring the chromium/codec dev here? so it's faster and i don't act as a relay | 23:01 |
asac | fta: why not ;) depends when he can make it though | 23:03 |
asac | http://pastebin.com/f65476f12 | 23:04 |
asac | that was the patch ... the other was about arm + fPIC afair | 23:04 |
fta | hold on, asking | 23:05 |
asac | fta: when basetest is running are all the debs already produced? | 23:11 |
fta | nope, debs are created at the end | 23:11 |
asac | oh no :( | 23:11 |
asac | this thing from yesterday is still running | 23:11 |
fta | lol | 23:12 |
asac | thats not acceptable ;) | 23:12 |
fta | maybe we should skip the test on arm then | 23:12 |
asac | maybe i should really try qemu ;) | 23:12 |
fta | testS | 23:12 |
asac | i think so ... for now | 23:12 |
asac | ipc_tests now | 23:13 |
fta | awong, welcome! | 23:14 |
awong | hello! | 23:14 |
asac | hi | 23:15 |
fta | asac, could you please summarize the situation? | 23:15 |
asac | situation is that the -codecs package failed because of relocation issue on armel ... adding -fPIC fixed that | 23:15 |
asac | then that revealed that there seems to be some code built that isnt built on other archs in libavdecoder/aac.c | 23:16 |
asac | this: http://pastebin.com/f65476f12 makes it build | 23:16 |
asac | but i am not sure if its right ... though i saw other tests for that constant | 23:16 |
awong | cool. | 23:16 |
asac | so now it actually built here | 23:16 |
asac | awong: so wonder if you know why -fPIC is not added for armel (though it is for amd64) | 23:17 |
asac | or any reason not to add that on i386 too? | 23:17 |
asac | e.g. everywhere? | 23:17 |
awong | I actually do know the reason. :D | 23:17 |
asac | cool. can you fix that ;)? | 23:17 |
asac | i think -fPIC everywhere should be ok | 23:18 |
awong | For arm, it's an oversight. We've only recently trying to fix the arm build. The -fPIC and -DPIC will probably get added soon (fbarchard has some changes in flight so I don't want to get into the mix quite yet). | 23:18 |
asac | also please ccheck the patch above and commit that too if thats ok ;) ... so we get building packages | 23:18 |
asac | hmm | 23:18 |
awong | As for i386, -fPIC can't work. | 23:18 |
asac | cant? | 23:18 |
awong | nope. | 23:18 |
asac | can you elaborate? | 23:19 |
awong | Not unless you disable a whole bunch of the ffmpeg assembly. | 23:19 |
awong | If you look at the current debian packages, -fPIC is actually disabled on ia32 builds of ffmpeg because of this. | 23:19 |
awong | Basically, they need the extra register (ebx specifically) free to do some of their more optimized routines. | 23:19 |
asac | i saw it wasnt used | 23:19 |
asac | and thats why i thoght that adding everywhere is wrong as it felt intentional | 23:19 |
asac | awong: so when is his arm work going to land? | 23:20 |
asac | (fbarchard) | 23:20 |
awong | i'll walk over and ask him in a sec... | 23:20 |
awong | For your patch, if you could submit a crbug.com with the patch, that'd be the easiest way to get it committed. | 23:21 |
asac | i hoped someone else can do that ;) | 23:21 |
awong | http://dev.chromium.org/developers/contributing-code | 23:21 |
asac | felt complicated when i last looked | 23:21 |
asac | at least for drive-by contributors | 23:22 |
fta | aac should be built for all arches with the non-free flags | 23:26 |
awong | asac: the issue is that since it's your patch, we want to make sure the attribution is right. | 23:27 |
awong | As for the arm stuff, fbarchard says he'll probably have something in a day or two. | 23:29 |
asac | ok thanks. | 23:29 |
