svglite does not work with saving a plot made with {ggblend}

[davidhodge931]

When I try run the following code, it just stalls until my Rstudio crashes

library(tidyverse)
library(palmerpenguins)
library(ggblend)
library(svglite)

f <- fs::path("test", ext = "svg")
svglite("test.png")

penguins |>   
	ggplot() +
	blend(geom_density(aes(x = flipper_length_mm, fill = species)), blend = "multiply")

dev.off()

[davidhodge931]

Realised this relates to #174

[trevorld]

When I try run the following code, it just stalls until my Rstudio crashes

If in the {ggblend} geom function you return a gTree() with a cl class argument and then add a grid::makeContent() S3 method for that class you can check at draw time if the active graphics device supports the features you need and if not throw a more user friendly error (or downgrade to a fallback approach).

• Example of affiner::affineGrob() using a makeContent() method to check for affine transformation support
• Example of oblicubes::geom_oblicubes() combining a grid::makeContent() method with a {ggplot2} geom to make calculations at draw time

However the fact that running your code with {svglite} makes RStudio stall and crash might be a bug. However I speculate this may not be a bug in {svglite} itself but instead may be a bug in base R or a bug in RStudio? In @pmur002's report about the new features he mentions:

> The remainder of the graphics devices in {grDevices} will run, but will (silently) not produce the correct output.

R knows that {svglite} doesn't support the features introduced in R 4.2 (since the device hasn't told it does as confirmed by dev.capabilities()) so I would speculate that when running your code with {svglite} then it should theoretically run but silently not produce the correct output?