reffy 6.2.0 → 6.2.1

package/src/lib/util.js

@@ -1,898 +1,898 @@
-/**
- * A bunch of utility functions common to multiple scripts
- */
-
-const fs = require('fs').promises;
-const { existsSync } = require('fs');
-const path = require('path');
-const puppeteer = require('puppeteer');
-const crypto = require('crypto');
-const { AbortController } = require('abortcontroller-polyfill/dist/cjs-ponyfill');
-const fetch = require('./fetch');
-const specEquivalents = require('../specs/spec-equivalents.json');
-
-const reffyModules = require('../browserlib/reffy.json');
-
-
-/**
- * Maximum depth difference supported between Reffy's install path and custom
- * modules that may be provided on the command-line
- *
- * TODO: Find a way to get right of that, there should be no limit
- */
-const maxPathDepth = 20;
-
-
-/**
- * Returns a range array from 0 to the number provided (not included)
- */
-const range = n => Array.from(Array(n).keys());
-
-
-/**
- * Shortcut that returns a property extractor iterator
- */
-const prop = p => x => x[p];
-
-
-/**
- * Wrapper around the "require" function to require files relative to the
- * current working directory (CWD), instead of relative to the current JS
- * file.
- *
- * This is typically needed to be able to use "require" to load JSON config
- * files provided as command-line arguments.
- *
- * @function
- * @param {String} filename The path to the file to require
- * @return {Object} The result of requiring the file relative to the current
- *   working directory.
- */
-function requireFromWorkingDirectory(filename) {
-    try {
-        return require(path.resolve(filename));
-    }
-    catch (err) {
-        return null;
-    }
-}
-
-
-/**
- * Determine the path to the "node_modules" folder to resolve relative links
- * in the ES6 browser lib modules. The path depends on whether Reffy is run
- * directly, or installed as a library.
- *
- * @function
- * @return {String} Path to the node_modules folder.
- */
-function getModulesFolder() {
-    const rootFolder = path.resolve(__dirname, '../..');
-    let folder = path.resolve(rootFolder, 'node_modules');
-    if (existsSync(folder)) {
-        return folder;
-    }
-    folder = path.resolve(rootFolder, '..');
-    return folder;
-}
-const modulesFolder = getModulesFolder();
-
-
-/**
- * Puppeteer browser instance used to load and process specifications
- */
-let browser = null;
-
-/**
- * Promise resolved when there is no running instance of Puppeteer. This allows
- * to serialize calls to setupBrowser (and thus to crawlList and crawlSpecs in
- * specs-crawler.js)
- */
-let browserClosed = Promise.resolve();
-let resolveBrowserClosed = null;
-
-/**
- * The browser JS library that will be loaded onto every crawled page
- */
-let browserlib = null;
-
-
-/**
- * Expand list of browser modules with right set of descriptive properties
- * 
- * User may specify a browser module as:
- * - a name which must match one of the existing modules in browserlib
- * - a relative path to an .mjs file which must exist
- * - an object with an "href" property that is a relative path to an .mjs file
- * which must exist
- * 
- * Relative paths provided by the user are interpreted as relative to the
- * current working directory, and converted to be relative to the browserlib
- * directory.
