libera/#devuan-arm/ Friday, 2023-02-17

gnarfaceagneli: not sure, you might have to rebuild it though04:15
gnarfacethat's using the armel port?04:15
gnarfacenot armhf right?04:15
agnelignarface: i have always been using armhf07:55
agnelion same machine07:55
agnelikindly forgive ubuntu link: https://wiki.ubuntu.com/ARM/TEGRA/AC10007:56
agneliin case of Firefox it is NEON that is required07:57
agnelipalemoon used to work but also gives me Illegal instruction....07:57
agneli:(07:58
gnarfaceyea it's coming up as a common problem, but i think other than switching to armel your only hope is to rebuild everything08:34
gnarfaceobviously neither solution is optimal08:34
gnarfacebut i do recall seeing the proper compiler flags around somewhere08:35
gnarfacethere is a way to preserve hardware floating point support without needing NEON, it's just that some time ago debian decided there wasn't enough market share of that type of hardware to care about (i think they were wrong and possibly biased by a disinformation campaign lead by organized crime but it's not really on-topic)08:36
agnelithank you for your answer gnarface, all in all: gentoo-arm?08:41
gnarfaceyea, i think this particular use case makes a good argument for the way gentoo is set up08:42
gnarfacethough a mean build script could certainly do the same thing on devuan08:42
agnelii have no idea how to do such a script :)08:43
agneliand gentoo I am using as a daily driver on all machines08:44
agnelithis particular machine was the only exception08:44
agnelii liked devuan...08:44
agnelinow question is - issue is only with daedalus08:45
agnelior earlier releases as well?08:45
gnarfacenot sure how far back the problem goes, but i do know that needing the NEON instruction set wasn't always there for palemoon08:46
gnarfacei think at some point the same issue came up for midori as well08:46
gnarfacei think individual developers are making the decision to add it08:46
agnelii used to use palemoon without problem08:49
agneliin 80386 time without fpu there were software solutions to work around that problem...08:50
gnarfacei don't know of such a thing for NEON support but that doesn't mean it doesn't exist08:52
gnarfaceit should be possible i think08:52
gnarfacemaybe look in the build options for a recent kenrel08:52
gnarfacekernel*08:52
agnelii kernels I build myself08:56
agnelino support for my board in devuan08:56
agnelipaz00 - it is removed08:56
agnelithough it is in vanila kernel08:56
gnarfacei had been thinking of trying to go through a proper hardfloat build for my old raspberry pi09:08
gnarfacebut even the suggestion of the proper compiler flags for that came from the gentoo wiki09:08
gnarfaceif i were to do this for devuan i'd likely be borrowing heavily from their documentation09:09
agnelignarface: https://wiki.debian.org/ArmHardFloatPort11:42
agnelithat seems official...11:43
agnelihttps://en.wikipedia.org/wiki/Tegra#Specifications11:44
agneliI am nVidia Tegra 211:45
agneli"Currently the Debian armhf port requires at least an Armv7 CPU with Thumb-2 and VFP3D16"11:46
agnelii have VFP311:46
agnelii am listed there:11:48
agnelihttps://wiki.debian.org/ArmHardFloatPort#Partial_reference_of_SoC_and_supported_ISAs11:48
gnarfacehmm12:08
gnarfacein that case you should definitely file a bug report about it with debian12:08
gnarfaceit won't help my rpi1 issue unfortunately, because it's actually Armv612:09
agneliit might be mesa actually needeing neon12:27
agnelijust like firefox does...12:27
gnarfacewell you might be able to test that by disabling accelerated rendering in firefox13:40
gnarfacebut i'm not sure it's just mesa13:40
agnelii might not gnarface :)13:48
agneliff requires neon13:48
agneliit does not even start13:48
agnelidives illegal instructions13:48
agneliunles you know something i do not know...13:49
agneliwhich is quite probable ;)13:49
bb|hcbhttps://wiki.debian.org/ArchitectureSpecificsMemo#armhf18:20
bb|hcbMy (highly uneducated guess) would be that recent gcc started generating code with NEON instructions and thus violated the baseline of armhf. Most of the people have NEON and this went unnoticed. Maybe it is a good idea to check if that is already reported as a debian bug18:22
agneli"armv7 with VFPv3-D16 floating point. NEON is not guaranteed."18:39
agnelinot guaranteed it does not mean it cannot be there...18:39
agneliit means it can be there but is not mandatory18:40
bb|hcbNot guaranteed means it is not in the baseline. Programs are OK to do a runtime check and use it, if present, but if not present, there must be a fallback18:42
agneliach OK18:44
agnelithank you bb|hcb for your explanation :)18:44
agneli"[...]you are bringing hope to those that have none[...]"18:45
bb|hcbNote that the baseline may have changed and those pages to be outdated. I wouldn't have noticed, because I only have arm64 devices. Search Debian bugs for "armhf neon" or similar, maybe also file bugs for the packages that do not work.18:47
bb|hcbI'd suppose that the problem is not in the packages themselves but in the toolchain18:47
agnelii am doing just that18:48
agnelibut what you say should mean that firefox also should have some sort of fallback18:48
agneliotherwise it violates baseline18:48
agnelifirefox in devuan is not working from me since long time18:49
agnelimaybe even since very begining of devuan18:49
bb|hcbI'd suppose the problem came with a new toolchain. Firefox is not forked and the binary package comes 1:1 from Debian19:58
agnelifirefox was never working for me in devuan20:13
agnelilast working was debian wheezy20:13
agnelimesa was ok till now20:13
agneli=> 0xb6ff07d4:  vmov    d16, r2, r020:14
agnelithis is the offending instrution in glxgears20:14
agnelii did not find anything relevant on debian bugs page yet20:14
agnelihttps://bugs.debian.org/cgi-bin/bugreport.cgi?bug=98279420:25
agnelisomebody _is_ chasing firefox issue...20:25
agnelihttps://lists.debian.org/debian-gcc/2022/06/msg00045.html20:38

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!