dsc_ | Wizzup: kindly add a repo for https://github.com/google/ruy | 06:46 |
---|---|---|
dsc_ | as -extras | 06:47 |
Wizzup | dsc_: will do | 11:15 |
Wizzup | freemangordon: I think qtwebengine build took down the jenkins web server hehe | 11:28 |
Wizzup | and I gave it 16GB of ram | 11:28 |
Wizzup | freemangordon: maybe we need to run it with one thread | 11:35 |
Wizzup | I guess it's all "ninja" now so just making it use one thread is hard | 11:52 |
Wizzup | even dpkg-buildpackage -b -uc -j1 doesn't work | 11:52 |
Wizzup | I'm moving the phoenix/leste .maemo.org back to an ssd | 11:57 |
Wizzup | downtime ~1hr | 11:57 |
freemangordon | Wizzup: can't you add swap? | 12:11 |
Wizzup | freemangordon: I'd rather have it not use 6-8 cores | 12:11 |
freemangordon | ENOMEM usually happens on linking, where the number of cores does not matter | 12:14 |
freemangordon | but ok | 12:14 |
Wizzup | freemangordon: hmm, I think that is not true, it can still be building other parts | 12:14 |
Wizzup | while linking some other component | 12:14 |
freemangordon | ok | 12:15 |
Wizzup | in any case --no-parallel doesn't seem to work for dh and dh_auto_build | 12:15 |
Wizzup | neither does -j1 | 12:15 |
Wizzup | but there's more to do, somehow the CI thinks that qtwebengine need not be built for arm | 12:15 |
Wizzup | I'll double check if I didn't forget to push my potential fix | 12:16 |
freemangordon | DEB_BUILD_OPTIONS="parallel=1"? | 12:16 |
Wizzup | nope, think I pushed myt fix | 12:16 |
Wizzup | https://github.com/maemo-leste-upstream-forks/qtwebengine/commit/668f864182b70469bbb819cade63207d9b44a438 | 12:17 |
Wizzup | this didn't help it seems | 12:17 |
Wizzup | also please check if you agree with version format in debian/changelog | 12:17 |
Wizzup | freemangordon: ^^ | 12:26 |
norayr | yesterday d4 again started heating up, and now it was kcompacd which was taking 99% of cpu. no programs were running, only empty hildon. | 12:39 |
norayr | but i believe such things were common even before chimaera. | 12:40 |
Wizzup | always check kernel messages when this happens | 12:41 |
Wizzup | dmesg | 12:41 |
freemangordon | Wizzup: I think you should have left it 5.15.2+dfsg-3 | 13:15 |
freemangordon | autobuilder will add +Nm7 | 13:16 |
Wizzup | freemangordon: does that make it 'newer' or 'higher'? | 13:17 |
Wizzup | if that is the case, then that's fine, but we should still make dfsg-4 in any case imo | 13:17 |
freemangordon | it makes it 'higher' | 13:17 |
freemangordon | no, it is debian that will eventually push dfsg-4 | 13:17 |
freemangordon | so, we either keep 5.15.2+dfsg-3 or make it 5.15.2+dfsg-3-maemo1 | 13:18 |
freemangordon | sorry, 5.15.2+dfsg-3-leste1 | 13:18 |
freemangordon | 5.15.2+maemo-1 is not really true, as we base on top of debian 5.15.2+dfsg-3 | 13:19 |
freemangordon | not that important though | 13:19 |
freemangordon | ah, it is already built | 13:20 |
freemangordon | leave it as it is | 13:20 |
Wizzup | freemangordon: nothing is built | 13:21 |
freemangordon | oh "ABORTED" | 13:21 |
freemangordon | :) | 13:21 |
Wizzup | yes, it refused to build anything for armhf and arm64 | 13:21 |
Wizzup | still refuses | 13:21 |
Wizzup | I'm waiting for it to build on my vm first, test if it works at all | 13:21 |
freemangordon | ok | 13:21 |
freemangordon | wtym 'refuses'? | 13:22 |
freemangordon | UI is still down? | 13:22 |