- * 
- * @function
- * @public
- * @return {Array(Object)} List of modules with an href, name and property keys
- */
-function expandBrowserModules(modules) {
-    // Helper function to create a camelCase name out of a module path
-    function getCamelCaseName(href) {
-        const filename = href.replace(/([^\/\\]+)\.mjs$/, '$1');
-        const nameParts = filename.split('-');
-        let name;
-        let namePart;
-        while (namePart = nameParts.shift()) {
-            namePart = namePart.replace(/\W/g, '');
-            if (name) {
-                name += namePart.substring(0, 1).toUpperCase() + namePart.substring(1);
-            }
-            else {
-                name = namePart;
-            }
-        }
-        return name;
-    }
-
-    const browserlibPath = path.resolve(__dirname, '..', 'browserlib');
-    if (!modules) {
-        return reffyModules.map(mod => Object.assign({
-            name: getCamelCaseName(mod.href),
-            expanded: true
-        }, mod));
-    }
-
-    modules = modules.map(mod => {
-        if (typeof mod === 'string') {
-            if (mod.endsWith('.mjs')) {
-                const name = getCamelCaseName(mod);
-                return {
-                    href: path.relative(browserlibPath, path.join(process.cwd(), mod)).replace(/\\/g, '/'),
-                    name,
-                    property: name,
-                    expanded: true
-                };
-            }
-            else if (mod === 'core') {
-                return reffyModules.map(mod => Object.assign({
-                    name: getCamelCaseName(mod.href),
-                    expanded: true
-                }, mod));
-            }
-            else {
-                const res = reffyModules.find(m => m.href === mod ||
-                    getCamelCaseName(m.href) === mod || m.property === mod);
-                if (!res) {
-                    throw new Error(`Unknown browserlib module ${mod}`);
-                }
-                return Object.assign({
-                    name: getCamelCaseName(res.href),
-                    expanded: true
-                }, res);
-            }
-        }
-        else if (mod.expanded) {
-            return mod;
-        }
-        else {
-            if (!mod.href) {
-                throw new Error('Browserlib module does not have an "href" property');
-            }
-            mod.href = path.relative(browserlibPath, path.join(process.cwd(), mod.href)).replace(/\\/g, '/');
-            if (!mod.name) {
-                mod.name = getCamelCaseName(mod.href);
-            }
-            if (!mod.property) {
-                mod.property = mod.name;
-            }
-            mod.expanded = true;
-            return mod;
-        }
-    });
-
-    return modules.flat();
-}
-
-
-/**
- * Prepare the browserlib script that will be loaded in every crawled page.
- * 
- * The script exposes a global reffy namespace with the requested modules.
- * 
- * The function must be called before any attempt to call `processSpecification`
- * and should only be called once. The `setupBrowser` function takes care of it.
- * 
- * @function
- * @private
- */
-function setupBrowserlib(modules) {
-    modules = expandBrowserModules(modules);
-    browserlib = 'window.reffy = window.reffy ?? {};\n';
-
-    if (modules.find(module => module.needsIdToHeadingMap)) {
-        browserlib += `
-import mapIdsToHeadings from './map-ids-to-headings.mjs';
-window.reffy.mapIdsToHeadings = mapIdsToHeadings;\n`;
-    }
-
-    browserlib += modules.map(module => `
-import ${module.name} from '${module.href}';
-window.reffy.${module.name} = ${module.name};
-`).join('\n');
-}
-
-
-/**
- * Setup and launch browser instance to use to load and process specifications.
- *
- * The function must be called before any attempt to call `processSpecification`
- * and should only be called once.
- *
- * The function also generates the code that will inject the `reffy` namespace
- * in each processed page.
- *
- * Note: Switch `headless` to `false` to access dev tools and debug processing
- *
- * @function
- * @public
- */
-async function setupBrowser(modules) {
-    // There can be only one crawl running at a time
-    await browserClosed;
-    browserClosed = new Promise(resolve => resolveBrowserClosed = resolve);
-
-    // Create browser instance
-    // Note: switch "headless" to "false" (and comment out the call to
-    // "browser.close()") to access dev tools in debug mode
-    browser = await puppeteer.launch({ headless: true });
-    setupBrowserlib(modules);
-}
-
-
-/**
- * Close and destroy browser instance.
- *
- * The function should be called once at the end of the processing.
- *
- * @function
- * @public
- */
-async function teardownBrowser() {
-    if (browser) {
-        await browser.close();
-        browser = null;
-        resolveBrowserClosed();
-        resolveBrowserClosed = null;
-    }
-}
-
-
-/**
- * Load and process the given specification.
- *
- * The method automatically exposes Reffy's library functions in a window.reffy
- * namespace (see setupBrowserlib) so that the callback function can
- * call them directly. Additional callback arguments that would need to be
- * passed to the browser context can be provided through the "args" parameter.
- *
- * A crawl will typically fetch and render hundreds of specs, triggering a lot
- * of network requests. Given that some of these requests (e.g. those on images)
- * are of no interest for the processing, that it is wasteful to fetch the same
- * resource again and again during a crawl, and that it is useful to have an
- * offline mode for debugging purpose, the method will intercept network
- * requests made by the browser, fail those that don't seem needed, and serve
- * requests on resources that have already been fetched from a local file cache
- * (the "cacheRefresh" setting in "config.json" allows to adjust this behavior).
