Environment

• Idris2 Version: 0.7.0-6e52f899b
• OS: Linux (Ubuntu)
• Locale: C.UTF-8 (UTF-8 enabled)
• Backend: Chez Scheme

Description

When compiling an Idris2 project located in a directory path containing non-ASCII Unicode characters (e.g., accented characters like é), Idris2 generates the compileChez script with incorrectly encoded path strings. Instead of properly escaping UTF-8 multi-byte sequences, Idris2 writes raw bytes that Chez Scheme cannot interpret correctly.

Steps to Reproduce

1. Create a directory with a non-ASCII character in its name:

mkdir tpRéférence
cd tpRéférence

2. Create a minimal Idris2 project with main.ipkg file declaring a runmain executable

3. Compile the project: idris2 --build main.ipkg

Expected Behavior

The compilation should succeed and generate a working executable. The generated compileChez script should contain properly escaped Unicode characters compatible with Chez Scheme's string syntax.

Actual Behavior

Compilation fails with:

Exception in compile-program: failed for /path/to/tpRfrence/build/exec/runmain_app/runmain.ss: no such file or directory
Error: INTERNAL ERROR: Chez exited with return code 255

Notice the tpRfrence instead of tpRéférence - the é characters have been corrupted.

Root Cause

Examining the generated build/exec/runmain_app/compileChez file reveals:

(parameterize ([optimize-level 3] [compile-file-message #f]) 
  (compile-program "/home/user/tpR\233f\233rence/build/exec/runmain_app/runmain.ss"))

The byte sequence for é in UTF-8 is 0xC3 0xA9, but Idris2 writes \233 (which is 0xE9, the ISO-8859-1 encoding). This appears to be writing the second byte of the UTF-8 sequence without the first byte, or incorrectly treating UTF-8 bytes as individual characters.

Impact

• Projects cannot be compiled if their path contains any non-ASCII characters
• This affects international users who may have accented characters in their usernames or project paths
• The error message is misleading as it shows a corrupted path rather than indicating an encoding issue

Workaround

Avoid using non-ASCII characters in project paths.