Wizzup | freemangordon: yes, another 30 mins or so | 13:25 |
freemangordon | ok | 13:25 |
Wizzup | freemangordon: well the jenkins debian glue decides there is nothing to build | 13:25 |
Wizzup | something based on grep Architecture:.*any on debian/control | 13:25 |
Wizzup | hence my commit above, but it didn't seem to have worked | 13:25 |
Wizzup | freemangordon: it's back | 13:50 |
Wizzup | freemangordon: https://phoenix.maemo.org/job/qtwebengine-binaries/architecture=armhf,label=armhf/2/consoleText | 13:51 |
freemangordon | is it possible that github was not 'ready'? | 13:52 |
freemangordon | happened once | 13:52 |
freemangordon | well, at least once | 13:53 |
freemangordon | or maybe re-tag | 13:53 |
sicelo | Wizzup: looks like n900 modem no longer causes insta-oops on v6.2-rc2 :-) | 13:54 |
sicelo | but it's not possible to online it yet, and there are still dma errors in dmesg, https://paste.debian.net/1266414/ | 13:57 |
Wizzup | sicelo: ah, great news, that will help narrow it down | 14:14 |
Wizzup | freemangordon: I tried twice, I'll try to re-tag as well | 14:14 |
Wizzup | freemangordon: still failing https://phoenix.maemo.org/job/qtwebengine-binaries/architecture=armhf,label=armhf/3/console | 14:30 |
Wizzup | freemangordon: as far as I can tell, looking at /usr/bin/build-and-provide-package on phoenix, there is a mystery | 14:36 |
Wizzup | it should find 'any' in the control file | 14:36 |
Wizzup | but doesn't | 14:36 |
Wizzup | + grep -q '^Architecture: all' qtwebengine-opensource-src-5.15.2+dfsg-3/debian/control | 14:37 |
Wizzup | + grep -q '^Architecture: .*any' qtwebengine-opensource-src-5.15.2+dfsg-3/debian/control | 14:37 |
Wizzup | + SKIP_ARCH_BUILD=true | 14:37 |
Wizzup | this clearly should be 0 and 0 as exit code, but then it sets this var, when it should not | 14:37 |
Wizzup | freemangordon: so apparently the control file is not what we expect https://phoenix.maemo.org/job/qtwebengine-binaries/architecture=armhf,label=armhf/4/console | 14:53 |
Wizzup | let me tag HEAD instead | 14:53 |
Wizzup | ok | 15:03 |
Wizzup | it's going now | 15:03 |
Wizzup | I suppose we could extend htis: | 15:03 |
Wizzup | if grep -q '^Architecture: .*any' "$control_file" ; then | 15:03 |
Wizzup | to also grep for ${architecture:-} | 15:03 |
Wizzup | it looks like pipewire can already do ofono + bt btw, assuming the sound system/modem works | 17:10 |
freemangordon | Wizzup: yeah, something github | 17:53 |
Wizzup | freemangordon: no | 17:54 |
Wizzup | freemangordon: this is something with debian/source/format I think | 17:54 |
Wizzup | I *changed* to to a unique tag | 17:55 |
freemangordon | ah | 17:55 |
Wizzup | and it still didn't take any changes to debian/control after the tag | 17:55 |
Wizzup | which is normally does | 17:55 |
Wizzup | which it* | 17:55 |
freemangordon | ok | 17:55 |
Wizzup | in any case | 17:55 |
Wizzup | freemangordon: ok, with local build qtwebengine now starts on vm | 19:17 |
freemangordon | you mean qtwebbrowser I guess | 20:30 |
freemangordon | Wizzup: BTW, do we plan to support non-NEON arm HW? | 20:30 |
Wizzup | freemangordon: not sure, why? | 20:32 |
Wizzup | freemangordon: yes I meant qtwebbrowser | 20:33 |
Wizzup | sicelo: any stuff going on on the ML regarding this oops btw? maybe it's more easy to get help now | 20:38 |
sicelo | Wizzup: yes, i will send it. do you think it's better to send it as a continuation of your previous thread, or a new one? | 20:56 |