- *
- * This triggers a few hiccups and needs for workarounds though:
- * - Puppeteer's page.setRequestInterception does not play nicely with workers
- * (which Respec typically uses) for the time being, so code uses the Chrome
- * DevTools Protocol (CDP) directly, see:
- * https://github.com/puppeteer/puppeteer/issues/4208
- * - Tampering with network requests means that the loaded page gets
- * automatically flagged as "non secure". That's mostly fine but means that
- * "window.crypto.subtle" is not available and Respec needs that to generate
- * hashes. The code re-creates that method manually.
- * - A few specs send HTTP requests that return "streams". This does not work
- * well with Puppeteer's "networkidle0" option (to detect when a spec is mostly
- * done loading), and that does not work with a file cache approach either.
- * These requests get intercepted.
- *
- * A couple of additional notes:
- * - Requests to CSS stylesheets are not intercepted because Respec dynamically
- * loads a few CSS resources, and intercepting them could perhaps impact the
- * rest of the generation.
- * - SVG images are not intercepted because a couple of specs have a PNG
- * fallback mechanism that, when interception is on, make the browser spin
- * forever, see discussion in: https://github.com/w3c/accelerometer/pull/55
- *
- * Strictly speaking, intercepting request is only needed to be able to use the
- * "networkidle0" option. The whole interception logic could be dropped (and
- * "networkidle2" could be used instead) if it proves too unstable.
- *
- * @function
- * @public
- * @param {Object|String} spec The spec to load. Must either be a URL string or
- *   an object with a "url" property. If the object contains an "html" property,
- *   the HTML content is loaded instead.
- * @param {function} processFunction Processing function that will be evaluated
- *   in the browser context where the spec gets loaded
- * @param {Arrays} args List of arguments to pass to the processing function.
- *   These arguments typically make it possible to pass contextual information
- *   to the processing function (such as the spec object that describes the
- *   spec being processed, or the list of processing modules to run)
- * @param {Object} options Processing options. The "quiet" flag tells the
- *   function not to report warnings to the console. The "forceLocalFetch"
- *   flag tells the function that all network requests need to be only handled
- *   by Node.js's "fetch" function (as opposed to falling back to Puppeteer's
- *   network and caching logic), which is useful to keep full control of network
- *   requests in tests.
- * @return {Promise} The promise to get the results of the processing function
- */
-async function processSpecification(spec, processFunction, args, options) {
-    spec = (typeof spec === 'string') ? { url: spec } : spec;
-    processFunction = processFunction || function () {};
-    args = args || [];
-    options = options || {};
-
-    if (!browser) {
-        throw new Error('Browser instance not initialized, setupBrowser() must be called before processSpecification().');
-    }
-
-    // Create an abort controller for network requests directly handled by the
-    // Node.js code (and not by Puppeteer)
-    const abortController = new AbortController();
-
-    // Inner function that returns a network interception method suitable for
-    // a given CDP session.
-    function interceptRequest(cdp, controller) {
-        return async function ({ requestId, request }) {
-            try {
-                // Abort network requests to common image formats
-                if (/\.(gif|ico|jpg|jpeg|png|ttf|woff)$/i.test(request.url)) {
-                    await cdp.send('Fetch.failRequest', { requestId, errorReason: 'Failed' });
-                    return;
-                }
-
-                // Abort network requests that return a "stream", they won't
-                // play well with Puppeteer's "networkidle0" option, and our
-                // custom "fetch" function does not handle streams in any case
-                if (request.url.startsWith('https://drafts.csswg.org/api/drafts/') ||
-                    request.url.startsWith('https://drafts.css-houdini.org/api/drafts/') ||
-                    request.url.startsWith('https://drafts.fxtf.org/api/drafts/') ||
-                    request.url.startsWith('https://api.csswg.org/shepherd/')) {
-                    await cdp.send('Fetch.failRequest', { requestId, errorReason: 'Failed' });
-                    return;
-                }
-
-                // The request needs to be intercepted, either because it
-                // targets one of the local script files, or because we would
-                // like to use our local cache to avoid sending network requests
-                // when possible.
