Monday, April 3, 2017

I hope that this week, I can finally resume my work on Spice and 3D acceleration instead of having to deal with broken hardware.

Continuous integration

One of the things that I want for the various projects I work on is continuous integration on multiple platforms.

Made sur that the recorder project runs correctly on all variants on my local Jenkins instance (i.e. for now Fedora 25/x86_64, Debian/Raspberry, macOS 10.12, Cygwin/WinXP and MinGW/Windows 7, more to come).

This implied a few minor fixes, mostly warnings detected by GCC and warnings related to size on 32-bit machines.

Added a .gitlab-ci.yml file to the project for continuous integration on GitLab. Had a first build pass, yay!

Updated my Windows7 VM image MinGW32 package with basic build tools. This machine will have both MinGW32 and MinGW64. But it's still problematic to have it run well. Have to wait on this one, had to reboot Big on the Linux side, because I realized I had some Spice work that I need to recover from that machine. While I'm at it, I may as well copy the three "ped VMs" to Shuttle.

Shuttle hang

Yet another Shuttle hang during the file copy.

Spent some time looking for BIOS updates, none in sight. Rebooted, installed lm_sensors, have it running in a loop. Temperatures look perfectly normal so far (about 60 degrees). Looked at the various Alt-SysReq commands, will do a kernel dump next time.

[Next time happens] Still another hang while doing nothing more than file copies. The physical screen went black, but the VNC session still shows the last update. Package temperature 55.0, Cores 55,51,50,49, video card 63.5. So definitely not a temperature emergency shutdown.

Something else is happening, but I am a bit at a loss on how to debug that. Short of regularly rebooting the machine, I don't see what I can do. I've swapped memory, video card, everything but the CPU.

Will try to maintain it at maximum temperature doing stupid stuff, to see if I can trigger the problem more rapidly.

Spice CI on GitLab

If I push the spice-gtk module to GitLab, the continuous integration fails:

Complete!
$ ./autogen.sh --enable-static
Submodule 'spice-common' (https://gitlab-ci-token:xxxxxxxxxxxxxxxxxxxx@gitlab.com/c3d/spice-common) registered for path 'spice-common'
Submodule 'src/keycodemapdb' (https://gitlab.com/keycodemap/keycodemapdb.git) registered for path 'src/keycodemapdb'
Cloning into '/builds/c3d/spice-gtk/spice-common'...
Cloning into '/builds/c3d/spice-gtk/src/keycodemapdb'...
error: no such remote ref 4423ea5d2ad5596533dddf83ac1c11660c9e0ca4
Fetched in submodule path 'spice-common', but it did not contain 4423ea5d2ad5596533dddf83ac1c11660c9e0ca4. Direct fetching of that commit failed.
ERROR: Job failed: exit code 1

I think I also need to push the update version of spice-common, it's not taking it from its original server. However, while trying to do that, I got error 500 from the server. Internal server error.

One more Shuttle repair

Went to the shop, bought some thermal paste, applied it liberally on the CPU. Immediately noticed that the fan is no longer spinning at boot. The temperatures look the same as before, but I could place Shuttle on heavy load for about 2 hours without a further hang. I could finally copy all my VM images to it and reboot Big on the Windows side. Yay!

[Update: The next day, Shuttle is still up. The fans are spinning moderately under heavy load. I suspect the issue was thermal imbalance on the CPU, so it did shut off without reporting excessive temperature on its cores. I don't know if that's possible, but what I know is that for now, Shuttle seems to finally stay up.]