On 2018-11-30 13:03:34 [-0500], Gene Heskett wrote: > Now, a progress report, making me think there is more than one bug, the > other is that I've tried 3 of these realtime kernels, and found that all > of them are quite good at throwing away keyboard/mouse events from their > own consoles and its boot sensitive in that if I reboot, I can > eventually reach a configuration that does work, and will work till the > next power bump. The keyboard/mouse is attached via USB, right? So this might be again a problem related to the USB controller. > So I've now found that using the full UUID in my fstab generates a drive > not found error buried in the logs, but that if a fresh > mkswap -c /dev/sda3(takes about half an hour for a 10GB swap) is done, > and the fstab uses PARTUUID="yadda-yadda", that error goes away. This shouldn't take that long. If I remember correctly, it writes the swap header and not much more. > So then I stopped the swapfile, leaving htop showing 9999 megs of swap > and empty, then from another console login, watched the git clone run > with htop. It took most of an hour, got 72 megs into swap, then died > from a bad index compression, just like before but this time complained > about the partition mounted as /media/slash. So I umounted it and ran > e2fsck /dev/sda3, quiet for several minutes, and gave it a clean > condition, no errors IOW. For testing I strongly recommend a local git server and not use kernel.org resources for this. Sebastian