-                //console.log(`intercept ${request.url}`);
-                const reffyPath = '/reffy/scripts/';
-                const webidl2Path = '/node_modules/webidl2/';
-                if (request.url.includes(reffyPath) || request.url.includes(webidl2Path)) {
-                    let body;
-                    if (request.url.endsWith('reffy.mjs')) {
-                        body = Buffer.from(browserlib);
-                    }
-                    else if (request.url.includes(webidl2Path)) {
-                        const file = path.resolve(modulesFolder, 'webidl2',
-                            request.url.substring(request.url.indexOf(webidl2Path) + webidl2Path.length));
-                        body = await fs.readFile(file);
-                    }
-                    else {
-                        // The "__" folders are just a means to resolve
-                        // relative paths that are higher than the "browserlib"
-                        // folder on the storage drive
-                        const requestPath = request.url.substring(request.url.indexOf(reffyPath) + reffyPath.length);
-                        let depth = requestPath.lastIndexOf('__/') / 3;
-                        const filename = requestPath.substring(requestPath.lastIndexOf('__/') + 3);
-                        let filePath = path.resolve(__dirname, '..', 'browserlib');
-                        while (depth < maxPathDepth - 1) {
-                            filePath = path.resolve(filePath, '..');
-                            depth += 1;
-                        }
-                        const file = path.resolve(filePath, filename);
-                        body = await fs.readFile(file);
-                    }
-                    await cdp.send('Fetch.fulfillRequest', {
-                        requestId,
-                        responseCode: 200,
-                        responseHeaders: [{ name: 'Content-Type', value: 'application/javascript' }],
-                        body: body.toString('base64')
-                    });
-                }
-                else {
-                    if ((request.method !== 'GET') ||
-                        (!request.url.startsWith('http:') && !request.url.startsWith('https:'))) {
-                        await cdp.send('Fetch.continueRequest', { requestId });
-                        return;
-                    }
-
-                    const response = await fetch(request.url, { signal: controller.signal });
-                    const body = await response.buffer();
-                    await cdp.send('Fetch.fulfillRequest', {
-                        requestId,
-                        responseCode: response.status,
-                        responseHeaders: Object.keys(response.headers.raw()).map(header => {
-                            return {
-                                name: header,
-                                value: response.headers.raw()[header].join(',')
-                            };
-                        }),
-                        body: body.toString('base64')
-                    });
-                }
-                //console.log(`intercept ${request.url} - done`);
-            }
-            catch (err) {
-                if (controller.signal.aborted) {
-                    // All is normal, processing was over, page and CDP session
-                    // have been closed, and network requests have been aborted
-                    // console.log(`intercept ${request.url} - aborted`);
-                    return;
-                }
-
-                // Fetch from file cache failed somehow
-                // Let Puppeteer handle the request as fallback unless
-                // calling function asked us not to do that
-                if (options.forceLocalFetch) {
-                    options.quiet ?? console.warn(`[warn] Network request for ${request.url} failed`, err);
-                    await cdp.send('Fetch.failRequest', { requestId, errorReason: 'Failed' });
-                }
-                else {
-                    options.quiet ?? console.warn(`[warn] Fall back to regular network request for ${request.url}`, err);
-                    try {
-                        await cdp.send('Fetch.continueRequest', { requestId });
-                    }
-                    catch (err) {
-                        if (!controller.signal.aborted) {
-                            options.quiet ?? console.warn(`[warn] Fall back to regular network request for ${request.url} failed`, err);
-                        }
-                    }
-                }
-            }
-        }
-    }
-
-    try {
-        const page = await browser.newPage();
-
-        // Disable cache if caller wants to handle all network requests
-        await page.setCacheEnabled(!options.forceLocalFetch);
-
-        // Intercept all network requests to use our own version of "fetch"
-        // that makes use of the local file cache.
