From aaf3dc5ba55d46b6b318bb761305013f4c533b26 Mon Sep 17 00:00:00 2001 From: Lukasz Kasprzak Date: Fri, 21 Aug 2026 12:09:35 +0200 Subject: fix(cli): diagnose an INI overlay passed to --overlay The INI overlay form is a SOURCE format that 'colitur convert' turns into the S-expression one. Handing the INI to --overlay instead failed deep in the sexp reader with 'more than one S-expression in file', which names neither the cause nor the cure -- and the shipped examples under data/ef/examples/ are exactly the files someone would try that with. The detector reads past the comment header to the first real line and looks for a [section]. Its first version capped that scan at 40 lines and poland.ini's own header is 43, so it silently never fired on the very file it was written for; the cap now bounds only how much of a binary gets read, and is not a guess at header length. --- bin/main.ml | 34 ++++++++++++++++++++++++++++++++++ test/cli.t | 11 +++++++++++ 2 files changed, 45 insertions(+) diff --git a/bin/main.ml b/bin/main.ml index 51c4989..c915255 100644 --- a/bin/main.ml +++ b/bin/main.ml @@ -169,8 +169,42 @@ let load_ef_layer ?(user_overlays = []) () = let dir = data_dir () in let sanctoral_path = Filename.concat dir "sanctoral.sexp" in let adjustments_path = Filename.concat dir "adjustments.sexp" in + (* An overlay handed to --overlay must be the S-expression form. The INI + form is real but is a SOURCE format: `colitur convert` turns it into one. + Feeding the INI here otherwise fails deep inside the sexp reader with + "more than one S-expression in file", which names neither the cause nor + the cure. Detect it and say both. *) + let looks_like_ini path = + match open_in path with + | exception _ -> false + | ic -> + (* Skip the comment header and stop at the first real line. The cap + only bounds how much of a binary file gets read; it is NOT a guess + at how long a header may be. The first version capped at 40 and + poland.ini's own header is 43 lines, so the detection silently + never fired on the very file it was written for. *) + let rec scan n = + if n > 2000 then false + else + match input_line ic with + | exception End_of_file -> false + | line -> + let l = String.trim line in + if l = "" || l.[0] = ';' || l.[0] = '#' then scan (n + 1) + else String.length l > 1 && l.[0] = '[' && l.[String.length l - 1] = ']' + in + let r = scan 0 in + close_in_noerr ic; + r + in let load_overlay path = match Colitur_kernel.Overlay.load Rite_ef.Vocab_ef.rank_of_sexp path with + | Error _ when looks_like_ini path -> + Error + (Printf.sprintf + "%s looks like an INI overlay, not an S-expression one.\n\ + colitur: convert it first: colitur convert %s > overlay.sexp" + path path) | Error e -> Error (Printf.sprintf "failed to load %s: %s" path e) | Ok o -> Ok o in diff --git a/test/cli.t b/test/cli.t index fe18800..b2c2263 100644 --- a/test/cli.t +++ b/test/cli.t @@ -1270,6 +1270,17 @@ slug the engine can produce over 2020-2045: partial.ini: 1 of 725 celebrations named, 724 missing, 0 unknown partial.ini: 0 of 104 book names, 104 missing +An INI overlay handed to `--overlay` is DIAGNOSED, not left to fail inside the +sexp reader. The INI form is a source format that `colitur convert` turns into +the S-expression one; feeding it directly otherwise produced "more than one +S-expression in file", which names neither the cause nor the cure. + + $ printf '; a comment header\n\n[overlay]\nid = x\n' > o.ini + $ colitur day 2027 --overlay o.ini + colitur: o.ini looks like an INI overlay, not an S-expression one. + colitur: convert it first: colitur convert o.ini > overlay.sexp + [2] + `--check` REJECTS an unknown slug (exit 1), so a typo is visible rather than silently dead -- its author would otherwise never learn why the name they wrote never appears: -- cgit v1.3