A git repo for my thoughts

Random findings and notes

Project maintained by paoloemilioselva Hosted on GitHub Pages — Theme by mattgraham

Content

OpenUSD

Compiling under windows

Currently using usd-25.08 at the time of writing this.

There is not that much more from the main docs here https://github.com/PixarAnimationStudios/OpenUSD/blob/release/README.md#windows </br>

What is a bit annoying is finding the right packages to be installed before compiling it, and I think it depends on how much you’ve “messed” up your current environment.

I have two systems, one has been going on for years, and it is currently a frankenstein of libraries which I admittedly have to say I don’t think I can repro it entirely; the other one is a fresh Windows 11 and here is what else is needed to compile OpenUSD via its build_script:

Then launched the build with

python ./build_scripts/build_usd.py --embree --openimageio C:\dev\usd-25.08

You might have an error on some rebuild, mainly in the tests, and you may want to force a BOOST rebuild too adding

--force BOOST

In case you need extra build-args, you can specify them like this for some specific modules

--build-args USD,"-DPXR_BUILD_GPU_SUPPORT=ON"

In case you have Renderman too, the following are useful to also compile the hydra delegate for renderman

--prman --prman-location "C:\Program Files\Pixar\RenderManProServer-26.3"

Executables and env var

When the build is done, it will tell you how to set various ENV VARs.

I haven’t done it as I wanted to build my executable and wanted it to be “portable”, so that anyone can run it without setting any env var.

I’m not an expert in Windows bundle-app, and what I found around is pretty cumbersome.

I wanted something simple, so I’ve decided to split my binary into 2 executables: a launcher and the actual app.

The launcher is basically setting all env vars just the execution of the actual app (kind of like a .bat file, but mostly an actual binary, I never liked .bat files fankly, but I mostly made Linux or MacOsX apps).

The launcher is a simple main(...) (not even a WinAPP), and it just takes the current folder

    TCHAR NPath[MAX_PATH];
    GetCurrentDirectory(MAX_PATH, NPath);

and it uses it to build the paths to pass to the env vars like

    _putenv_s("USDROOT", usdRoot.c_str());
    _putenv_s("PXR_PLUGINPATH_NAME", usdPluginPath.c_str());

it also takes care of “appending” to existing PATH and PYTHONPATH like

    // get vars
    char* origPath = getenv("PATH");
    char* origPythonPath = getenv("PYTHONPATH");
    
    // check for nullptr and append and then

    _putenv_s("PATH", newPath.c_str());
    _putenv_s("PYTHONPATH", newPythonPath.c_str());

In case you need to enable extra env var for some USD features, you can do this too

    _putenv_s("HDGP_INCLUDE_DEFAULT_RESOLVER", "True");
    _putenv_s("USDIMAGINGGL_ENGINE_ENABLE_SCENE_INDEX", "True");

after all of this, it then launches the actual app with

    system("awesomeApp.exe");

or you can also just make it setup a shell to work with USD (like to launch a python-server and do some cool USD web-apps!)

    system("cmd");

RenderSettings, RenderProducts, RenderVars

One of the topics that I worked on for a while, to try to make actual offline-renders with OpenUSD.

What is a RenderSetting prim

soon

What is a RenderProduct prim

soon

What is a RenderVar prim

soon

how it currently works in OpenUSD with Renderman

soon

How to use usdrecord

soob

What you might want to lookout for when implementing your own render-delegate for offline renders

soon

How to use husk from Houdini/Solaris for your offline renders

soon

Python Server

You can create a server running in a shell by simply calling

python -m http.server 8080 --cgi

and you can make python code in files in a sub-folder cgi_bin so that you can access it from a browser via

http://localhost:8080/cgi_bin/hello.py

Using ngrok you can also tunnel it and use a fixed address to access your server.

This may not be the safest way possible to test with some python-code on server-side, but it’s a pretty good starting point.

Once you have generated your unique ngrok identifier, you can start the server in python and then tell ngrok to access your current local server from the unique identifier

ngrok http 8080

You can specify the unique identifier for your website to use a consider link, or use one of the automatically generated links (see output of ngrok command, and help for more options).