-        const cdp = await page.target().createCDPSession();
-        await cdp.send('Fetch.enable');
-        cdp.on('Fetch.requestPaused', interceptRequest(cdp, abortController));
-
-        // Quick and dirty workaround to re-create the "window.crypto.digest"
-        // function that Respec needs (context is seen as unsecure because we're
-        // tampering with network requests)
-        await page.exposeFunction('hashdigest', (algorithm, buffer) => {
-            return crypto.createHash(algorithm).update(Buffer.from(Object.values(buffer))).digest();
-        });
-        await page.evaluateOnNewDocument(() => {
-            window.crypto.subtle = {
-                digest: function (algorithm, buffer) {
-                    const res = window.hashdigest('sha1', buffer);
-                    return res.then(buf => {
-                        return Uint8Array.from(buf.data);
-                    });
-                }
-            };
-        });
-
-        // Common loading option to give the browser enough time to load large
-        // specs, and to consider navigation done when there haven't been
-        // network connections in the past 500ms. This should be enough to
-        // handle "redirection" through JS or meta refresh (which would not
-        // have time to run if we used "load").
-        const loadOptions = {
-            timeout: 120000,
-            waitUntil: 'networkidle0'
-        };
-
-        // Load the page
-        // (note HTTP status is 0 when `file://` URLs are loaded)
-        if (spec.html) {
-            await page.setContent(spec.html, loadOptions);
-        }
-        else {
-            const result = await page.goto(spec.url, loadOptions);
-            if ((result.status() !== 200) && (!spec.url.startsWith('file://') || (result.status() !== 0))) {
-                throw new Error(`Loading ${spec.url} triggered HTTP status ${result.status()}`);
-            }
-        }
-
-        // Handle multi-page specs
-        const pageUrls = spec.pages || [];
-
-        if (pageUrls.length > 0) {
-            const pages = [];
-            for (const url of pageUrls) {
-                const subAbort = new AbortController();
-                const subPage = await browser.newPage();
-                await subPage.setCacheEnabled(!options.forceLocalFetch);
-                const subCdp = await subPage.target().createCDPSession();
-                await subCdp.send('Fetch.enable');
-                subCdp.on('Fetch.requestPaused', interceptRequest(subCdp, subAbort));
-                try {
-                    // (Note HTTP status is 0 when `file://` URLs are loaded)
-                    const subresult = await subPage.goto(url, loadOptions);
-                    if ((subresult.status() !== 200) && (!url.startsWith('file://') || (subresult.status() !== 0))) {
-                        throw new Error(`Loading ${spec.url} triggered HTTP status ${result.status()}`);
-                    }
-                    const html = await subPage.evaluate(() => {
-                        return document.body.outerHTML
-                            .replace(/<body/, '<section')
-                            .replace(/<\/body/, '</section');
-                    });
-                    pages.push({ url, html });
-                }
-                finally {
-                    subAbort.abort();
-                    await subCdp.detach();
-                    await subPage.close();
-                }
-            }
-            await page.evaluate(pages => {
-                for (const subPage of pages) {
-                    const section = document.createElement('section');
-                    section.setAttribute('data-reffy-page', subPage.url);
-                    section.innerHTML = subPage.html;
-                    document.body.appendChild(section);
-                }
-            }, pages);
-        }
-
-        // Wait until the generation of the spec is completely over
-        await page.evaluate(async () => {
-            // Detect draft CSS server hiccups as done in browser-specs:
-            // https://github.com/w3c/browser-specs/blob/b31fc0b03ba67a19162883afc30e01fcec3c600d/src/fetch-info.js#L292
-            const title = (window.document.querySelector('h1')?.textContent || '')
-                .replace(/\n/g, '').trim();
-            if (title.startsWith('Index of ')) {
-                throw new Error(`CSS server issue detected`);
-            }
-
-            const usesRespec = (window.respecConfig || window.eval('typeof respecConfig !== "undefined"')) &&
-                window.document.head.querySelector("script[src*='respec']");
-
-            function sleep(ms) {
-                return new Promise(resolve => setTimeout(resolve, ms));
-            }
-
-            async function isReady(counter) {
-                counter = counter || 0;
-                if (counter > 60) {
-                    throw new Error('Respec generation took too long');
-                }
-                if (window.document.respec?.ready) {
-                    await window.document.respec.ready;
-                }
-                else if (usesRespec) {
-                    await sleep(1000);
-                    await isReady(counter + 1);
-                }
-            }
-
-            await isReady();
-        });
-
-        // Capture and report Reffy's browserlib warnings
-        page.on('console', msg => {
-            const text = msg.text();
-            if (text.startsWith('[reffy] ')) {
-                options.quiet ?? console.warn(spec.url, `[${msg.type()}]`, msg.text().substr('[reffy] '.length));
-            }
-        });
-
-        // Capture and report when page throws an error
-        page.on('pageerror', err => {
-            options.quiet ?? console.warn(err);
-        });
-
-        // Expose additional functions defined in src/browserlib/ to the
-        // browser context, under a window.reffy namespace, so that processing
-        // script may call them. The script is an ES6 module and needs to be
-        // loaded as such.
