Crash with Japanese text

[dmurdoch]

I'm on macOS Sequoia 15.7.2 on an M4, running CRAN R 4.5.2. When I try to run shape_text("\u6771\u4eac") I get a crash. Running in lldb gives this message:

> library(textshaping)
> shape_text("\u6771\u4eac")
Failed to get face: /System/Library/PrivateFrameworks/FontServices.framework/Resources/Reserved/PingFangUI.ttc, 0
Process 36582 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0xc0)
    frame #0: 0x000000011117032c textshaping.so`hb_ft_face_create(ft_face=0x0000000000000000, destroy=(textshaping.so`_hb_ft_face_destroy(void*) at hb-ft.cc:125:3)) at hb-ft.cc:1137:17 [opt]
Target 0: (R) stopped.

The font file listed here "PingFangUI.ttc" is the same one that systemfonts::font_fallback() says I should use, and I get a similar error if I specify it explicitly as path and index values.

Here's session info following library(textshaping) in a new session:

> sessionInfo()
R version 4.5.2 (2025-10-31)
Platform: aarch64-apple-darwin20
Running under: macOS Sequoia 15.7.2

Matrix products: default
BLAS:   /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib 
LAPACK: /Library/Frameworks/R.framework/Versions/4.5-arm64/Resources/lib/libRlapack.dylib;  LAPACK version 3.12.1

locale:
[1] en_CA.UTF-8/en_CA.UTF-8/en_CA.UTF-8/C/en_CA.UTF-8/en_CA.UTF-8

time zone: America/Toronto
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] textshaping_1.0.4

loaded via a namespace (and not attached):
[1] compiler_4.5.2    cli_3.6.5         lifecycle_1.0.4   systemfonts_1.3.1
[5] rlang_1.1.6   

[dmurdoch]

I don't know why the font was not found, but the problem appears to be around here:

textshaping/src/string_shape.cpp

Line 894 in 36143b3

FT_Face face = get_font_sizing(fallbacks[font], shape_info.size, shape_info.res, fallback_sizes, fallback_scales);

The function is returning a null pointer, but no check is done, and it is used as if it is a valid hb_font_t pointer.

[dmurdoch]

The deep underlying cause here is that Apple created a new undocumented format for that file: see https://developer.apple.com/forums/thread/758189 . Apparently Harfbuzz can read it, but probably only in versions 11.3.0 or later. I'm using 8.5.0 (from mac.r-project.org/bin). I don't know how to find out what version is used in the textshaping CRAN binary, but I'd guess it's the same one. You'd need different code for this anyway.

I don't know if Freetype has been updated to read it or not. I'm using 2.13.2, from the same source. I think based on the timing you'd need at least 2.14.0 for this to be fixed.

[thomasp85]

Thanks for the additional context - I did find that PingFangUI was broken on newer Mac systems but still need to find a good way to opt out of it. systemfonts rely on the macOS font substitution so I'm not directly in control over which font gets returned. I hope to have a fix soon though

[dmurdoch]

I've been working directly with Pango/Cairo & Harfbuzz, and that combination doesn't try to read the PingFangUI file. So maybe looking at what they do for font substitution would be helpful.

[thomasp85]

Should be fixed on latest version of systemfonts

[dmurdoch]

Thanks, it works for me now.