Posts by frakkin64

    Raspberry Pi did not pay mesa; they paid a well-known and respected developer shop with tons of mesa experience.

    Yeah that's what the context of "Mesa developers" meant as in someone with Mesa development experience, not specifically the project.

    The problem is consumers seem to either be chasing the cheapest SBC/TV box or the most powerful SBC, and then the complaints start rolling in on every Linux distro about support. I played around enough boards, and pretty much the realization is it's either a Raspbery Pi or a N100 mini PC at this point. Everything else is just too much of a time sink, and giving them any money just continues to encourage this behavior.

    LE is just a Kodi-focused Linux distribution. It seems quite apparent to me that what drives new LE versions is Kodi new versions (not board support, not to say it doesn't happen but the major version cycle is tied to Kodi's).

    I think your expectations are misdirected, you should talk to Orange Pi about why they don't get their board supported by Mesa, Linux, and other projects that get rolled into LE as a distribution. Best way to deal with it is just don't buy these boards that don't make the effort. Raspberry Pi 5 has support because Raspberry Pi paid Mesa, Linux, and other developers to write the software and push it to those same projects.

    I ran into a problem 4 years ago, when I was hosting files via HTTP to Kodi. The bug was in Kodi, it's HTML parser was incomplete, it didn't handle character entities properly. It's possible the issue is with your SFTP server, or in the SFTP vfs component in Kodi, a bit hard to say.

    I can tell you what does work fine, with UTF-8 encoded file names, ampersands, quotes, dashes, etc, etc, etc is NFS. You might want to start with using sftp on your Kodi box (and perhaps your Windows PC or other computers) to see how the filename is presented. SFTP is supposed to be support UTF-8 encoded filenames.

    Are you talking about the box image? It includes the same DTBs in the board specific images, more than likely your eMMC is wiped to the level that there is no boot loader, so using the Khadas/Le Potato image includes u-boot bootloader in the SD image. Amlogic boards will look for a bootloader on the eMMC, and fall back to the SD/USB.

    The box image does not have a bootloader, which is probably why it won't boot, it relies/assumes your eMMC has a working bootloader.

    I am going to support Vero 4K and 4K + for the v21 series at a minimum. See https://osmc.tv/2022/10/vero-4…turing-updates/.

    I figured that would be coming, CE is ending support for S905x devices after 21 (at least that is the latest announcement, I think they were trying to end it with 20), so expecting it is coming to a close eventually on the OSMC side as well.

    That's why part of my interest is in mainline and v4l2 video pipeline (unfortunately that's a slow road for Amlogic), but I boot Armbian as well on the device using the stock bootloader and modified boot scripts (similar to what LE does), just need some postinst triggers on the kernel to build a uImage (since they dropped that). The other part with CE is mainly to get access to bleeding edge Kodi with a fully working video playback. There are options for folks to repurpose the device, but would be nice to get mainline u-boot eventually (and then reformat the partition and install there). Chainloading mainline u-boot works fine, it's all the bl2 and bl3x bits.

    >This, in my opinion, is because they (OSMC) didn't modify the default bootloader script to remove the "wipeisb" & decided to use multiple Android partitions in an LVM group.

    Thanks for the report. This is fixed in this commit:

    Hopefully that will help folks experimenting with the toothpick method from having their eMMC install corrupted, it's fairly easy to recover using the recovery media. It's never been a big deal for me.

    Try a web search on PVR manager is starting up 0%.

    You do understand how search engines work, right? Wow.

    By the way, I don't think you read any of the search results, which I did. The problems varied, ACL problems, missing user problems, connectivity problems. But what really annoys me about you is you exaggerate the problem for attention, then provide no logs, and act like you don't have the time to investigate it. Why are you here then? Is it just to air your grievances for software you did pay for, support you don't pay for? I feel bad for folks like chewitt who try to help users like you. Honestly it's all a bit insane and at this point I just ended up blocking you and not going to waste another second on you or this thread. I hope you figure out your problem, good luck.

    Just me and the many folk who have reported this problem since 2015!

    Where are all of these people? You are trying to describe this as a software problem that affects everyone, yet I am telling you I use the same software and it works fine here. Not only that, I used to use MythTV with LE, and guess what it also worked fine.