-        // Note that we're using a fake relative URL on purpose. In practice,
-        // the request will be processed by "interceptRequest", which will
-        // respond with the contents of the script file. Also, there are
-        // multiple path levels in that fake URL on purpose as well, because
-        // scripts import the WebIDL2.js library with a URL like
-        // "../../node_modules/[...]" and may import other scripts that are
-        // higher in the folder tree.
-        await page.addScriptTag({
-            url: `reffy/scripts/${range(maxPathDepth).map(n => '__').join('/')}/reffy.mjs`,
-            type: 'module'
-        });
-
-        // Run the processFunction method in the browser context
-        const results = await page.evaluate(processFunction, ...args);
-
-        // Pending network requests may still be in the queue, flag the page
-        // as closed not to send commands on a CDP session that's no longer
-        // attached to anything
-        abortController.abort();
-
-        // Close CDP session and page
-        // Note that gets done no matter what when browser.close() gets called.
-        await cdp.detach();
-        await page.close();
-
-        return results;
-    }
-    finally {
-        // Signal abortion again (in case an exception was thrown)
-        abortController.abort();
-    }
-}
-
-
-/**
- * Enrich the spec description with alternative URLs (versions and equivalents)
- *
- * TODO: The list used to contain published versions of TR specs retrieved from
- * the W3C API. They are useful to improve the relevance of reported anomalies.
- *
- * @function
- * @param {Object} spec Spec description structure (only the URL is useful)
- * @return {Object} The same structure, enriched with the URL of the editor's
- *   draft when one is found
- */
-function completeWithAlternativeUrls(spec) {
-    spec.versions = new Set();
-    spec.versions.add(spec.url);
-    if (spec.release) {
-        spec.versions.add(spec.release.url);
-    }
-    if (spec.nightly) {
-        spec.versions.add(spec.nightly.url);
-    }
-    if (specEquivalents[spec.url]) {
-        spec.versions = new Set([
-            ...spec.versions,
-            ...specEquivalents[spec.url]
-        ]);
-    }
-    spec.versions = [...spec.versions];
-    return spec;
-}
-
-
-/**
- * Returns true when the given spec is the latest "fullest" level of that spec
- * in the given list of specs that passes the given predicate.
- *
- * "Fullest" means "not a delta spec, unless that is the only level that passes
- * the predicate".
- *
- * @function
- * @public
- * @param {Object} spec Spec to check
- * @param {Array(Object)} list List of specs (must include the spec to check)
- * @param {function} predicate Predicate function that the spec must pass. Must
- *   be a function that takes a spec as argument and returns a boolean.
- * @return {Boolean} true if the spec is the latest "fullest" level in the list
- *   that passes the predicate.
- */
-function isLatestLevelThatPasses(spec, list, predicate) {
-    predicate = predicate || (_ => true);
-    if (!predicate(spec)) {
-        return false;
-    }
-    if (spec.seriesComposition === 'delta') {
-        while (spec.seriesPrevious) {
-            spec = list.find(s => s.shortname === spec.seriesPrevious);
-            if (!spec) {
-                break;
-            }
-            if ((spec.seriesComposition === 'full') && predicate(spec)) {
-                return false;
-            }
-        }
-        return true;
-    }
-    while (spec.seriesNext) {
-        if (!spec) {
-            break;
-        }
-        spec = list.find(s => s.shortname === spec.seriesNext);
-        if ((spec.seriesComposition === 'full') && predicate(spec)) {
-            return false;
-        }
-    }
-    return true;
-}
-
-
-/**
- * Takes the results of a crawl for a given spec and expands it to include the
- * contents of referenced files.
- *
- * The function handles both files and HTTPS resources, using either filesystem
- * functions (for files) or fetch (for HTTPS resources).
- *
- * Note the spec object is expanded in place.
- *
- * @function
- * @public
- * @param {Object} spec Spec crawl result that needs to be expanded
- * @param {string} baseFolder The base folder that contains the crawl file, or
- *   the base HTTPS URI to resolve relative links in the crawl object.
