Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open Beta v8.0 | Please help testing and hardening the upcoming release #5137

Closed
MichaIng opened this issue Jan 4, 2022 · 39 comments · Fixed by #5141, #5101, #5146 or #5147
Closed

Open Beta v8.0 | Please help testing and hardening the upcoming release #5137

MichaIng opened this issue Jan 4, 2022 · 39 comments · Fixed by #5141, #5101, #5146 or #5147
Labels
Beta 🧪 Issues specific to the Beta branch testing Information ℹ️
Milestone

Comments

@MichaIng
Copy link
Owner

MichaIng commented Jan 4, 2022

RC version v8.0.2
v8.0.0 PR #5101
v8.0.1 PR #5141
v8.0.1 PR #5147
Raw changelog https://github.com/MichaIng/DietPi/blob/beta/CHANGELOG.txt
Code changes master...beta
Release date 2022-01-08

How to apply: https://github.com/MichaIng/DietPi/blob/master/BRANCH_SYSTEM.md

Related/solved issues: https://github.com/MichaIng/DietPi/issues?q=is%3Aissue+milestone%3Av8.0


Beta v8.0.2

(2022-01-08)

Stretch support

  • Support for Debian Stretch has now been removed, which allowed us to do a major code cleanup and migrate to some newer methods in different areas which are only available on Buster and above.

Changes

  • Network | For new images, the default DHCP timeout is not reduced to 10 seconds anymore. This could have been too short, in which case network targets of the boot sequence are reached before an IP has been actually assigned. Especially in case of AUTO_SETUP_AUTOMATED=1 this could have lead to connection test timeouts and hence abortion of the automated first run setup. Many thanks to @jpeg2600 for reporting such a case: Slow DHCP server causing trouble with auto installation (with workaround)  #5143
  • DietPi-Config | When configuring a first WiFi slot via scanning for SSIDs, the WiFi adapter is now not brought up via "ifup" but "ip l dev wlanX up". This way no DHCP and WPA client starts are triggered, which are doomed to fail when the WiFi adapter is not connected to any access point yet. Especially in case of the default 60 seconds DHCP timeout, this would otherwise cause an unnecessary long delay.
  • DietPi-Dashboard | Added an option to view multiple dashboard nodes from one frontend web interface. Related to this, the backend can now be installed only, reducing memory usage and making it impossible to manually access the node outside of the backend API.
  • DietPi-Print_large | This new script has been added which can be executed or sourced from /boot/dietpi/func/dietpi-print_large to print the string passed via first argument in large figlet style fonts. It currently only supports the characters a-z, A-Z, 0-9, dot and dash, i.e. those commonly allowed in hostnames.
  • DietPi-Banner | Added an option to print the system's hostname in large figlet style fonts, right below the banner header. In case it is enabled as well, the regular/small hostname line will then be skipped. Many thanks to @matellis for implementing this feature: DietPi-Banner | Adds large style banner #5113
  • DietPi-Software | Mono: "mono-complete" won't be installed anymore but only "mono-devel" instead. This skips the XSP4 webserver service, which listens on port 8084 by default where it conflicts with File Browser. This only affects new instals. You can manually apply this change via "apt-mark manual mono-devel && apt --autoremove purge mono-complete". Many thanks to @jaguar489 for reporting this issue: Dietpi-Software File browser won't start #5093
  • DietPi-Software | FuguHub: The outdated official installer has been replaced with a fully automated own setup, removing the obsolete or even harmful interactive dialogues. On fresh installs, an admin account "dietpi" with global software password is created as well.
  • DietPi-Software | myMPD: Installation is now done via official APT repository, which means quicker install compared to compiling from source, fewer dependencies and easier updates via "apt upgrade": Error installing myMPD #5115
  • DietPi-Software | Airsonic: Since the project has been archived by its maintainer and suffers from the critical Log4Shell security vulnerability, we moved to the advertised fork "Airsonic-Advanced" which also enables compatibility with Java 17 and hence Debian Bullseye. Airsonic will be reinstalled as part of the DietPi update to apply the migration. All data and settings will be preserved and migrated automatically.

