| sadoon_albader[m | I found a crucial bug in the version of mergerfs shipped in chimaera | 05:40 |
|---|---|---|
| sadoon_albader[m | It duplicates all the directories at first, and then it just causes the mountpoint to be inaccessible | 05:40 |
| sadoon_albader[m | I updated manually to version 2.32.6 from the github page https://github.com/trapexit/mergerfs/releases using their deb package for bullseye and it works fine now | 05:41 |
| sadoon_albader[m | The version shipped is 2.31.0-1 btw | 05:42 |
| bb|hcb | sadoon_albader[m: Thanks for reporting that. Please note that mergerfs comes from Debian without any changes and the problem should be reported there | 05:49 |
| sadoon_albader[m | I figured that might be the case but I thought I should let you guys know first | 05:49 |
| sadoon_albader[m | I'm going to spin a vm of bullseye and see if it has the same behavior, I'm expecting it to | 05:50 |
| sadoon_albader[m | I also have another issue with flashrom + libusb but I need to test it there first as well | 05:50 |
| bb|hcb | One easy way to check if a package is forked by Devuan is https://pkginfo.devuan.org/cgi-bin/policy-query.html - Devuan specific packages contain +devuanX in their version | 05:53 |
| bb|hcb | Testing in bullseye vm is the proper way to do that :) Both flashrom + libusb are not forked | 05:57 |
| xrogaan | sadoon_albader[m: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=989565 | 06:15 |
| xrogaan | So they probably need to backport the patch. | 06:16 |
| sadoon_albader[m | Yeah this is a major issue | 06:17 |
| xrogaan | only because of the so called "freeze" | 06:17 |
| sadoon_albader[m | I thought the whole point was to only update for major bugs and security issues | 06:18 |
| sadoon_albader[m | this is both | 06:18 |
| sadoon_albader[m | potentially | 06:18 |
| bb|hcb | Two things can be done in Debian - add the latest ver to bullseye-backports and cherrypick the serious problems (only the ones that prevent usage) to stable-proposed-updates | 06:21 |
| bb|hcb | But the later shall use the older version package. It would be helpful for the maintainer to file bugs for the other problems that are fixed upstream | 06:22 |
| bb|hcb | And problems that do not prevent usage will not be included. The whole process is really slow | 06:23 |
| xrogaan | it kinds of duplicate the work. | 06:24 |
| bb|hcb | mergerfs is already in Debian testing: https://qa.debian.org/developer.php?login=rrs@debian.org#mergerfs | 06:28 |
| bb|hcb | That is what stable is defined to be :) | 06:29 |
| xrogaan | it's not stable if it's broken :P | 06:31 |
| sadoon_albader[m | Guys I would really appreciate some help if someone has any experience with this | 15:56 |
| sadoon_albader[m | Can't find info anywhere | 15:56 |
| sadoon_albader[m | I'm trying to build webkit2gtk from source using apt source and dpkg-buildpackage | 15:57 |
| sadoon_albader[m | I want to build an optimized version with -O3 and -flto if possible too | 15:57 |
| sadoon_albader[m | Because on ppc64le there is no JIT for JavaScript | 15:57 |
| sadoon_albader[m | How do I pass those options? And how do I make sure that dpkg-buildpackage respects my options? | 15:58 |
| bgstack15 | sadoon_albader[m: you will want to modify the code within a section of debian/rules | 16:26 |
| sadoon_albader[m | I'll check it out, thanks! | 16:27 |
| bgstack15 | override_dh_auto_build: | 16:27 |
| bgstack15 | dh_auto_build -O3 -flto | 16:27 |
| bgstack15 | iirc | 16:27 |
| bgstack15 | One of my examples replaces the contents of override_dh_auto_build: with | 16:28 |
| bgstack15 | make -f client.mk build (so the exact instructions I want, rather than dh_auto_build with some extra params) | 16:28 |
| bgstack15 | And actually if you go with the dh_auto_build line, then you should pass parameters to the make or gcc command, after a double dash -- | 16:29 |
| bgstack15 | override_dh_auto_build: | 16:30 |
| bgstack15 | dh_auto_build -- VERBOSE=1 NOSTRIP=1 ARCH=$(DEB_HOST_ARCH) | 16:30 |