- * @param {Array(string)} properties An explicit list of properties to expand
- *   (no value means "expand all possible properties")
- * @return {Promise(object)} The promise to get an expanded crawl object that
- *   contains the contents of referenced files and no longer references external
- *   files (for the requested properties)
- */
-async function expandSpecResult(spec, baseFolder, properties) {
-    baseFolder = baseFolder || '';
-    await Promise.all(Object.keys(spec).map(async property => {
-        // Only consider properties explicitly requested
-        if (properties && !properties.includes(property)) {
-            return;
-        }
-
-        // Only consider properties that link to an extract, i.e. an IDL
-        // or JSON file in subfolder.
-        if (!spec[property] ||
-                (typeof spec[property] !== 'string') ||
-                !spec[property].match(/^[^\/]+\/[^\/]+\.(json|idl)$/)) {
-            return;
-        }
-        let contents = null;
-        if (baseFolder.startsWith('https:')) {
-            const url = (new URL(spec[property], baseFolder)).toString();
-            const response = await fetch(url, { nolog: true });
-            contents = await response.text();
-        }
-        else {
-            const filename = path.join(baseFolder, spec[property]);
-            contents = await fs.readFile(filename, 'utf8');
-        }
-        if (spec[property].endsWith('.json')) {
-            contents = JSON.parse(contents);
-        }
-        if (property === 'css') {
-            // Special case for CSS where the "css" level does not exist
-            // in the generated files
-            const css = Object.assign({}, contents);
-            delete css.spec;
-            spec[property] = css;
-        }
-        else if (property === 'idl') {
-            // Special case for raw IDL extracts, which are text extracts.
-            // Also drop header that may have been added when extract was
-            // serialized.
-            if (contents.startsWith('// GENERATED CONTENT - DO NOT EDIT')) {
-                const endOfHeader = contents.indexOf('\n\n');
-                contents = contents.substring(endOfHeader + 2);
-            }
-            spec.idl = contents;
-        }
-        else {
-            spec[property] = contents[property];
-        }
-    }));
-    return spec;
-}
-
-
-/**
- * Takes the results of a crawl (typically the contents of the index.json file)
- * and expands it to include the contents of all referenced files.
- *
- * The function handles both files and HTTPS resources, using either filesystem
- * functions (for files) or fetch (for HTTPS resources).
- *
- * Note the crawl object is expanded in place.
- *
- * @function
- * @public
- * @param {Object} crawl Crawl index object that needs to be expanded
- * @param {string} baseFolder The base folder that contains the crawl file, or
- *   the base HTTPS URI to resolve relative links in the crawl object.
- * @param {Array(string)} properties An explicit list of properties to expand
- *   (no value means "expand all possible properties")
- * @return {Promise(object)} The promise to get an expanded crawl object that
- *   contains the entire crawl report (and no longer references external files)
- */
-async function expandCrawlResult(crawl, baseFolder, properties) {
-    baseFolder = baseFolder || '';
-    crawl.results = await Promise.all(
-        crawl.results.map(spec => expandSpecResult(spec, baseFolder, properties))
-    );
-    return crawl;
-}
-
-
-/**
- * Retrieves the list of IDL attribute names that the CSS property generates
- * per the CSSOM spec, see:
- * https://drafts.csswg.org/cssom/#ref-for-css-property-to-idl-attribute
- *
- * @function
- * @param {String} property CSS property name
- * @return {Array(String)} An array of IDL attribute names, dashed attribute
- *   first, then camel-cased attribute if different, then webkit-cased attribute
- *   name if needed
- */
-function getGeneratedIDLNamesByCSSProperty(property) {
-    // Converts a CSS property to an IDL attribute name per the CSSOM spec:
-    // https://drafts.csswg.org/cssom/#css-property-to-idl-attribute
-    function cssPropertyToIDLAttribute(property, lowercaseFirst) {
-        let output = '';
-        let uppercaseNext = false;
-        if (lowercaseFirst) {
-            property = property.substr(1);
-        }
-        for (const c of property) {
-            if (c === '-') {
-                uppercaseNext = true;
-            } else if (uppercaseNext) {
-                uppercaseNext = false;
-                output += c.toUpperCase();
-            } else {
-                output += c;
-            }
-        }
-        return output;
-    }
-
-    // Start with dashed attribute
-    const res = [property];
-
-    // Add camel-cased attribute if different
-    const camelCased = cssPropertyToIDLAttribute(property, false);
-    if (camelCased !== property) {
-        res.push(camelCased);
-    }
-
-    // Add webkit-cased attribute if needed
-    if (property.startsWith('-webkit-')) {
-        res.push(cssPropertyToIDLAttribute(property, true));
-    }
-
-    return res;
-};
-
-
-/**
- * Creates the given folder if it does not exist yet.
