| josch | erle: i already profiled it. Results above. The slow parts are hundreds of calls to modinfo, cp and mkdir | 00:02 |
|---|---|---|
| - Ar|stote|is (QUIT: Read error: Connection reset by peer) (~linx@149.210.0.13) | 00:06 | |
| erle | josch what tool do you use to profile shell scripts if i may ask? (i use annotate-output-relative and strace and ltrace). | 00:07 |
| + Ar|stote|is (~linx@149.210.0.52) | 00:12 | |
| - Ar|stote|is (QUIT: Ping timeout: 268 seconds) (~linx@149.210.0.52) | 00:22 | |
| + Ar|stote|is (~linx@149.210.0.216) | 00:26 | |
| - mjw (QUIT: Ping timeout: 260 seconds) (~mjw@gnu.wildebeest.org) | 01:35 | |
| - Ar|stote|is (QUIT: Quit: https://quassel-irc.org - Chat comfortably. Anywhere.) (~linx@149.210.0.216) | 01:47 | |
| + Ar|stote|is (~linx@149.210.0.216) | 01:48 | |
| - Ar|stote|is (QUIT: Ping timeout: 255 seconds) (~linx@149.210.0.216) | 01:54 | |
| + Ar|stote|is (~linx@149.210.3.168) | 01:58 | |
| - Ar|stote|is (QUIT: Ping timeout: 268 seconds) (~linx@149.210.3.168) | 02:19 | |
| + Ar|stote|is (~linx@149.210.1.251) | 02:23 | |
| - cobra (QUIT: Ping timeout: 260 seconds) (~cobra@user/Cobra) | 03:54 | |
| - yankcrime (QUIT: Quit: "") (~nick@gw.tetromino.io) | 04:17 | |
| + yankcrime (~nick@gw.tetromino.io) | 04:18 | |
| + chomwitt (~chomwitt@2a02:587:7a21:2700:1ac0:4dff:fedb:a3f1) | 05:16 | |
| - robin (QUIT: Ping timeout: 268 seconds) (~robin@user/terpri) | 07:17 | |
| + robin (~robin@user/terpri) | 08:06 | |
| - Ar|stote|is (QUIT: Ping timeout: 268 seconds) (~linx@149.210.1.251) | 08:22 | |
| + Ar|stote|is (~linx@149.210.26.5) | 08:42 | |
| - Ar|stote|is (QUIT: Ping timeout: 255 seconds) (~linx@149.210.26.5) | 08:49 | |
| + Ar|stote|is (~linx@149.210.24.250) | 08:53 | |
| ex-parrot | hello everyone | 09:14 |
| josch | _o/ | 09:33 |
| josch | erle: what is annotate-output-relative? | 09:33 |
| minute | ex-parrot: hello! | 09:57 |
| - chomwitt (QUIT: Ping timeout: 256 seconds) (~chomwitt@2a02:587:7a21:2700:1ac0:4dff:fedb:a3f1) | 10:20 | |
| - Ar|stote|is (QUIT: Ping timeout: 255 seconds) (~linx@149.210.24.250) | 10:47 | |
| + Ar|stote|is (~linx@149.210.27.203) | 10:52 | |
| + mjw (~mjw@gnu.wildebeest.org) | 11:46 | |
| - iank (QUIT: Ping timeout: 268 seconds) (~iank@fsf/staff/iank) | 12:10 | |
| + iank (~iank@fsf/staff/iank) | 12:22 | |
| erle | josch it's annotate-output, but with relative time in milliseconds http://news.dieweltistgarnichtso.net/bin/annotate-output-relative | 12:56 |
| erle | josch i don't *think* annotate-output can do this. can it? | 12:57 |
| erle | josch output is like this: https://mister-muffin.de/p/SQGH.txt | 12:59 |
| - mjw (QUIT: Ping timeout: 264 seconds) (~mjw@gnu.wildebeest.org) | 13:29 | |
| josch | erle: could you submit your changes as a merge request to the annotate-output upstream? thanks! | 13:38 |
| * Guest8519 -> mjw | 13:38 | |
| josch | erle: seems we have independently created very similar shell horrors: https://salsa.debian.org/debian/devscripts/-/blob/fa86bf6d6d0683198f8c08da8ce69bb928a53306/scripts/annotate-output.sh#L79 | 13:42 |
| erle | josch oh, i did not know you were the author of annotate-output | 14:21 |