Fixes

  • Raspberry Pi | Resolved an issue where unintentionally the turbo mode was enabled. This got backported to v7.9, to our existing RPi images and via live patch. Many thanks to @ayo-x and @whyisthisbroken for reporting this issue: Raspberry Pi | CPU frequency scaling does not work #5088
  • DietPi-FS_partition_resize | Resolved an issue where in some cases the boot partition failed to mount on first boot since after the partition resize the system in some cases didn't get sufficient time to fully apply the change before continuing with the boot mount attempt. Many thanks to @Mausy5043 and @sistemicorp for reporting this issue: rebooting after PREP-script #5006
  • DietPi-Config | Resolved an issue where on Raspberry Pi, when the full KMS display driver is used, the sound card selection could have been falsely applied. With full KMS enabled, for each HDMI port an additional HDMI sound card appears (replacing the regular firmware HDMI sound device), even when no HDMI device is attached and with onboard audio disabled. This lead to an incrementation of external sound card indices. The appearance of the KMS HDMI sound cards is now aligned with the onboard firmware HDMI sound cards, i.e. if not onboard "auto" or HDMI is selected in dietpi-config, now the KMS HDMI sound devices are disabled as well, making use of the "noaudio" device tree overlay parameter.
  • DietPi-Software | Resolved an issue where on first boot, choice and preference menu selections triggered an error as the install state file did not exist yet. Many thanks to @bsheeres for reporting this issue: Clean Install Errors on Software - Log System (AUTO_SETUP_LOGGING_INDEX) #5080
  • DietPi-Software | Resolved an issue where the deprecated AUTO_SETUP_FILE_SERVER_INDEX dietpi.txt option was applied inverted. "-1" should have installed ProFTPD and "-2" Samba, while it was done the opposite way. However, note that this setting is not present on current images and deprecated. Use AUTO_SETUP_INSTALL_SOFTWARE_ID instead to mark any one or multiple file servers for automatic install. Many thanks to @bsheeres for reporting this issue: AUTO_SETUP_FILE_SERVER_INDEX options reversed #5081
  • DietPi-Software | Browsers: On the ARMv6 Raspberry Pi models Raspberry Pi 1 and Zero (1) since Bullseye, Chromium and Firefox fail to start due to hardware and build limitations which are outside of our control. Since those are currently the only two browser available from dietpi-software, they have been disabled together with the browser preference menu for those systems, until we find and implement a good additional compatible browser. Further info: [Bullseye] Chromium fails to start on ARMv6 RPi-Distro/chromium#21
  • DietPi-Software | Kodi: Resolved an issue where subsequent Kodi reinstalls created duplicates of the CMA parameter for the KMS overlay setting in config.txt. The duplicates are patched out during DietPi update.
  • DietPi-Software | Docker: Resolved an issue where the install was aborted due to an incorrect check for missing kernel modules. Many thanks to @dragonandy for reporting this issue: Docker and its dependants will be deselected Modules for the loaded kernel are missing. #5061
  • DietPi-Software | Pi-hole: Resolved an issue where the teleporter import log was not shown but denied by the X-Frame-Options header: https://discourse.pi-hole.net/t/unable-to-restore-teleporter-backup-fresh-install-no-funky-changes-made/51573
  • DietPi-Software | Blynk Server: Resolved an issue on ARMv6 RPi models where the install failed as the latest Blynk Server release does not contain a Java 8 build. Since the older Java 8 build does not contain a native Log4Shell vulnerability mitigation, it is added server-wise on these systems.
  • DietPi-Software | Mycroft AI: Resolved an issue on Bullseye (and above) systems where "mycroft-cli-client" command fails with a permissions issue, even as root user. Many thanks to @berndverhofstadt for reporting this issue: DietPi-Software | Mycroft AI: mycroft-cli-client fails as root on Bullseye #5100
  • DietPi-Software | Nukkit: Resolved an issue where the install failed due to a changed download URL. The Jenkins instance has moved to ci.opencollab.dev, where also the Geyser and Floodgate projects are hosted.
  • DietPi-Software | FuguHub: Resolved an issue where the uninstall failed as the process was not stopped as intended. Many thanks to @kd9352 for reporting this issue: Concurrent execution of DietPi-Software detected #5058
  • DietPi-Software | myMPD: Resolved an issue where the installation failed due to an updated dependency. Many thanks to @supersexy for reporting this issue: Error installing myMPD #5115
  • DietPi-Software | Python 3: Worked around an issue where the pip reinstall failed. Many thanks to @hueppinr for reporting this issue: Dietpi-software | Python3 - reinstall failed due to conflict between Setuptools and Distutils #5117
  • DietPi-Software | Unbound: Resolved an issue where the "unbound-resolvconf" service applied localhost automatically as local nameserver, if the "resolvconf" package was installed. Since often Unbound is installed in combination with Pi-hole or AdGuard Home, and those are usually used by network clients only, not the server itself, "unbound-resolvconf" is disabled now on Unbound installs. If Unbound is wanted as local resolver as well, it should be explicitly configured, e.g. via dietpi-config network options. Many thanks to @Ianszh for reporting this issue: DietPi-Software | Unbound: Disable unbound-resolvconf.service #5133
  • DietPi-Software | Mosquitto: Applied a workaround on ARMv6 where the latest Mosquitto packages from the official APT repository are not ARMv6-compatible. Many thanks to @thomasmockridge for reporting this issue: DietPi-Software | Mosquitto fails to install on ARMv6 #5140
@MichaIng MichaIng added Testing/testers required 🔽 Information ℹ️ Beta 🧪 Issues specific to the Beta branch testing labels Jan 4, 2022
@MichaIng MichaIng added this to the v8.0 milestone Jan 4, 2022
@MichaIng MichaIng pinned this issue Jan 4, 2022
@billouetaudrey
Copy link

