• Editor
  • Continuous integration

Hi!
Spine has command line mode for export and documentation says that it is intended for use with scripts and build servers.
We tried two different approaches to setup our build server:
1) Headless linux VM with Xvfb. At first we got LWJGL exception because Xvfb doesn't support xrandr, but manual rebuild with xrandr patch solve only this problem and we got another one.
2) Real Windows server with separate videocard. We can run Spine export only on the machine itself, but it throws error "OpenGL 2.0 Unavailable" if we try to run it via jenkins or rdp.
So the question is: Is there any success story about Spine running on build server? I tried to find it on google and this forum, but I didn't find anything.
Thank you in advance.

Related Discussions
...
  • 編集済み

We have the same problem with our build environment.

We commit the exported json files in subversion as part of our build process but this is not a viable solution!

Further issues with the command line export is needing to run the editor to update.

23日 後

We solved this issue by using physical Windows build server with always active VNC-session for jenkins user. This way Spine export works as expected.

We have a solution for headless export, but only for JSON and binary export since image and video export requires OpenGL. The only remaining task is to decide how to package it. Likely the launcher will need a headless mode so it can be used as the entry point to headless export. I imagine the first time it is run it will prompt for an activation code (at the command line) and download the appropriate Spine update which will contain the actual headless export functionality. This way headless export can be kept up to date. The launcher can be locked to a specific Spine version the usual way.

I've added a Trello task for it:
https://trello.com/c/lttUDH6M/151-headless-json-binary-export-for-headless-build-servers