| erle | josch i would have to work a bit on the command line handling to even propose a merge | 14:22 |
| minute | josch: new kernel appears to have a regression with hdmi on imx8mp, investigating (drivers don't load/probe) | 14:27 |
| minute | instead we get > platform 32fd8000.hdmi: deferred probe pending: platform: supplier 32fc0000.blk-ctrl not ready | 14:29 |
| erle | josch btw you can avoid path lookups for commands executed often like date(1), by first capturing it's path with WHATEVER=$(command -v whatever) and then just using $WHATEVER instead | 14:32 |
| erle | obviously this incurs the overhead of one subshell, so it's only ever useful if you execute something a lot | 14:33 |
| erle | josch you can easily see this by comparing “strace -c sh -c 'factor 2'” and “strace -c sh -c '/usr/bin/factor 2'” – on my system the not-absolute path incurs an overhead of 3 stat64 syscalls. | 14:37 |
| erle | so 6 syscalls overall vs 3 syscalls overall | 14:38 |
| grimmware | erle: oh that’s interesting, I thought that all path searches were done by just blindly shipping an execve on the prepended paths in order but I guess that’s shell implementation specific and sh stats instead? Had to do a bunch of work to get around bash and zsh not being an absolute spam cannon of path searches when trying gather executed commands under eBPF | 14:47 |
| erle | grimmware i have no idea if this is shell implementation specific, but it seems to me the overhead is obviously there due to $PATH with fallback paths existing at all. if you go the plan 9 route and use overlay filesystems, all of that is handled somewhere else i guess. | 14:51 |
| erle | (the path can always be absolute then, filesystem gives you the top-most binary) | 14:52 |
| minute | the thing that is pending is > hdmi_blk_ctrl: blk-ctrl@32fc0000 > compatible = "fsl,imx8mp-hdmi-blk-ctrl", "syscon"; | 14:53 |
| minute | the first 9pocket reform https://mastodon.social/@1g0r/112524469149027701 | 14:55 |
| grimmware | Yeah plan 9 just bind mounts it all into /bin | 14:55 |
| minute | http://9lab.org/plan9/mnt-pocket-reform-arrived/ | 14:55 |
| - Gooberpatrol66 (QUIT: Quit: Konversation terminated!) (~Gooberpat@user/gooberpatrol66) | 14:58 | |
| grimmware | minute: nice, looks like he simulates processors for a living so that’s good news for the porting effort, especially given the imx8 already works | 14:59 |
| AbortRetryFail | does it count with just drawterm? :) | 15:08 |
| + chomwitt (~chomwitt@2a02:587:7a21:2700:1ac0:4dff:fedb:a3f1) | 15:09 | |
| AbortRetryFail | man that rgb keyboard looks so cool | 15:09 |
| - chomwitt (QUIT: Ping timeout: 272 seconds) (~chomwitt@2a02:587:7a21:2700:1ac0:4dff:fedb:a3f1) | 15:28 | |
| minute | hdmi patches uses "power-domains@14" | 15:29 |
| minute | but everything else "power-domain@10" etc (singular) | 15:30 |
| + cobra (~cobra@user/Cobra) | 15:41 | |
| + mark_ (~mjw@gnu.wildebeest.org) | 15:58 | |
| AbortRetryFail | how new is the HDMI breakage? I just updated mine Saturday and HDMI was working fine. | 16:17 |
| minute | AbortRetryFail: you have pocket already since saturday? | 16:20 |
| AbortRetryFail | oh, nope. imx8 reform | 16:21 |