| bgstack15 | Devuan meet today at 20:30 UTC. Pad is here: https://pad.dyne.org/code/#/2/code/edit/y-12l4gGEDXyACn2ZB13sLvW/ | 16:36 |
| sadoon_albader[m | bgstack15 thanks I'll try that too | 16:37 |
| sadoon_albader[m | currently I'm actually trying to force it to build JIT first | 16:38 |
| sadoon_albader[m | Might have some luck there | 16:38 |
| sadoon_albader[m | It's just a theory, but I think it's not building JIT and instead using CLoop because we're on 64kb pages | 16:39 |
| sadoon_albader[m | There's an explicit option in webkit2gtk-2.32.3/Source/cmake/WebKitFeatures.cmake on line 82 that disables JIT for 64kb systems | 16:40 |
| bgstack15 | sorry, that is way beyond my understanding. | 16:45 |
| bgstack15 | Chances are it was disabled for a reason on that architecture... | 16:45 |
| sadoon_albader[m | It's not disabled specifically for ppc64le | 16:47 |
| sadoon_albader[m | I think it's only triggering because of the option for 64kb pages, and that option is there as described by the people who wrote it, for aarch64 | 16:48 |
| sadoon_albader[m | Otherwise if that doesn't work I could just pass -O3 -mtune -mcpu -flto and the whole nine yards, just to get CLoop to be a bit faster | 16:49 |
| bgstack15 | Good! It sounds like you know what needs to be done, so just getting it configured that way in the dpkg is all that's left. | 16:49 |
| bgstack15 | That's the part I understand. | 16:49 |
| sadoon_albader[m | And that was the part I didn't understand :D | 17:03 |
| bb|hcb | sadoon_albader[m, bgstack15: I remember seeing something about gtk stuff not playing nicely with -flto | 17:26 |
| sadoon_albader[m | won't hurt to try | 17:28 |
| sadoon_albader[m | It really doesn't want to build JIT.. bummer | 17:29 |
| sadoon_albader[m | time to try the options | 17:29 |
| bb|hcb | yep, just some build time. if the flto problem is there (I may be wrong about it being in gtk) the build will fail | 17:56 |
| sadoon_albader[m | It reached ~4700 out of ~4800 objects built and no errors yet :D | 18:17 |
| sadoon_albader[m | Wow the webpage object file has taken more than 5 minutes and still | 18:22 |
| sadoon_albader[m | took 30 minutes heh | 18:38 |
| sadoon_albader[m | fml, jdupes not found. something I didn't install screwed it up | 18:40 |
| sadoon_albader[m | All the files are built though | 18:40 |
| sadoon_albader[m | How can I tell it to package it without rebuilding? | 18:41 |
| sadoon_albader[m | does the -nc option do that? | 18:41 |
| sadoon_albader[m | ok yes it does, sorry guys I'm kinda new to the debian package building, used to arch and gentoo | 18:42 |
| sadoon_albader[m | Ok wow that's quite a bit faster :D | 18:45 |
| sadoon_albader[m | If anyone is on ppc64le and wants to try an LTOd and fully optimized webkit2gtk, let me know! | 18:45 |
| bgstack15 | I don't think there's a dpkg-buildpackage option for assembling a package, that doesn't also run the build steps. | 18:54 |
| bgstack15 | Or did you actually discover the way? | 18:55 |
| sadoon_albader[m | it's in the help output :D | 18:57 |
| sadoon_albader[m | So it probably ran the build steps but seeing as the project was already built it did nothing and packaged it | 18:57 |
| sadoon_albader[m | I used the -nc option | 18:57 |
| sadoon_albader[m | which instructs it not to clean | 18:57 |
| bgstack15 | oh, so you had to run the previous command with a -nc | 20:11 |
| sadoon_albader[m | yep | 20:46 |
| sadoon_albader[m | I'm getting an error on another package build, it's telling me "open: no such file or directory" and "elf_open no such..." | 20:46 |
| sadoon_albader[m | I searched online and can't find any info on those | 20:46 |
| sadoon_albader[m | nvm I messed up lol, needs some patches | 20:54 |
| sadoon_albader[m | Trying to build qtwebengine | 20:54 |
| UsL | it was nice to see you all in the meet. Too bad my toaster couldn't stay: http://0x0.st/-38Y.png (upper left corner for toaster levels) | 23:32 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
