Fix headless screenshot rendering (partially)

[bwasty]

Should solve #12.

Current state: Works fine on macOS, on Linux the resulting image is always black.

TODOs:

• [ ] Fix Linux (black/transparent image - issue in glutin?) (postponed)
• [ ] Test on Windows (postponed)
• If not working on all platforms: add CLI switch to choose between 'real' headless and hidden window mode

[siddhartpai]

I tried pulling in this branch and taking a screenshot.
But i just ended up getting a black screen.

[bwasty]

Hmm - what OS? And do you mean 'screen' or the resulting PNG image?

[siddhartpai]

I'm using Arch Linux ,
On running the below command
./target/debug/gltf-viewer ./gltf/models/DamagedHelmet.gltf --screenshot a.png

a.png is just a black image.

[siddhartpai]

[ec2-user@ip-172-31-23-71 gltf-viewer]$ RUST_BACKTRACE=1 ./target/debug/gltf-viewer ./models/DamagedHelmet.gltf --screenshot a.png thread 'main' panicked at 'called Result::unwrap()on anErrvalue: NoBackendAvailable(LoadingError("LibraryNotFound { descr: \"libOSMesa.so: cannot open shared object file: No such file or directory\" }"))', /checkout/src/libcore/result.rs:906:4 stack backtrace: 0: std::sys::imp::backtrace::tracing::imp::unwind_backtrace at /checkout/src/libstd/sys/unix/backtrace/tracing/gcc_s.rs:49 1: std::sys_common::backtrace::_print at /checkout/src/libstd/sys_common/backtrace.rs:68 2: std::panicking::default_hook::{{closure}} at /checkout/src/libstd/sys_common/backtrace.rs:57 at /checkout/src/libstd/panicking.rs:381 3: std::panicking::default_hook at /checkout/src/libstd/panicking.rs:397 4: std::panicking::rust_panic_with_hook at /checkout/src/libstd/panicking.rs:577 5: std::panicking::begin_panic at /checkout/src/libstd/panicking.rs:538 6: std::panicking::begin_panic_fmt at /checkout/src/libstd/panicking.rs:522 7: rust_begin_unwind at /checkout/src/libstd/panicking.rs:498 8: core::panicking::panic_fmt at /checkout/src/libcore/panicking.rs:71 9: core::result::unwrap_failed at /checkout/src/libcore/macros.rs:23 10: <core::result::Result<T, E>>::unwrap at /checkout/src/libcore/result.rs:772 11: gltf_viewer::viewer::GltfViewer::new at src/viewer.rs:64 12: gltf_viewer::main at src/main.rs:87 13: __rust_maybe_catch_panic at /checkout/src/libpanic_unwind/lib.rs:101 14: std::rt::lang_start at /checkout/src/libstd/panicking.rs:459 at /checkout/src/libstd/panic.rs:365 at /checkout/src/libstd/rt.rs:58 15: main 16: __libc_start_main 17: <unknown>

This is what happens when i run it on an RHEL amazon aws instance which is where i'm trying to do headless rendering .

[bwasty]

I did some tests - on macOS everything works fine, on Ubuntu (16.04 Desktop) I first got the same error about libOSMesa.so. Fixed by installing libosmesa6-dev (apt-get - I don't know RHEL, but I guess there's an equivalent package).
But then I also just got black images. I tried increasing log level to max (-vvv) and adding another gl_check_error!() call after the screenshot, but it's no help. Not sure what else to try (perhaps open an issue for glutin?).

Alternative approach I've successfully used on a different project you might try to adapt to RHEL:

$ sudo apt-get install xvfb # depends on mesa, so it doesn't have to be installed explicitly
$ xvfb-run --auto-servernum --server-args="-screen 0 640x480x24" \
    cargo run -- Box.glb --screenshot b.png

This should work with the current 'faux' headless mode.

Perhaps I'll add a switch to choose between 'real' and 'faux' headless mode.

[siddhartpai]

I saw a similar issue opened on glutin,

rust-windowing/glutin#946

I can try the xfvb but i need this to work on an AWS Lambda; but Lambda doesn't have xfvb preinstalled . I'll see what i can do . Thank you for all your help.

[bwasty]

You definitely need mesa even if you can get around xvfb, so I'm not sure it can work on AWS Lambda if it's not already installed (found this related question).

[siddhartpai]

xvfb-run --auto-servernum --server-args="-screen 0 640x480x24" -e logs.log target/debug/gltf-viewer models/DamagedHelmet.gltf --screenshot a.png -w 640 -h 480
thread 'main' panicked at 'No backend is available', /home/ec2-user/.cargo/registry/src/github.com-1ecc6299db9ec823/winit-0.10.1/src/platform/linux/mod.rs:340:8
note: Run with RUST_BACKTRACE=1 for a backtrace.

Even xvfb has issues with the faux headless mode . I'll give this a few more go's

[bwasty]

Today I had some time do to my own experiments on AWS (Ubuntu 16.04 instance) - I fixed the above error by installing xorg.

Then I started to put everything in Docker -> develop...headless_docker
Needs some cleanup still, but works - like this:
./screenshot_docker.sh ../gltf/glTF-Sample-Models/2.0/Box/glTF-Binary/Box.glb (only glb at the moment)

[siddhartpai]

I'll have a look at this @bwasty. Thank you .

[bwasty]

Some notes from further experiments:

• changed screenshot from RGB to RGBA - screenshots are still empty but now transparent instead of black
• after explicitly specifying OpenGL version + profile (3.3), I now get an error that might help with further debugging:

thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: OsError("OSMesaCreateContextAttribs failed")', /checkout/src/libcore/result.rs:906:4
note: Run with `RUST_BACKTRACE=1` for a backtrace.

(from here)

Next step: generate more debug output with these queries: https://www.khronos.org/opengl/wiki/OpenGL_Context#Context_information_queries

[bwasty]

I've decided to merge this as it is since I've made quite some progress on the docker branch.