Bugfix for job cancellation (reported by Carsten in email)
Ended up recoding part of the communication pipe (use json more consistently)
Fix bpy data modifications where it shouldn't happen (as a bonus, thumbnailing is now done out of process)
[#24601] Net rendering master node fails to send/retrieve files to/from slaves
Both Reported by Dan McGrath
The second might not actually be fixed with this, but now there is better error checking and reporting.
Misc: add poll method in Panel base class, to better support mixins with different inheritance order.
New Feature:
VCS job type
Render a file (with dependencies) from a version control system (currently only supports subversion, but system is already generic).
On client, working path, server path and current revision can be guessed from data on disk or entered manually.
On slave, a working copy is created (if needed) where specified by the job and updated to the proper revision.
On master web page, job types now appear in the job lists. The job page shows the list of dependencies for "normal" jobs or versioning information for VCS jobs.
Limitations: Need to have command line tools "svn" and "svnversion". Working copy path must be the same on client and slaves (the client gets the job path relative to the working copy). When guessing, working copy path is set to the folder where the current file is (this can be changed manually after). On the slave, it will update the working copy AS SPECIFIED to the revision, so if that path is too deep, some dependencies will not be updated properly. Doesn't support mixed revisions (and will not give any warnings for that), it will always use the first revision specified by "svnversion"
Bugfix:
Thumbnail generation doesn't chew down memory anymore and always gives correct result (thumbnail on master especially could mess up between jobs with the name result filename)
Users can set their machine name to something containing non-ascii characters. In Python this currently causes problem due to
socket.gethostname() throwing UnicodeEncodingError. Work around this by not using platform.system() (which uses internally socket.gethostname()).
See http://www.pasteall.org/16215 for backtrace
netrender client's functions to get slaves and jobs info didn't work anymore (was reading response content twice).
At the same time, I switched to json for the dump and load, instead of using eval and repr (for obvious security reasons). I should have done this much earlier.
slave use output log option correctly
use os.path.join more consistently (reported by Dalai Felinto)
unify frame range (was 4 digits on master, 6 elsewhere, now 6 everywhere).
Fix occasional broken pipe between master and slave (didn't affect result, just polluted log).
- Temporary fix for linked libs repathing (didn't work correctly when using relative links).
- autorefresh only for main page (on job page, it's really annoying because it keeps reseting results preview, so until that's fixed, off it is).
Repath functionality: Edit paths in blend file on slaves to match the transfered files.
Tested with textures and point caches. Fluid should work. Point Caches with the External option might not.
This should fix previous problems when using absolute paths in blend files.
- multires cache files and image .tex cache support in dependency list
- Compare md5 of files before using a local copy (not one transfered by netrender). Could be changed to a simpler CRC if speed is an issue. The goal is not to have a strong crypto signature but just to detect outdated local files.
- Reduce slave timeout to 5 minutes (down from 30). Slaves should report at most every 30s, there's no reason for a value to be that high.
- Reorder the presentation tables on the main web page (job list is more important)
- Collapse dependency list by default on job page (only show main file and headers for other files, point cache and fluid cache)
- Slave option (default: True) to also output render log to the console (as well as the usual copy to the master)
* Client connection is made from panel poll functions even if
netrender is disabled
* Network render slave should print when it has connected to server,
now it says "node initiation", I initially thought this meant it was
not connected yet.
* Web interface should auto-refresh, I got confused if job and slaves
were actually found or not.
Slave and Master options to delete files when closed (default True for slave, False for Master)
Web interface option to remove files (on master) when deleting a job (or all jobs)
Web interface button to pause a job
- Patch by Olivier Amrein (prettying the web interface with css, fixing some notification bugs)
- More error checks on the slave (better behavior when job is canceled)
- Client: when using "Animate on Network" and canceling render midway, also cancel job. Use Send Job and Animate if you want a real background job.
use a real object for files instead of a tuple
unique urls for files, logs and render results (just missing the proper mime type for exr files)
fix bug with slaves not getting the correct machine name