dgilmore changed the topic of #fedora-riscv to: Fedora on RISC-V https://fedoraproject.org/wiki/Architectures/RISC-V || Logs: https://libera.irclog.whitequark.org/fedora-riscv || Alt Arch discussions are welcome in #fedora-alt-arches
<davidlt[m]> 5:57 AM. I am late today :)
<davidlt[m]> Lack of sleep is catching me already :)
<davidlt[m]> 🧟‍♂️
<thefossguy> Haha, I recommend a sleep schedule of 21:30/22:00 to 04:00/05:00
<thefossguy> It has been working great for me
<davidlt[m]> Yeah, something like is what I do.
<davidlt[m]> The sun is up at ~5:30 AM. It's bright outside before 5 AM already.
<davidlt[m]> It feels I am aligned to that :)
<thefossguy> XD
<davidlt[m]> rwmjones: another thing, I managed to unblock antlr* packages too.
<davidlt[m]> Golang is in a better shape, too. I managed to get into a state where I am good to submit 250 of them to build and see what happens.
<thefossguy> cool :)
<davidlt[m]> There aren't that many golang packages that we didn't have built.
<davidlt[m]> 250 from <3K ain't many.
<davidlt[m]> I will scan Koji logs from the last 1-2 months and check if any failed build tried to pull golang packages and submit those too.
<davidlt[m]> Majority of golang* packages so far are green. antr* packages are also good.
<davidlt[m]> I cannot believe I got a break from mass rebuild, and my brain refreshed enough to notice how to "hack" these two long-term blockers.
<davidlt[m]> Pulling logs for FAILED tasks in the last 60 days now.
<davidlt[m]> Btw, Nelson Chu (Rivos) patches TEXTREL issue in binutils. I plan to start testing the patch most likely next week.
<davidlt[m]> As this golang +antlr stuff will take some time to pass through pipeline.
<davidlt[m]> I might start sooner (like today) as I have one board still disconnected.
<davidlt[m]> djdelorie: any news about your board?
<davidlt[m]> Ok, let's increase that log window to 180 days.
<davidlt[m]> rwmjones: are you using both of your boards? I see those are disconnected from Koji.
<davidlt[m]> 47 packages built.
<davidlt[m]> Another small cycle, easy to fix.
<rwmjones> davidlt[m]: morning, reading back
<rwmjones> awesome news about antlr & golang
<rwmjones> what is the TEXTREL issue in binutils?
<rwmjones> davidlt[m]: my boards should be connected ... let me check
<davidlt[m]> Binaries have TEXTREL bit set in the headers by binutils, but no TEXTREL relocations are actually generated.
<davidlt[m]> SELinux will block such binaries as .text is RO.
<rwmjones> nujive seems fine
<davidlt[m]> If TEXTRELs exist dynamic loader needs to make .text RW for a short period of time to handle relocs.
<rwmjones> nufive has a python error in the kojid logs, let me restart kojid ...
<davidlt[m]> SELinux will detect that and kill it.
<rwmjones> both boards *should* be up now
<davidlt[m]> Still waiting for nujive ping.
<rwmjones> ok I have selinux disabled so not seen that ..
<davidlt[m]> It got worse with Fedora 38 (and binutils 2.40 probably).
<rwmjones> I'm going to restart nujive kojid too, as there is nothing running
<davidlt[m]> Things like systemd-logind, NetworkManager, etc. got affected that didn't allow proper booting.
<rwmjones> shouldn't need it, but who knows ..
<rwmjones> ok restarted
<davidlt[m]> rwmjones: any news on pandoc?
<davidlt[m]> Well, GHC issue.
<rwmjones> davidlt[m]: yeah the build failed because of missing ld.gold
<davidlt[m]> Ah you didn't build our SRPM.
<rwmjones> I did ..
<rwmjones> with a few modifications
<davidlt[m]> That disabled ld.gold
<rwmjones> oh there's another one?
<rwmjones> I'm using the fedora one
<rwmjones> tbh I've been in meetings all morning and I've got another one in 5 mins
<rwmjones> that being the traditional way big companies use to ensure no actual work gets done ...
<rwmjones> IIRC you did mention another git repo, but I've not looked at it & forgotten the location now
zsun has joined #fedora-riscv
<davidlt[m]> rwmjones: it's in koji build information
<davidlt[m]> 101 package built.
<davidlt[m]> alexsaezm: new golang is incoming to F38?
<davidlt[m]> 1.20.4, I think.
<alexsaezm> Yep
<alexsaezm> I’ll finish the builds today
<davidlt[m]> Nice, I will get that building ASAP.
<davidlt[m]> As soon as I see that green on Koji :)
zsun has quit [Quit: Leaving.]
zsun has joined #fedora-riscv
zsun has quit [Quit: Leaving.]
zsun has joined #fedora-riscv
zsun has quit [Client Quit]
<davidlt[m]> 123 packages built.
<djdelorie> davidlt[m]: not yet. I'll go fetch it now and poke at it.
<davidlt[m]> djdelorie: fix it, and I will send a GCC build your way 😉
<djdelorie> ok, I removed the video card and set it up in my office with serial ports. he cpu fan sounds horrible and the eht0 port doesn't show up
<djdelorie> there are some bios errors at boot, I'll capture them and put them somewhere
<davidlt[m]> That doesn't sound good. Hopefully it's still alive :)
<davidlt[m]> So it boots!