asac | i dont even know where the ffmpeg code i patched is in svn ;) | 23:30 |
awong | hah | 23:30 |
asac | its not in third_party/ffmpeg for sure | 23:31 |
asac | fta: where do you produce that ffmpeg-mt thing from? | 23:32 |
fta | the tarballs inside the tarball? | 23:32 |
asac | yes | 23:32 |
asac | ffmpeg-mt | 23:32 |
fta | i just get them from chromium | 23:32 |
asac | where? | 23:32 |
fta | untouched | 23:32 |
asac | i dont see them in svn | 23:32 |
asac | http://src.chromium.org/viewvc/chrome/trunk/src/third_party/ffmpeg/ | 23:32 |
fta | http://src.chromium.org/svn/trunk/deps/third_party/ffmpeg/ | 23:33 |
fta | yep, same | 23:33 |
asac | nothingthere is nothing in it | 23:33 |
asac | no libavdecoder/aac.c ;) | 23:33 |
asac | http://src.chromium.org/viewvc/chrome/trunk/deps/third_party/ffmpeg/ | 23:34 |
asac | its different | 23:34 |
asac | crazy ;) | 23:34 |
asac | so this whole patch is not possible ;) | 23:35 |
awong | http://src.chromium.org/svn/trunk/deps/third_party/ffmpeg/patched-ffmpeg-mt/libavcodec/aac.c | 23:35 |
asac | hmm | 23:35 |
asac | yeah | 23:35 |
asac | ok | 23:35 |
asac | i can do that i hope | 23:35 |
asac | ;) | 23:35 |
awong | so, I can do the uploading of the patch and landing | 23:35 |
awong | but I think you still need to follow a few of the steops in http://dev.chromium.org/developers/contributing-code about the Individual COntributor License Agreement, and getting your name into the Authors file. | 23:36 |
asac | just for this minimal patch. cant you shuffle it a bit and then submit ;) | 23:37 |
awong | hah. | 23:37 |
awong | unforutnately, I don't htink that's quite cool. Frank's got a few changes for ARM in the air currently and will likely have fixed this same issue since it doesn't compile w/o it. Let's let his stuff land first and then we can hopefully just side-step this. | 23:42 |
awong | it's silly, I know...but not being a lawyer, I don't know the consequences, blah blah. | 23:43 |
asac | i now did all the gcl stuff ;) | 23:46 |
asac | so you want me to abort that? | 23:46 |
awong | heh | 23:46 |
awong | up to you man. If you're most of the way there, I'm happy to review and land for you. | 23:47 |
asac | awong: right. wonder if its a problem that i only checked out the libav*/ directory ... i assume gcl gets it right from svn info? | 23:48 |
awong | btw, on your guys's build, are you doing the full gyp/make thing? If so...how are you building our ffmpeg for arm at all? I don't think we have a config.h checked in for arm. | 23:48 |
awong | asac: I think gcl can handle it. However, bigger issue is that you're going to want to submit a patch to the source tree. | 23:50 |
awong | and add it to http://src.chromium.org/svn/trunk/deps/third_party/ffmpeg/patches/to_upstream/ as patch 51. | 23:50 |
awong | The "patched-ffmpeg-mt-source" directory is autogenerated from the tarball and these patches. | 23:50 |
asac | right | 23:50 |
asac | thats what i planned to do | 23:50 |
awong | ah, ok | 23:50 |
asac | you say on the patch in patches/? | 23:51 |
asac | or both? | 23:51 |
asac | oh | 23:51 |
asac | only the patch in patches/? | 23:51 |
asac | that was the question ;) | 23:51 |
asac | sorry | 23:51 |
awong | ah. yes, that's all you need. | 23:51 |
awong | I'll regenerated the patches tree for you afterwards since Ihave everything already setup. | 23:52 |
asac | ok ... good | 23:52 |
asac | out for a few minutes getting some stuff | 23:54 |
awong | np. I'm here for another hour or so. Othewrise, feel free to e-mail me (ajwong@chromium.org) | 23:55 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!