billouetaudrey commented Jan 5, 2022

Hello, install done, will tell if any bug.

For now, this is the bug when I install it :

[FAILED] DietPi-Survey | Purging survey data
 - Command: curl --key  -m 20 -sT 80017429-532b-437d-9b75-256a91d206ec.txt sftp://dietpi-survey:[email protected]/survey/

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 5, 2022

Can you try to run this manually and past the output here?

cd /tmp
> 80017429-532b-437d-9b75-256a91d206ec.txt
curl --key  -m 20 -vT 80017429-532b-437d-9b75-256a91d206ec.txt sftp://dietpi-survey:[email protected]/survey/

@billouetaudrey
Copy link

billouetaudrey commented Jan 5, 2022

Seem stuck here :

root@raspberrypi:~# cd /tmp
> 80017429-532b-437d-9b75-256a91d206ec.txt
curl --key  -m 20 -vT 80017429-532b-437d-9b75-256a91d206ec.txt sftp://dietpi-survey:[email protected]/survey/
Warning: The file name argument '-m' looks like a flag.
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0*   Trying 0.0.0.20:80...
  0     0    0     0    0     0      0      0 --:--:--  0:00:16 --:--:--     0

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 5, 2022

Ah sorry, I forgot the empty --key argument:

cd /tmp
> 80017429-532b-437d-9b75-256a91d206ec.txt
curl --key '' -m 20 -vT 80017429-532b-437d-9b75-256a91d206ec.txt 'sftp://dietpi-survey:[email protected]/survey/'

@MichaIng MichaIng linked a pull request Jan 5, 2022 that will close this issue
@billouetaudrey
Copy link

billouetaudrey commented Jan 5, 2022

* Closing connection 0
curl: (60) SSL peer certificate or SSH remote key was not OK
More details here: https://curl.se/docs/sslcerts.html

curl failed to verify the legitimacy of the server and therefore could not
establish a secure connection to it. To learn more about this situation and
how to fix it, please visit the web page mentioned above.

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 5, 2022

Is our host key still present?

grep 'ssh.dietpi.com' /root/.ssh/known_hosts

@billouetaudrey
Copy link

Nope :(

root@raspberrypi:/tmp# grep 'ssh.dietpi.com' /root/.ssh/known_hosts grep: /root/.ssh/known_hosts: Aucun fichier ou dossier de ce type root@raspberrypi:/tmp#

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 5, 2022

Recreate it with this command:

echo 'ssh.dietpi.com ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDE6aw3r6aOEqendNu376iiCHr9tGBIWPgfrLkzjXjEsHGyVSUFNnZt6pftrDeK7UX+qX4FxOwQlugG4fymOHbimRCFiv6cf7VpYg1Ednquq9TLb7/cIIbX8a6AuRmX4fjdGuqwmBq3OG7ZksFcYEFKt5U4mAJIaL8hXiM2iXjgY02LqiQY/QWATsHI4ie9ZOnwrQE+Rr6mASN1BVFuIgyHIbwX54jsFSnZ/7CdBMkuAd9B8JkxppWVYpYIFHE9oWNfjh/epdK8yv9Oo6r0w5Rb+4qaAc5g+RAaknHeV6Gp75d2lxBdCm5XknKKbGma2+/DfoE8WZTSgzXrYcRlStYN' > /root/.ssh/known_hosts

@Joulinar
Copy link
Collaborator

Joulinar commented Jan 5, 2022

@MichaIng
we had already a couple of cases where the ssh-rsa key was missing. Should we include some error handle within the survey function?