Wizzup | sicelo: maybe a new one? | 20:57 |
sicelo | ok | 20:59 |
sicelo | https://marc.info/?l=linux-omap&m=163927204009431&w=2 "setting the dev->dma_mask in the driver is the right fix" ... | 21:04 |
sicelo | you meant adding 'dma_set_mask_and_coherent(&ssi->device, DMA_BIT_MASK(32));' in the `ssi_start_dma(struct hsi_msg *msg, int lch)` function? | 21:05 |
Wizzup | sicelo: who is this aimed at? me? | 21:08 |
Wizzup | sicelo: god I totally forgot about this mail | 21:09 |
Wizzup | I don't even remember if I tried this fix or not | 21:09 |
sicelo | :-D | 21:09 |
Wizzup | :( | 21:10 |
Wizzup | too much going on :) | 21:11 |
sicelo | that says you did ... anyway, i'll post and see what kernel people have to say. maybe the output from 'ssi-protocol' will ring a bell to sre, if he'll have time | 21:11 |
Wizzup | sicelo: reading my email I don't mention that I tested it | 21:12 |
sicelo | ah, ok. i misread. | 21:13 |
Wizzup | if it is helpful I can try to test tomorrow perhaps, although I'd prefer to do it based on our kernel tree somehow | 21:13 |
Wizzup | our being the shared mapphone/n900 one | 21:13 |
sicelo | sure | 21:15 |
sicelo | although it didn't make a noticeable difference in my case, unless i made a mistake | 21:16 |
sicelo | https://paste.debian.net/1266447/ hope this is correct? | 21:16 |
Wizzup | sicelo: I am not sure, I might not have tested it because I had to first figure out what the right mask was | 21:17 |
Wizzup | probably others know better than I do | 21:17 |
sicelo | that came from Pavel, but then it also looks like he was only making an example :-) | 21:18 |
Wizzup | might make sense to cc him, he's usually quick in replies | 21:21 |
sicelo | will do. so new email, and maybe i can refer/link to your previous one? | 21:22 |
Wizzup | yeah :) | 21:25 |
Wizzup | please cc me if you can | 21:25 |
sicelo | of course :-) | 21:25 |
sicelo | i usually cc the ML mailing list, but it seems slow to show mails | 21:25 |
sicelo | currently certificate broken too | 21:28 |
sicelo | Wizzup: yes you did try that dma_mask thingy ... i was wondering if i'm going crazy :p | 21:40 |
sicelo | https://marc.info/?l=linux-omap&m=163930832517778&w=2 | 21:40 |
Wizzup | However, looking at the omap3-n900.dts it looks to me that the ssi_pins | 21:40 |
Wizzup | definition suggests that it should use pio, rather than DMA. | 21:40 |
Wizzup | hmm | 21:41 |
sicelo | anyway, let's see what new info we will get | 21:41 |
Wizzup | mhm | 21:41 |
freemangordon | Wizzup: sicelo: yes, IIRC on n900 ssi shall use PIO | 22:02 |
Wizzup | sicelo: so it looks like it to me that the whole dma path should not be taken | 22:06 |
freemangordon | :nod: | 22:07 |
Wizzup | and something changed I guess in dma code that makes it not hard error and suddenly the n900 tries to take dma path | 22:07 |
freemangordon | Wizzup: why is modest no installed by default (at least on VM)? | 22:09 |
Wizzup | freemangordon: is it not? let's add it if it's not in hildon-meta | 22:11 |
Wizzup | I think it should be fwiw | 22:11 |
Wizzup | maybe it wasn't ported yet in your vm image | 22:12 |
freemangordon | could be | 22:12 |
freemangordon | maemo-leste-1.0-amd64-20221213 | 22:12 |
Wizzup | quite likely I think | 22:12 |
freemangordon | ok | 22:12 |
Wizzup | let me verify | 22:12 |
freemangordon | but yeat, my VM is upgraded | 22:13 |