    In any case, what your doing isn't going to solve your problem. If you want meaningful feedback then you have to provide debug logs, if you don't want to do that then I really don't see how your going to get any help.

    CNAME should work without a problem. I use DNS to name "services" and then CNAME them to the host, so Kodi uses "kodi--fs1" which points to my server.

    My guess is machine #2 is not using DNS, but probably using mDNS (Avahi). So what I do is (on the server) create mDNS entries as well via a systemd unit avahi-alias. So you might want to make sure your machine #2 is using a FQDN or your DHCP server is pushing the DNS search order. Which will be a problem since the hostname is embedded in the database.

    LE also prefers mDNS over DNS, so you may have to create a systemd unit on your server hosting your file server to create avahi entries. I have a systemd unit called avahi-alias:

    Code
    # cat /etc/systemd/system/[email protected]
    [Unit]
    Description=Publish %i as alias for %H.local via mdns
    [Service]
    Type=simple
    ExecStart=/bin/bash -c "/usr/bin/avahi-publish -a -R %i $(avahi-resolve -4 -n %H.local | cut -f 2)"
    [Install]
    WantedBy=multi-user.target


    It is enabled as

    sudo systemctl enable --now [email protected]

    kodi-fs1 being the mDNS alias, ".local" is the mDNS domain namespace. The simplest solution is already from chewitt, which is change the hostname of your file server. This solution would also require your old file server to be offline, or Avahi not publishing there.

    The other simple solution is creating a hosts entry on the LE device in /storage/.config/hosts.conf. You would have to dig into what's causing the problem here to figure out the appropriate solution, it's probably as simple as using avahi-resolve, nslookup, ping, on the LE device to figure out what's being resolved and what is resolving it.

    frakkin64  vpeter Thanks for the update.

    Much better solution for me, MariaDB 10.5.2 won' install on my server [W2008R2] so I was a bit stuck.

    Sounds like it's getting to the time to upgrade the OS :). I am still waffling myself on moving these things to containers on my Linux server, at least the advantage there is you can run a newer software package on an older OS/kernel, but upgrading Ubuntu has been relatively painless so fortunately it's pretty current.

    KOPRajs Easiest way to find out is try to push a PR and see if they accept it. Some users probably also want some more device mapper modules, like dm-integrity if you care about bit rot and data loss, to be turned on.

    I would also guess there are init scripts changes to assemble the raid at boot. The other catch which I assume hasn't changed is the boot device can't be part of the RAID or it would have to be only in a mirror, that was a GRUB limitation and I have to imagine that u-boot, Raspberry Pi's bootloader, and others have similar limitations. Not sure if folks use eMMC/SD/USB drive as a boot, and then have 2 or more additional devices attached to do some sort of mirror or RAID 5-6.

    Thanks very much for the info.

    My suggestion would be to open a ticket on Kodi's Github or Forum. Not sure if the Kodi developers are aware they created this implicit dependency, it seems unusual to me that they are renaming the column as I thought the way the upgrade typically worked is they created the new schema and then used insert/select statements to migrate the data from the old DB. So I don't quite know why they didn't just create the new schema with the new column name and migrate the data over with an insert/select mapping from old to new column.

    I bet this won't be the last time we hear of this. I didn't run into any problems, but I checked and I am using MariaDB 10.6.

    my hard drive died - all data were lost

    I hear you, just playing devil's advocate, and at the same token you should re-read post #16 as a counter to your argument of no big deal doesn't bother anyone. I can just imagine the users coming in asking how to grow/shrink their array, how to replace a disk, how to expand the array, how to reshape the array (i.e. RAID 5 to 6), hey I did something stupid why doesn't my RAID assemble, etc. Not to mention the feature parity complaints on low-memory footprint devices, have to imagine most of the devices are less than 2G ram and pretty limited I/O bandwidth.

    Frankly, raid isn't going to help you with data loss. RAID is not a backup, there are many cases where a RAID can also result in a complete failure with a single drive failure, it's not unusual to have drive failures in pairs or at the same time or while recovering the RAID. So now your talking a minimum of having 2 or 3 disk redundancy to cover for that, but in reality you really need a backup. You can go wander on the RAID-specific forums on reddit and all they preach about is the 3-2-1 rule.