Re: [Bug: 21.5-b24] Problems with coding systems autodetect

Joachim Schrod wrote: I have some additional information, since I learned about debug-coding-detection in the mean time. Turning it on yields the following output on stderr: detected coding system: nil detect_coding_type: processing 88 bytes First 16: .Mastab whlt u 09 4D 61 DF 73 74 61 62 20 77 E4 68 6C 74 20 75 Last 16: e Fachkonzept.). 65 20 46 61 63 68 6B 6F 6E 7A 65 70 74 2E 29 0A seen_non_ascii: 1 no-conversion: slightly-likely utf-8: nearly-impossible utf-8-bom: nearly-impossible ucs-4: as-likely-as-unlikely utf-16: quite-improbable utf-16-little-endian: quite-improbable utf-16-bom: quite-improbable utf-16-little-endian-bom: quite-improbable iso-7: somewhat-unlikely iso-8-designate: somewhat-unlikely iso-8-1: somewhat-unlikely iso-8-2: somewhat-unlikely iso-lock-shift: somewhat-unlikely shift-jis: quite-improbable big5: somewhat-likely detect_coding_type: returning 0 (keep going) detected coding system: #<coding-system big5 big5> detected coding system: nil << deleted more than 31000 lines with the same output >> I'm more and more convinced that this is a problem with auto-detection. The test file has only latin1-encoded German umlauts beyond ASCII, and iso-8-1 gets a tag `somewhat-unlikely'. That doesn't seem to be correct. And the >30,000 lines of `detected coding system: nil' are looking suspicious as well. They don't appear when I visit a file with just the second line (and iso-8859-1 is properly selected then). They also don't appear when I visit a file with just the first line (when raw-text is selected as coding system). Perhaps this helps to categorize my problem. Where is the place where this likely/unlikely decision is made? In Lisp or in the C core? Cheers, Joachim -- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Joachim Schrod Email: jschrod(a)acm.org Roedermark, Germany