| AbortRetryFail | unaffected i guess :) | 16:21 |
| - Ar|stote|is (QUIT: Ping timeout: 268 seconds) (~linx@149.210.27.203) | 16:22 | |
| josch | minute: sorry for the imx8mq hdmi troubles -- it is very likely that i did a bad rebase :/ | 16:38 |
| + Ar|stote|is (~linx@149.210.12.195) | 16:41 | |
| - Ar|stote|is (QUIT: Ping timeout: 268 seconds) (~linx@149.210.12.195) | 16:59 | |
| + Ar|stote|is (~linx@149.210.32.215) | 17:02 | |
| + chomwitt (~chomwitt@2a02:587:7a21:2700:1ac0:4dff:fedb:a3f1) | 17:06 | |
| minute | josch: i don't think so... some device tree nodes just evolved, i think they added some clocks in the kernel that weren't activated now | 17:17 |
| minute | josch: https://source.mnt.re/reform/reform-debian-packages/-/merge_requests/38 | 17:17 |
| minute | (need to test on device after build process, so far i only tested with modified .dtb) | 17:17 |
| - Ar|stote|is (QUIT: Read error: Connection reset by peer) (~linx@149.210.32.215) | 17:21 | |
| + Ar|stote|is (~linx@149.210.32.218) | 17:27 | |
| minute | argh https://source.mnt.re/reform/reform-debian-packages/-/jobs/4415#L2112 | 17:44 |
| minute | josch: do you have some magic for realigning patches like that? | 17:45 |
| josch | minute: i wish i had -- i manually inspect and edit the patch because usually, fuzz means that the hunk is applied at the wrong location | 18:14 |
| josch | and even if it's the correct location, i have to read it manually to make sure that that is indeed the case | 18:14 |
| josch | minute: want me to have a look or are you taking care of it? | 18:16 |
| minute | josch: if you have time, it would be wonderful, i have to run away for today | 18:28 |
| minute | but no stress | 18:28 |
| * mjw -> Guest162 | 18:49 | |
| - Guest162 (QUIT: Killed (tantalum.libera.chat (Nickname regained by services))) (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 18:49 | |
| * mark_ -> mjw | 18:49 | |
| + Guest162 (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 18:50 | |
| - chomwitt (QUIT: Ping timeout: 256 seconds) (~chomwitt@2a02:587:7a21:2700:1ac0:4dff:fedb:a3f1) | 18:55 | |
| + chomwitt (~chomwitt@2a02:587:7a21:2700:1ac0:4dff:fedb:a3f1) | 19:49 | |
| + gustav28 (~gustav@c-2834524e.019-141-67626730.bbcust.telenor.se) | 20:02 | |
| josch | minute: it seems that two spaces were missing and thus the patch only applied with fuzz: https://source.mnt.re/reform/reform-debian-packages/-/commit/5110500a6b6df1fdeb3d1dc0caf1b30bb4b0ead5 | 20:27 |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:20) | 20:57 | |
| + reform13173 (~reform@185.240.55.234) | 21:13 | |
| - reform13173 (PART: !!unknown attribute: msg!!) (~reform@185.240.55.234) | 21:14 | |
| + Guest95 (~Guest95@149.210.32.218) | 21:43 | |
| - Guest95 (QUIT: Client Quit) (~Guest95@149.210.32.218) | 21:43 | |
| - gustav28 (QUIT: Quit: Quit) (~gustav@c-2834524e.019-141-67626730.bbcust.telenor.se) | 22:15 | |
| - robin (QUIT: Read error: Connection reset by peer) (~robin@user/terpri) | 22:46 | |
| - chomwitt (QUIT: Quit: WeeChat 3.8) (~chomwitt@2a02:587:7a21:2700:1ac0:4dff:fedb:a3f1) | 23:21 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!