feat: Added Houdini 20.5 conda recipe #76
Conversation
| condaPlatforms: | ||
| - platform: linux-64 | ||
| defaultSubmit: true | ||
| sourceArchiveFilename: houdini-20.5.332-linux_x86_64_gcc11.2.tar.gz |
There was a problem hiding this comment.
I got a failure running ./submit-package-job Houdini-20.5/ without the sourceDownloadInstructions note.
| export "HOUDINI_HOUDINI_PATH=\$HOUDINI_LOCATION/houdini" | ||
| export "HOUDINI_INCLUDE_PATH=\$HOUDINI_LOCATION/toolkit/include" | ||
| export "HOUDINI_LIBRARY_PATH=\$HOUDINI_LOCATION/bin" | ||
| export "HB=\$HOUDINI_LOCATION/dsolib" |
There was a problem hiding this comment.
What's the purpose of this environment variable? Since it's so short and not prefixed by HOUDINI_ like the others, a comment describing how it's used would be helpful.
| export "HOUDINI_INCLUDE_PATH=\$HOUDINI_LOCATION/toolkit/include" | ||
| export "HOUDINI_LIBRARY_PATH=\$HOUDINI_LOCATION/bin" | ||
| export "HB=\$HOUDINI_LOCATION/dsolib" | ||
| export "LD_LIBRARY_PATH=\$HOUDINI_LOCATION/dsolib" |
There was a problem hiding this comment.
For this conda package to interact well in virtual environments that include many other packages, it shouldn't touch the LD_LIBRARY_PATH variable. See https://docs.conda.io/projects/conda-build/en/latest/resources/use-shared-libraries.html for more information about that.
| export "HOUDINI_LIBRARY_PATH=\$HOUDINI_LOCATION/bin" | ||
| export "HB=\$HOUDINI_LOCATION/dsolib" | ||
| export "LD_LIBRARY_PATH=\$HOUDINI_LOCATION/dsolib" | ||
| export "SESI_LMHOST=localhost" |
There was a problem hiding this comment.
Since licensing is a configuration that can change from environment to environment, I think it's better to exclude this variable from the recipe. That way a package built from the recipe won't be specific to one farm.
There was a problem hiding this comment.
I found that using rattler-build instead of conda-build made a big difference in performance. Converting a recipe is not hard, I'll give that a try and let you know the performance difference.
There was a problem hiding this comment.
I found with conda-build, it took 42 minutes, and with rattler-build, it took 9 minutes. That suggests it's better to make it a rattler-build recipe so anyone trying it gets more reasonable build times.
I'll push my adjustments to a branch for you once I've finished doing some tweaks.
There was a problem hiding this comment.
Don't know what rattler-build is Can you explain and show how to modify the code. Unfortunately this package I built like four or five months ago had some of the guys at AWS help me with it so sounds like there's some new better ways.
There was a problem hiding this comment.
Yeah, it's described in a buildTool option that was added since then. See https://github.com/aws-deadline/deadline-cloud-samples/tree/mainline/conda_recipes#the-buildtool-option. I found that rattler-build was consistently faster than conda-build, and for large packages like Maya and now Houdini the difference is significant. See the documentation at https://rattler.build/latest/.
I haven't tested beyond doing the package build, but here's a commit in my branch that converts to rattler-build: mwiebe@af0f503
If you don't have a Deadline Cloud farm for building packages, vs building locally, you might look at https://github.com/aws-deadline/deadline-cloud-samples/tree/mainline/cloudformation/farm_templates/starter_farm that we added recently as well.
| {% set version_minor = "332" %} | ||
| {% set version = version_partial + "." + version_minor %} | ||
|
|
||
| {% set bucket_name = "houdinihive-bucket-deadlinecloud" %} |
There was a problem hiding this comment.
This is using a different convention for how the archive is provided than all the other samples. For the samples to work out of the box, they need this consistency.
| dnf download --resolve -y alsa-lib fontconfig libXScrnSaver libX* libGL libXcomposite libxkbcommon | ||
|
|
||
| # Install python deadline package | ||
| pip install deadline-cloud-for-houdini |
There was a problem hiding this comment.
In all the other samples, we've been packaging this separately. See https://github.com/aws-deadline/deadline-cloud-samples/tree/mainline/conda_recipes/maya-openjd for an example recipe doing this.
There was a problem hiding this comment.
Just the pip install command.
There was a problem hiding this comment.
Yes if I don't do the PIP install then it doesn't seem to work.
I have talked to somebody else several months back in AWS and they weren't sure either but when I installed it it would work.
There was a problem hiding this comment.
That makes sense, there's probably some settings to tweak for it to work as expected. If you look in the submitter code
you can see that it's requesting two packages, houdini and houdini-openjd. The second package is from deadline-cloud-for-houdini just like https://github.com/aws-deadline/deadline-cloud-samples/tree/mainline/conda_recipes/maya-openjd is from deadline-cloud-for-maya.
mwiebe
changed the title
Fixes: #69
What was the problem/requirement? (What/Why)
Wanted to add conda package for Houdini-20.5
What was the solution? (How)
N/A
What is the impact of this change?
N/A
How was this change tested?
Ran the build process ./submit-package-job Houdini-20.5
And submitted jobs from Houdini to the new build ran with no errors
Was this change documented?
This requires no change to the conda build package documentation.
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.