From 9d28f724620db27dc8de72bff9ea37ad5f6e9c6d Mon Sep 17 00:00:00 2001 From: Nicholas Paun Date: Thu, 29 Oct 2026 07:43:45 +0000 Subject: [PATCH] Add WPT mimesniff/mime-types test parsing.any.js runs the WHATWG MIME Sniffing parser vectors (mime-types.json and generated-mime-types.json) through two paths: new Blob/File({type}).type, and the type of a Blob read from a Request or Response with that Content-Type header. 1285 of 1898 subtests pass. Request/Response passes 933 of 942. Its failures are workerd MIME parser bugs (backslash treated as a token code point; an invalid parameter name does not consume its quoted value), Fetch's comma splitting in "extract a MIME type", and workerd accepting non-Latin-1 header values. The other 604 failures are Blob/File cases: the test expects MIME parsing, but the File API spec only lowercases the type, as workerd does. Browsers fail these heavily too. charset-parameter.window.js is omitted because it needs iframes. Co-Authored-By: Claude Opus 5.5 --- src/wpt/BUILD.bazel | 6 ++ src/wpt/mimesniff/mime-types-test.ts | 116 +++++++++++++++++++++++++++ 2 files changed, 122 insertions(+) create mode 100644 src/wpt/mimesniff/mime-types-test.ts diff --git a/src/wpt/BUILD.bazel b/src/wpt/BUILD.bazel index b385c59afb7..4bd0a8fb0c4 100644 --- a/src/wpt/BUILD.bazel +++ b/src/wpt/BUILD.bazel @@ -287,3 +287,9 @@ wpt_test( config = "console-test.ts", wpt_directory = "@wpt//:console@module", ) + +wpt_test( + name = "mimesniff/mime-types", + config = "mimesniff/mime-types-test.ts", + wpt_directory = "@wpt//:mimesniff/mime-types@module", +) diff --git a/src/wpt/mimesniff/mime-types-test.ts b/src/wpt/mimesniff/mime-types-test.ts new file mode 100644 index 00000000000..09c84f41140 --- /dev/null +++ b/src/wpt/mimesniff/mime-types-test.ts @@ -0,0 +1,116 @@ +// Copyright (c) 2017-2026 Cloudflare, Inc. +// Licensed under the Apache 2.0 license found in the LICENSE file or at: +// https://opensource.org/licenses/Apache-2.0 + +import { type TestRunnerConfig } from 'harness/harness'; + +export default { + 'charset-parameter.window.js': { + comment: + 'Loads each MIME type into an iframe and reads document.characterSet; workerd has no iframes or documents', + omittedTests: true, + }, + 'parsing.any.js': { + comment: + 'Blob/File: the test expects new Blob({type}).type to be the parsed and serialized MIME type, ' + + 'but the File API spec (and workerd, src/workerd/api/blob.c++ normalizeType) only lowercases ' + + 'printable-ASCII types and discards others, without MIME parsing. Browsers fail most of these ' + + 'too. The regexes cover the generated-mime-types.json families that fail for every listed ' + + 'code point (or for the ranges given); the strings are the failing mime-types.json cases. ' + + 'Request/Response: see the comment on each group below.', + expectedFailures: [ + // Blob/File, generated-mime-types.json: "/x" and "x/" for printable non-token + // code points. Control and non-ASCII code points pass because both sides give "". + // U+007F would pass too if normalizeType discarded it as the File API spec requires. + /^[ "(),:-@[\]\\{}\x7F]\/x \(Blob\/File\)$/, + /^x\/[ "(),:-@[\]\\{}\x7F] \(Blob\/File\)$/, + // Blob/File, generated-mime-types.json: "x/x;=x;bonus=x" and "x/x;x=;bonus=x" + // only exist for code points that make the parameter invalid, and all of them fail. + /^x\/x;[\s\S]=x;bonus=x \(Blob\/File\)$/, + /^x\/x;x=[\s\S];bonus=x \(Blob\/File\)$/, + // Blob/File, generated-mime-types.json: quoted parameter values containing control or + // non-ASCII code points. + /^x\/x;x="[^ -~]";bonus=x \(Blob\/File\)$/, + // Blob/File, mime-types.json. + 'TEXT/HTML;CHARSET=GBK (Blob/File)', + 'text/html;charset=gbk( (Blob/File)', + 'text/html;x=(;charset=gbk (Blob/File)', + 'text/html;charset=gbk;charset=windows-1255 (Blob/File)', + 'text/html;charset=();charset=GBK (Blob/File)', + 'text/html;charset =gbk (Blob/File)', + 'text/html ;charset=gbk (Blob/File)', + 