<djdelorie> it does. Maybe the video card is bad (I removed it when I brought it to my office)
<davidlt[m]> Kick NetworkManager, and see what happens.
<djdelorie> with no video card and no ethernet, it would have appeared dead
<davidlt[m]> It started Network Manager, and seems it not networking working.
<djdelorie> yeah, ip can see eth0 but can't start it
<davidlt[m]> Thus I would look into Network Manager logs, and I would restart it.
<davidlt[m]> See systemd NetworkManager.service journal.
<davidlt[m]> and just restart the service.
<djdelorie> device (eth0): state change: unmanaged -> unavailable (reason 'managed', sys-iface-state: 'external')
<davidlt[m]> Really? It think that something else is setting up the network and bails out.
<davidlt[m]> I have seen it before on FUSDK IIRC.
<djdelorie> would it also give that error if the device was broken?
<davidlt[m]> No IIRC.
<djdelorie> [root@unmatched ~]# nmcli d
<djdelorie> DEVICE TYPE STATE CONNECTION
<djdelorie> eth0 ethernet unavailable --
<djdelorie> lo loopback unmanaged --
<davidlt[m]> Let me check FUSDK commits.
<davidlt[m]> Is systemd resolved installed?
<djdelorie> Error: Connection activation failed: No suitable device found for this connection (device eth0 not available because device has no carrier).
<djdelorie> I know the cable is good because I borrowed it from my printer
<djdelorie> and it's the second I tried
<djdelorie> heh, I have a pcie ethernet card from the hifive...
<davidlt[m]> Could you try to launch NM TUI gui and force connect on eth0?
<djdelorie> that was the error I posted
<davidlt[m]> nmcli conn show --active
<davidlt[m]> nmcli conn down 'Wired connection 1' && nmcli conn up 'Wired connection 1'
<davidlt[m]> Something like that.
<djdelorie> ok, this hurts - I plug in the ethernet pcie card, *its* carrier light light up, but the system doesn't see the pcie card at all
<davidlt[m]> Up, ah.
<djdelorie> so it smells like a dead pcie slot (i.e. no video) and a dead eth0 port (no carrier)
<davidlt[m]> Alternatively idea would be to try a newer disk image, just in case data is somewhere corrupted.
<djdelorie> he microsdcard?
<davidlt[m]> Is USB working?
<davidlt[m]> Pretty much everything is connected via PCIe expect Ethernet.
<davidlt[m]> Is NVMe working? I think you are booting with it.
<djdelorie> usb seems to work, as does nvme
<davidlt[m]> Test all the bits that should work, all USBs, microSD (are you using it?
<davidlt[m]> So PCIe works.
<djdelorie> microsd works (at least, I can fdisk it)
<davidlt[m]> It's all-in-one (firmware, /boot, rootfs).
<djdelorie> I put all my sdcards in one box, so now I can lose them all at once. Efficiency!
<djdelorie> found it
<djdelorie> the new image sees the pcie card
<djdelorie> still no eth0 but work-aroundable
<djdelorie> new image did not fix the cpu fan ;-P
<davidlt[m]> What do you mean? :)
<djdelorie> I mean pcie bus is not hosed (good) but eth0 probably is (bad)
<davidlt[m]> Cheap USB Ethernet dongle/
<djdelorie> pcie ethernetcard is working <- eth1 is up, eth0 is down
<djdelorie> "just" need to replace the cpu fan and it should be usable
<djdelorie> but I need to rebuild the old mmc card with the new image so it boots *and* uses the nvme
<djdelorie> oh wait, it is usuing the nvme... smart card :-)
<davidlt[m]> Nah,
<djdelorie> it appears to be doing a koji build already
<davidlt[m]> Ah, shoot, it already booted into NVMe?
<davidlt[m]> NVMe has a higher boot priority in U-Boot IIRC.
<davidlt[m]> You are right, Koji shows it's active.
<djdelorie> killed. I want to replace the cpu fan first
<davidlt[m]> The more proper way would be to flash SPI-NOR flash, change DIP switch settings, flash NVme with a new image.
<davidlt[m]> Configure new rootfs for Koji.
<davidlt[m]> and then it will not use microSD card anymore.
<djdelorie> I'm ok with it using microsd. I just don't want to use the 128G temporary card I put the new image on
<djdelorie> I'll put it on the original 32G one next, now that I know it works
<davidlt[m]> only contains firmware GPT partitions to be used for microSD card (only).
<djdelorie> even better :-)
<djdelorie> oh look, I can just 3D print a fan adapter so I don't have to buy a fan...
<davidlt[m]> You can use traditional PC fan (3-pin, 2-pin) all works.
<davidlt[m]> I have Noctua pushing air from the side instead top-down.
<davidlt[m]> 120mm
g-tenko-r[m] has joined #fedora-riscv
<djdelorie> oh crap, it's a 12v fan, isn't it? sigh.
<djdelorie> I'm sure I can find 5v somewhere... hack, hack hack ;-)
<davidlt[m]> Yes
<palmer> IIRC you want to make sure there's some airflow over the PCIe switch, otherwise it gets very hot
<djdelorie> yeah, plan B is to put an 80mm in the back case and print a duct that hits both of them
fuwei has quit [Ping timeout: 240 seconds]
fuwei has joined #fedora-riscv
iooi has joined #fedora-riscv