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 5, 2022

Or we re-create it with a patch on next update? But a check for grep '^ssh.dietpi.com' /root/.ssh/known_hosts within survey and bug report scripts is actually a good idea as well, so the upload is not attempted in the first place when doomed to fail but a meaningful error message shown instead.

@Joulinar
Copy link
Collaborator

Joulinar commented Jan 6, 2022

question is it missing in some cases? By accident or by purpose?

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 6, 2022

Another reason to go with the check + error message, so users who intentionally remove it get an info for what it is required.

@billouetaudrey
Copy link

billouetaudrey commented Jan 6, 2022

Oh maybe my fault, when I had ssh issue, maybe I delete all folder .ssh/

Sorry

Question : Do you have any Telegram Group ? :)

@Joulinar
Copy link
Collaborator

Joulinar commented Jan 6, 2022

Nope we don't have such a group. There is a small user community on Facebook.

@billouetaudrey
Copy link

Nope we don't have such a group. There is a small user community on Facebook.

Ok ;)

@sheeepdev
Copy link

Nope we don't have such a group. There is a small user community on Facebook.

Maybe you could make a discord server or such?

@tandy-1000
Copy link

Nope we don't have such a group. There is a small user community on Facebook.

Maybe you could make a discord server or such?

No thank you, Matrix please, no proprietary services for open source projects :/

@sheeepdev
Copy link

Nope we don't have such a group. There is a small user community on Facebook.

Maybe you could make a discord server or such?

No thank you, Matrix please, no proprietary services for open source projects :/

Yeah, I know. Discord is poprietary, but still, a lot of open source projects use it.
I don't like Matrix, its very clunky and not user-friendly at all.

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 6, 2022

I haven't used it much, but there are a lot of Matrix clients, the benefit of an open protocol. So I guess there will be one which suites your expectations, or is it somehow a limitation of the protocol itself compared to Discord?

Generally an additional platform is not always better, it splits reports and solutions for common problems, making it more difficult for us and for users to quickly find or link solutions or contribute to discussions/issues. And it requires us to watch and react on an additional platform. I have a Discord account, but am, to be true, not very reactive there, as forum and in my case especially GitHub keeps me busy. We try to be very reactive on both platforms, so not sure whether a chat platforms adds so much value. However, we were playing with the idea to host an own Synapse Matrix server, being the obvious solution as our server runs DietPi (of course) and we have Synapse integrated into dietpi-software 😉.

@sheeepdev
Copy link

sheeepdev commented Jan 6, 2022

I haven't used it much, but there are a lot of Matrix clients, the benefit of an open protocol. So I guess there will be one which suites your expectations, or is it somehow a limitation of the protocol itself compared to Discord?

Generally an additional platform is not always better, it splits reports and solutions for common problems, making it more difficult for us and for users to quickly find or link solutions or contribute to discussions/issues. And it requires us to watch and react on an additional platform. I have a Discord account, but am, to be true, not very reactive there, as forum and in my case especially GitHub keeps me busy. We try to be very reactive on both platforms, so not sure whether a chat platforms adds so much value. However, we were playing with the idea to host an own Synapse Matrix server, being the obvious solution as our server runs DietPi (of course) and we have Synapse integrated into dietpi-software wink.

Pretty much all of the matrix clients are the same, just the look is different. It is confusing because of all the codes, rooms, etc. I find discord more user-friendly, you just join the server and pick a channel to talk. In matrix you need to pick a homeserver, then make your account, then join the room (or whatever, I don't know) then join the channel and finally you can chat with others.
Also, installing matrix and connecting and verifying your session is also a trouble. You would need to save the codes somewhere, which I can do but then I forget where I stored them and I need to re-do it all again.

It just that to a normal user who wants to chat about DietPI Matrix is too complicated. With discord, you just register and join the server and all the information, channels and etc. are there.

But I agree that Matrix is more private, open and secure. I'm not hating on matrix, it is just too complex for someone who just wants to chat.

