Mime.detect treats truncated multi-byte UTF-8 as text/plain #56183

Activity

1. opencode-agent commented on Oct 16, 2026

   

   opencode-agentBot

   on Oct 16, 2026 – with OpenCode Agent

   Contributor

   More actions

   Thanks for the report. I could reproduce this: on v2.0.31 and on the latest dev (0a03b4b), Mime.detect returns text/plain for [0x61, 0x62, 0xC3] and [0x61, 0xE2, 0x82]. Invalid bytes in the middle of the buffer ([0xC3, 0x41, 0x42]) are correctly reported as application/octet-stream. The script below shows this.

   One thing to consider for #55946: the read tool calls Mime.detect only on the first 256 KiB of a file (FIRST_CHUNK in read-filesystem.ts). That cut can land in the middle of a valid multi-byte character, so { stream: true } may be deliberate there. Always flushing would make some valid UTF-8 files look binary. The fix may need to flush only when the bytes are the whole file, or allow an incomplete character only at the very end of the buffer.

   Reproduction script (mime-truncated.ts)

   // Usage: bun /workspace/repro/mime-truncated.ts <path to packages/core/src/mime.ts>
   const { detect } = await import(process.argv[2] ?? "/workspace/repo/packages/core/src/mime.ts")
   const cases: [string, number[]][] = [
     ["valid 'abc'", [0x61, 0x62, 0x63]],
     ["complete 'é' (C3 A9)", [0x61, 0xc3, 0xa9]],
     ["truncated 2-byte (trailing C3)", [0x61, 0x62, 0xc3]],
     ["truncated 3-byte (trailing E2 82)", [0x61, 0xe2, 0x82]],
     ["invalid mid-buffer (C3 41)", [0xc3, 0x41, 0x42]],
   ]
   for (const [name, b] of cases) console.log(name.padEnd(36), detect(new Uint8Array(b)))

   Run it: bun mime-truncated.ts
2. opencode-agent

   added

   reproduced

   on Oct 16, 2026