rgl not displaying plots in XQuartz on Mac OS Tahoe

[jdpampush-netizen]

Hi Duncan, I have run into an issue and looking for some general advice. On the latest mac OS, Tahoe 26.0.1. It seems the informative error is during the rgl install. Throwing: X11 error:

GLXBadContext
'rgl.init' failed, will use the null device.

So rgl wont produce xquartz windows when plotting on a mac. The rglwidget() function is a useful workaround allowing visualization in a web browser, but the background prints which include legends are not printing to the web browser.

So I am wondering, since this is a known issue, will a patch become available so that us mac users can keep using XQuartz with rgl or should I start modifying functions to be sure to print to the background of web browsers with rglwidget(). Any thoughts would be appreciated. Thanks.

[dmurdoch]

I am out of town now, without my computer, so it will be a few days before I get to this. But general advice is: reinstall xquartz after any macos version change. Do not use the homebrew builds of r or rgl, use the binaries from cran.

…

On Wed, Oct 1, 2025, 23:55 jdpampush-netizen ***@***.***> wrote: *jdpampush-netizen* created an issue (dmurdoch/rgl#488) <#488> Hi Duncan, I have run into an issue and looking for some general advice. On the latest mac OS, Tahoe 26.0.1. It seems the informative error is during the rgl install. Throwing: X11 error: GLXBadContext 'rgl.init' failed, will use the null device. So rgl wont produce xquartz windows when plotting on a mac. The rglwidget() function is a useful workaround allowing visualization in a web browser, but the background prints which include legends are not printing to the web browser. So I am wondering, since this is a known issue, will a patch become available so that us mac users can keep using XQuartz with rgl or should I start modifying functions to be sure to print to the background of web browsers with rglwidget(). Any thoughts would be appreciated. Thanks. — Reply to this email directly, view it on GitHub <#488>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOYSQDO6T7W3OJZW6ZZ5Y33VRETTAVCNFSM6AAAAACIBR3YROVHI2DSMVQWIX3LMV43ASLTON2WKOZTGQ3TKMRWG44DENI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[dmurdoch]

I am back home now, with some questions.

• You say this is "a known issue". What makes you say that? Have you read about it elsewhere?
• Did you try my general advice? What happened?

[jason-muhlbauer]

I have also had the same issue running rgl on 26.0.1. I tried reinstalling XQuartz but it did not address the issue. Below is the exact warning message:

1: In rgl.init(initValue, onlyNULL) : X11 error: GLXBadContext
2: 'rgl.init' failed, will use the null device.
See '?rgl.useNULL' for ways to avoid this warning.

[jimjam-slam]

I've also been getting this error since I upgraded to Tahoe. I've reinstalled XQuartz, rebooted the system and installed rgl, but no luck. I'm on:

• rgl 1.3.24
• R 4.5.1
• macOS 26.0.1

[s-nicolas]

Hello,

Same for me, since Tahoe, impossible to une rgl expect from widget with the In rgl.init(initValue, onlyNULL) : X11 error: GLXBadContext.

rgl 1.3.24
R 4.5.2
macOS 26.1

[dmurdoch]

I don't have Tahoe yet, so I can't debug this. If someone who is seeing it can put together a PR, I'll take a look.

[dmurdoch]

Reading this thread: https://discussions.apple.com/thread/256157657?sortBy=rank, it appears that Tahoe has made changes that are incompatible with XQuartz, and the XQuartz project has no activity, so no fix is expected there. It's possible that rgl's X11 display will not be able to be made compatible with Tahoe.

[s-nicolas]

@dmurdoch also XQuartz/XQuartz#438

[dmurdoch]

The only workaround I can suggest for now is setting these options before running rgl:

options(rgl.useNULL = TRUE, rgl.printRglwidget = TRUE)

These will get rgl to display in a web browser instead of trying to use X11. The display is not identical, but generally it's acceptable, sometimes better than the X11 display. If you find examples where it doesn't work, please let me know.

[jason-muhlbauer]

The only workaround I can suggest for now is setting these options before running rgl:

options(rgl.useNULL = TRUE, rgl.printRglwidget = TRUE)

These will get rgl to display in a web browser instead of trying to use X11. The display is not identical, but generally it's acceptable, sometimes better than the X11 display. If you find examples where it doesn't work, please let me know.

This worked great for my application, thank you!

[jimjam-slam]

The only workaround I can suggest for now is setting these options before running rgl:

options(rgl.useNULL = TRUE, rgl.printRglwidget = TRUE)

These will get rgl to display in a web browser instead of trying to use X11. The display is not identical, but generally it's acceptable, sometimes better than the X11 display. If you find examples where it doesn't work, please let me know.

I'm hitting this issue when using rgl via the rayshader package. When I use the above workaround, I get:

Warning in rgl.snapshot(filename, fmt, top): this build of rgl does not support
snapshots

(And the video — which I'm assuming rayshader compiles from a series of static frames — doesn't render.)

Does rgl need to be installed from source with a certain configuration to enable snapshots? I probably have the CRAN version.

EDIT: looking at the source of the error, it seems like this occurs when OpenGL isn't available — so maybe not something I can work around (I can see you just described the problem on the README as OpenGL not being available to Quartz on Tahoe, @dmurdoch). Thank you for working through this nonetheless! I'll also open up an issue with rayshader to see if there're any workarounds at that level.

[tylermorganwall]

To solve this issue inside of rayshader, you can set software_render = TRUE which will call a software renderer that doesn't require OpenGL and will pull all the geometry as it is supposed to be rendered from rgl when rgl.useNULL = TRUE.

[jimjam-slam]

Thanks very much for the tip, @tylermorganwall! I'm using rayshader::render_movie() rather than rayshader::render_snapshot() — is the software_render option passed through there too? (We can render out an image series if necessary, but ideally we'd keep the Closeread demo we had in mind here as simple as possible.)

[dmurdoch]

Yes, rgl.snapshot requires the internal display. snapshot3d(webshot = TRUE) doesn't, but it is a lot slower. I don't know the details of Tyler's software rendering, but it could well be a lot faster than the Webshot rendering.