@Joulinar
Copy link
Collaborator

Joulinar commented Jan 6, 2022

As we are 2 people only who do user support, there is quite a good chance that there is nobody you could chat with. Doesn't matter if Matrix or Discord. 😉

@tandy-1000
Copy link

I haven't used it much, but there are a lot of Matrix clients, the benefit of an open protocol. So I guess there will be one which suites your expectations, or is it somehow a limitation of the protocol itself compared to Discord?
Generally an additional platform is not always better, it splits reports and solutions for common problems, making it more difficult for us and for users to quickly find or link solutions or contribute to discussions/issues. And it requires us to watch and react on an additional platform. I have a Discord account, but am, to be true, not very reactive there, as forum and in my case especially GitHub keeps me busy. We try to be very reactive on both platforms, so not sure whether a chat platforms adds so much value. However, we were playing with the idea to host an own Synapse Matrix server, being the obvious solution as our server runs DietPi (of course) and we have Synapse integrated into dietpi-software wink.

Pretty much all of the matrix clients are the same, just the look is different. It is confusing because of all the codes, rooms, etc. I find discord more user-friendly, you just join the server and pick a channel to talk. In matrix you need to pick a homeserver, then make your account, then join the room (or whatever, I don't know) then join the channel and finally you can chat with others. Also, installing matrix and connecting and verifying your session is also a trouble. You would need to save the codes somewhere, which I can do but then I forget where I stored them and I need to re-do it all again.

It just that to a normal user who wants to chat about DietPI Matrix is too complicated. With discord, you just register and join the server and all the information, channels and etc. are there.

But I agree that Matrix is more private, open and secure. I'm not hating on matrix, it is just too complex for someone who just wants to chat.

you can just use the default matrix.org homeserver, so its about as simple..

As we are 2 people only who do user support, there is quite a good chance that there is nobody you could chat with. Doesn't matter if Matrix or Discord. wink

I agree with this though, any chat platform will add to the burden of maintenance and moderation, the forum / github is probably fine enough until there are people ready to moderate it.

@dasrocky
Copy link

dasrocky commented Jan 6, 2022

Another major version of dietpi, hurray. Thank you.
Luckily had no issue with this beta update. Jump from v7.4 to v8.0. RPi4 - Home media server

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 6, 2022

For those who are interested, why a major version increment: We decided to do this whenever support for an older Debian version (Stretch in this case) or older DietPi versions (DietPi prior to v6.14 in this case) are dropped, hence follow the intended meaning of major versions in semantic versioning to do a breaking change. Other than that, which allowed us to drop and enhance a lot of code to keep up Stretch compatibility, there is not further meaning of it.

Old DietPi systems however only need to call dietpi-update two times to get to v8.0, first migration to the pre-v6.14 branch, updating to v7.9 is done, second from v7.9 the update to v8.0 from master branch is offered 😉. Debian Stretch systems can be quite painlessly upgraded to Buster: https://dietpi.com/docs/usage/#how-to-upgrade-to-buster

@m0jek
Copy link

m0jek commented Jan 7, 2022

Nope we don't have such a group. There is a small user community on Facebook.

We do need something for those that don't use Meta / Facebook. My vote will be Telegram or Discord or Matrix. What is useful with Matrix is that there are bridges for Telegram and Discord ...

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 7, 2022

We do need something for those that don't use Meta / Facebook.

GitHub and forum.

What is useful with Matrix is that there are bridges for Telegram and Discord ...

That is actually very nice, indeed. How does it work, is this something that can/needs to be setup on the Matrix server with a Telegram/Discord account which then exposes/syncs Matrix chats on Telegram/Discord automatically?

@billouetaudrey
Copy link

I think I started a discussion about Telegram etc... lol

I know with Telegram ( certainly with Discord too ) you can sync Github changed

@tandy-1000
Copy link

That is actually very nice, indeed. How does it work, is this something that can/needs to be setup on the Matrix server with a Telegram/Discord account which then exposes/syncs Matrix chats on Telegram/Discord automatically?

you can selfhost the bridge bot, or use t2bot.io which hosts telegram and discord bots

This was linked to pull requests Jan 7, 2022
@MichaIng
Copy link
Owner Author

MichaIng commented Jan 7, 2022

Beta v8.0.1 has been merged with a few changes and fixes: #5141

@MichaIng MichaIng linked a pull request Jan 8, 2022 that will close this issue
@Mausy5043
Copy link

My 2 cents regarding hardening: you might want to consider disabling SSH-logins by root. The dietpi user has sudo capabities so root really shouldn't be exposed.