freemangordon | *yet | 22:13 |
freemangordon | and no modest | 22:13 |
Wizzup | I don't see it is in hildon-meta | 22:13 |
Wizzup | I'll add it then | 22:13 |
freemangordon | mhm | 22:13 |
sicelo | what makes you guys say this should be PIO? not disagreeing - just trying to understand where it comes from (this stuff is higher than my pay grade, tbh) | 22:13 |
freemangordon | sicelo: my memories back from the days I was plying with that | 22:13 |
freemangordon | *playing | 22:13 |
freemangordon | I remember we had discussions back then | 22:14 |
freemangordon | argh | 22:14 |
freemangordon | no /seen command here | 22:14 |
freemangordon | anyway | 22:14 |
freemangordon | also, IIRC fremantle kernel does PIO | 22:14 |
Wizzup | sicelo: maybe you can force nerf the dma path just as a test | 22:14 |
freemangordon | so DMA was never tested | 22:14 |
sicelo | i see | 22:16 |
freemangordon | sicelo: BTW, for reference, please check if the last working upstream kernel does DMA or PIO | 22:18 |
Wizzup | yeah that was also where I was at at the time | 22:18 |
sicelo | yes, in fact i'm considering to compile 5.7, which i know worked fine | 22:19 |
sicelo | how will i tell if it's dma or pio btw? | 22:19 |
sicelo | printks in each of the functions? or there's easier/better way? | 22:19 |
freemangordon | sec | 22:19 |
freemangordon | hmm, not sure why | 22:22 |
freemangordon | maybe put printks in ssi_start_transfer | 22:23 |
freemangordon | ssi_start_dma/ssi_start_pio | 22:23 |
sicelo | mmm, seems gcc (or something else) not happy with kernel 5.7 | 22:42 |
sicelo | https://paste.debian.net/1266455/ | 22:43 |
Wizzup | sicelo: sorry no idea, how an older gcc you can CC= ? | 22:46 |
sicelo | i have ARCH=arm CROSS_COMPILE=arm-linux-gnueabihf- | 22:47 |
sicelo | gcc version 12.2.0 (Debian 12.2.0-13) | 22:47 |
sicelo | mmm, debian specific problem seemingly, https://groups.google.com/g/linux.debian.bugs.dist/c/2AWbk5Dxy54 | 22:49 |
Wizzup | I guess you can use this then: https://lore.kernel.org/linux-arm-kernel/20211018140735.3714254-1-arnd@kernel.org/ | 22:49 |
sicelo | yes, i'm applying that already :-) | 22:50 |
Wizzup | I'm finally getting somewhere with the news post btw | 22:50 |
Wizzup | probably tomorrow it will be done | 22:50 |
sicelo | compilation in progress, finally | 22:51 |
sicelo | great @post | 22:51 |
Wizzup | sicelo: sweet, let's see if it works, if it does use pio, a quick hack would be to change ssi_start_transfer to have it always call ssi_start_pio | 23:05 |
sicelo | mmm, bootloader (or something else) doesn't like my kernel. i get big green letters shortly after kernel log starts scrolling past: "Malfunction - device shutdown in 10 s" | 23:09 |
Wizzup | :/ | 23:09 |
sicelo | maybe low battery. even maemo won't boot now. i hope device is not bork | 23:11 |
sicelo | will see after charging for a bit | 23:13 |
Wizzup | btw, it would be kinda cute if we perhaps made some page that shows the latest built packages | 23:28 |
Wizzup | freemangordon: wow qtwebbrowser feels much more snappy with the 3d backend I think | 23:30 |
Wizzup | feels like it anyway | 23:30 |
sicelo | i really wonder what's up with that 'malfunction' ... anyway, building 6.2 again, with dma path commented out in ssi_start_transfer | 23:45 |
Wizzup | sicelo: let's see | 23:46 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!