'text/html; charset=gbk (Blob/File)', + 'text/html;charset= gbk (Blob/File)', + 'text/html;charset= "gbk" (Blob/File)', + 'text/html;charset=\x0Bgbk (Blob/File)', + 'text/html;charset=\x0Cgbk (Blob/File)', + 'text/html;\x0Bcharset=gbk (Blob/File)', + 'text/html;\x0Ccharset=gbk (Blob/File)', + "text/html;charset=';charset=GBK (Blob/File)", + 'text/html;test;charset=gbk (Blob/File)', + 'text/html;test=;charset=gbk (Blob/File)', + "text/html;';charset=gbk (Blob/File)", + 'text/html;";charset=gbk (Blob/File)', + 'text/html ; ; charset=gbk (Blob/File)', + 'text/html;;;;charset=gbk (Blob/File)', + 'text/html;charset= "\x7F;charset=GBK (Blob/File)', + 'text/html;charset="\x7F;charset=foo";charset=GBK (Blob/File)', + 'text/html;charset="gbk" (Blob/File)', + 'text/html;charset="gbk (Blob/File)', + 'text/html;charset=gbk" (Blob/File)', + 'text/html;charset="\\ gbk" (Blob/File)', + 'text/html;charset="\\g\\b\\k" (Blob/File)', + 'text/html;charset="gbk"x (Blob/File)', + 'text/html;charset="";charset=GBK (Blob/File)', + 'text/html;charset=";charset=GBK (Blob/File)', + 'text/html;charset={gbk} (Blob/File)', + 'text/html;a]=bar;b[=bar;c=bar (Blob/File)', + 'text/html;in]valid=";asd=foo";foo=bar (Blob/File)', + "!#$%&'*+-.^_`|~0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz/!#$%&'*+-.^_`|~0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz;!#$%&'*+-.^_`|~0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz=!#$%&'*+-.^_`|~0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz (Blob/File)", + 'x/x;x="\t !\\"#$%&\'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\\\]^_`abcdefghijklmnopqrstuvwxyz{|}~\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8A\x8B\x8C\x8D\x8E\x8F\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9A\x9B\x9C\x9D\x9E\x9F\xA0¡¢£¤¥¦§¨©ª«¬\xAD®¯°±²³´µ¶·¸¹º»¼½¾¿ÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖ×ØÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõö÷øùúûüýþÿ" (Blob/File)', + 'x/x;test (Blob/File)', + 'x/x;test="\\ (Blob/File)', + 'x/x;x= (Blob/File)', + 'x/x;x=\t (Blob/File)', + 'x/x\n\r\t ;x=x (Blob/File)', + '\n\r\t x/x;x=x\n\r\t (Blob/File)', + 'x/x;\n\r\t x=x\n\r\t ;x=y (Blob/File)', + 'text/html;test=ÿ;charset=gbk (Blob/File)', + 'x/x;test=\uFFFD;x=x (Blob/File)', + '/ (Blob/File)', + 'bogus (Blob/File)', + 'bogus/ (Blob/File)', + 'bogus/ (Blob/File)', + 'bogus/bogus/; (Blob/File)', + ' (Blob/File)', + '(/) (Blob/File)', + 'text/html(;doesnot=matter (Blob/File)', + '{/} (Blob/File)', + 'text /html (Blob/File)', + 'text/ html (Blob/File)', + '"text/html" (Blob/File)', + + // Request/Response: workerd's MIME parser treats '\' as an HTTP token code point + // (token_table in src/workerd/util/mimetype.c++), so it accepts '\' in types, subtypes + // and parameter names, and serializes a '\' parameter value without quoting it. + 'x/x;test="\\ (Request/Response)', + '\\/x (Request/Response)', + 'x/\\ (Request/Response)', + 'x/x;\\=x;bonus=x (Request/Response)', + // Request/Response: when a parameter name is invalid, workerd's MIME parser skips to + // the next ';' instead of first collecting the quoted-string value, so a ';' inside the + // quotes starts a new parameter (src/workerd/util/mimetype.c++, MimeType::tryParse). + 'text/html;in]valid=";asd=foo";foo=bar (Request/Response)', + // Request/Response: Body.blob() uses Fetch's "extract a MIME type", which splits the + // Content-Type value on commas outside quoted strings. The test expects the whole value + // to be parsed as one MIME type. workerd follows Fetch here. + 'x/x;,=x;bonus=x (Request/Response)', + 'x/x;x=,;bonus=x (Request/Response)', + // Request/Response: workerd deliberately accepts header values with code points above + // U+00FF (encoding them as UTF-8 and logging a warning) instead of throwing a TypeError + // (src/workerd/api/headers.c++, maybeWarnIfBadHeaderString). + 'x/x;test=\uFFFD;x=x (Request/Response)', + 'Ā/Ā (Request/Response)', + ], + }, +} satisfies TestRunnerConfig;