.ISPC files arent rendered as c++, unlike cuda files

I traced this through the current TUI highlighting path.

diff_render.rs already extracts the file extension and passes it through the existing syntax-highlighting path, so this does not appear to require any diff-rendering changes.

The missing piece is find_syntax in codex-rs/tui/src/render/highlight.rs. two-face does not resolve ispc/isph, while Codex already has a small fallback alias table there for unsupported C++-like extensions such as cu/cuh and cppm/cxxm/ixx.

The minimal change appears to be:

• map ispc | isph to cpp in find_syntax
• extend find_syntax_resolves_languages_and_aliases
• extend the existing cpp_module_extensions_use_cpp_highlighting rendered-diff test and its insta snapshot

This keeps the existing unknown-extension fallback unchanged and follows the same implementation/testing pattern as the CUDA change in #34570.

Happy to submit the focused patch if this matches the intended direction and you'd like an external PR.

I traced this through the current TUI highlighting path. diff_render.rs already extracts the file extension and passes it through the existing syntax-highlighting path, so this does not appear to require any diff-rendering changes. The missing piece is find_syntax in codex-rs/tui/src/render/highlight.rs. two-face does not resolve ispc/isph, while Codex already has a small fallback alias table there for unsupported C++-like extensions such as cu/cuh and cppm/cxxm/ixx. The minimal change appears to be: map ispc | isph to cpp in find_syntax extend find_syntax_resolves_languages_and_aliases * extend the existing cpp_module_extensions_use_cpp_highlighting rendered-diff test and its insta snapshot This keeps the existing unknown-extension fallback unchanged and follows the same implementation/testing pattern as the CUDA change in #34570. Happy to submit the focused patch if this matches the intended direction and you'd like an external PR.

Its a small enhancement - the use ispc or isph is rare. but is a legitmate format as its one of the bigger compilers

sounds good, i’ll put together the focused patch.

For anyone picking this up: it's a one-line change in the same alias table the CUDA commit touched. find_syntax patches extensions two-face can't resolve (https://github.com/openai/codex/blob/1f41cc5d92/codex-rs/tui/src/render/highlight.rs#L541-L550), so ISPC just needs "ispc" | "isph" => "cpp", next to the existing "cu" | "cuh" => "cpp", line. Other C-family candidates with no dedicated syntect definition (.metal, .cl/OpenCL, .hip) could join the same arm; anything two-face already knows should be left out so it keeps its proper grammar.

For anyone picking this up: it's a one-line change in the same alias table the CUDA commit touched. find_syntax patches extensions two-face can't resolve (https://github.com/openai/codex/blob/1f41cc5d92/codex-rs/tui/src/render/highlight.rs#L541-L550), so ISPC just needs "ispc" | "isph" => "cpp", next to the existing "cu" | "cuh" => "cpp", line. Other C-family candidates with no dedicated syntect definition (.metal, .cl/OpenCL, .hip) could join the same arm; anything two-face already knows should be left out so it keeps its proper grammar.

yeah its a bit painful compiling a huge codebase like Codex - but still - provided another commit by @etraut-openai , on how the codex repo likely want for the maintainers for quicker patch

as for using cargo and fetching all of the whatever seems a bit painful too

fairpoint .cl .hip .metal should be supported?
ispc? is pretty much c

maybe its fair that @ttaylorr-oai would push this commit?

<img width="887" height="462" alt="Image" src="https://github.com/user-attachments/assets/0659abcc-3a3f-45d2-8eaa-733aa60e35b8" />