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 8, 2022

root login is currently required for first run setup, especially when AUTO_SETUP_AUTOMATED=1 has been set. Since dietpi has full passwordless sudo permissions, both imply exactly the same security risk, with the only different that sudo calls are generally logged.

Furthermore we plan to remove the dietpi user in the future and instead add an option to create a new/custom login user during first run setup. When doing so, we can additionally offer to have root SSH logins disabled, or lock any root logins completely (usermod -L root), given that the new user has sudo permissions, of course.

But I agree in any case that hardening SSH logins (disabling root login and especially also enabling key authentication) needs to be more easy/transparent to users. While it can be easily done manually, I guess many don't know about this and leave root password login enabled. For this release it is of course too late, but probably you can open an issue so we can track for next release.

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 8, 2022

Beta v8.0.2 has been merged: #5147

We'll do a last code review and in case some quick tests, before doing the DietPi v8.0 release until midnight.

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 8, 2022

DietPi v8.0 has been released, many thanks to all testers ❤️!

@MichaIng MichaIng unpinned this issue Jan 8, 2022
@dasrocky
Copy link

dasrocky commented Jan 9, 2022

Congrats again on major release. Just a question. Wasn't sure should I be making another issue for a question. I am planning to have a clean install for Bullseye upgrade.
Question:

  1. Is it right time to switch 64bit/v8 or stay with 32bit/v7? (I went through rpi forum for kodi, noticed it has support hardware decoding for 64bit)

Details:

  • SBC: RPi4 B
  • RAM: 4 GB
  • SD: 32GB
  • Purpose: Home media server
  • Software used: Kodi, Emby, Sonarr, Radarr, Deluge, qbittorrent.
    *****Hardware de/encoding is a must

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 9, 2022

Congrats again on major release.

Many thanks 🙂.

  1. Is it right time to switch 64bit/v8 or stay with 32bit/v7? (I went through rpi forum for kodi, noticed it has support hardware decoding for 64bit)

Yes you can do that, indeed since Bullseye there is an arm64 build from the RPi repository with hardware acceleration available. It uses the KMS display driver via GBM and video de/encoding is done via FFmpeg and V4L2 libraries. The same is however true for the 32-bit/armhf Kodi package, so the 64-bit image won't give you much benefit as long as you do not need it for other software, or run such software which makes vast use of 64-bit integers. The ARMv7 image should be however preferred over the ARMv6 image in every case where you do not plan to use the same image on ARMv6-only RPi models (RPi 1/Zero).

@dasrocky
Copy link

The same is however true for the 32-bit/armhf Kodi package, so the 64-bit image won't give you much benefit as long as you do not need it for other software, or run such software which makes vast use of 64-bit integers.

Even as per my research on it, didn't find any mentions of huge differences other than small performance differences and package issues. So I think I will stick with 32bit for now.

@MichaIng
Copy link
Owner Author

MichaIng commented Jan 10, 2022

A small number of software titles is not available for 64-bit ARM, another small number not for 32-bit ARM, it may have reached quality. But also from performance point of view, 64-bit is faster only in certain cases, as mentioned when software code deals with 64-bit size integers often, while it consumes usually notably more disk and RAM space, which has again a negative impact on performance. So I'd say one cannot give a generic preference (for RPi 4 where RAM size is not inherently a problem) for either the one or the other, but more important is whether all aimed software titles are supported 😉.

Reminds me that we need to add supported architectures to e.g. the software list in our Wiki, so one can check this before flashing the image.

@dasrocky
Copy link

A small number of software titles is not available for 64-bit ARM, another small number not for 32-bit ARM, it may have reached quality. But also from performance point of view, 64-bit is faster only in certain cases, as mentioned when software code deals with 64-bit size integers often, while it consumes usually notably more disk and RAM space, which has again a negative impact on performance.

Understood, and got a clean install of dietpi for RPi. Luckily, no issue. Way fewer crashes for Kodi compared to previously. I will stick with custom Kodi service rather than autostart. I noticed ur comments rpi forum previously and I didn't understand some parts initially until faced that issue(sound issue). Luckily it was an easy fix.

Reminds me that we need to add supported architectures to e.g. the software list in our Wiki, so one can check this before flashing the image.

Reminder for this.

@MichaIng
Copy link
Owner Author

Jep, on RPi using an unprivileged user for Kodi (via service) should be pretty easy, using video, render and audio groups.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Beta 🧪 Issues specific to the Beta branch testing Information ℹ️
Projects
None yet
8 participants