- *
- * @function
- * @public
- * @param {String} folder Path to folder to create
- *   (from current working directory)
- */
-async function createFolderIfNeeded(folder) {
-    try {
-        await fs.mkdir(folder);
-    }
-    catch (err) {
-        if (err.code !== 'EEXIST') {
-            throw err;
-        }
-    }
-}
-
-
-module.exports = {
-    fetch,
-    requireFromWorkingDirectory,
-    expandBrowserModules,
-    setupBrowser,
-    teardownBrowser,
-    processSpecification,
-    completeWithAlternativeUrls,
-    isLatestLevelThatPasses,
-    expandCrawlResult,
-    expandSpecResult,
-    getGeneratedIDLNamesByCSSProperty,
-    createFolderIfNeeded
-};
+/**
+ * A bunch of utility functions common to multiple scripts
+ */
+
+const fs = require('fs').promises;
+const { existsSync } = require('fs');
+const path = require('path');
+const puppeteer = require('puppeteer');
+const crypto = require('crypto');
+const { AbortController } = require('abortcontroller-polyfill/dist/cjs-ponyfill');
+const fetch = require('./fetch');
+const specEquivalents = require('../specs/spec-equivalents.json');
+
+const reffyModules = require('../browserlib/reffy.json');
+
+
+/**
+ * Maximum depth difference supported between Reffy's install path and custom
+ * modules that may be provided on the command-line
+ *
+ * TODO: Find a way to get right of that, there should be no limit
+ */
+const maxPathDepth = 20;
+
+
+/**
+ * Returns a range array from 0 to the number provided (not included)
+ */
+const range = n => Array.from(Array(n).keys());
+
+
+/**
+ * Shortcut that returns a property extractor iterator
+ */
+const prop = p => x => x[p];
+
+
+/**
+ * Wrapper around the "require" function to require files relative to the
+ * current working directory (CWD), instead of relative to the current JS
+ * file.
+ *
+ * This is typically needed to be able to use "require" to load JSON config
+ * files provided as command-line arguments.
+ *
+ * @function
+ * @param {String} filename The path to the file to require
+ * @return {Object} The result of requiring the file relative to the current
+ *   working directory.
+ */
+function requireFromWorkingDirectory(filename) {
+    try {
+        return require(path.resolve(filename));
+    }
+    catch (err) {
+        return null;
+    }
+}
+
+
+/**
+ * Determine the path to the "node_modules" folder to resolve relative links
+ * in the ES6 browser lib modules. The path depends on whether Reffy is run
+ * directly, or installed as a library.
+ *
+ * @function
+ * @return {String} Path to the node_modules folder.
+ */
+function getModulesFolder() {
+    const rootFolder = path.resolve(__dirname, '../..');
+    let folder = path.resolve(rootFolder, 'node_modules');
+    if (existsSync(folder)) {
+        return folder;
+    }
+    folder = path.resolve(rootFolder, '..');
+    return folder;
+}
+const modulesFolder = getModulesFolder();
+
+
+/**
+ * Puppeteer browser instance used to load and process specifications
+ */
+let browser = null;
+
+/**
+ * Promise resolved when there is no running instance of Puppeteer. This allows
+ * to serialize calls to setupBrowser (and thus to crawlList and crawlSpecs in
+ * specs-crawler.js)
+ */
+let browserClosed = Promise.resolve();
+let resolveBrowserClosed = null;
+
+/**
+ * The browser JS library that will be loaded onto every crawled page
+ */
+let browserlib = null;
+
+
+/**
+ * Expand list of browser modules with right set of descriptive properties
+ * 
+ * User may specify a browser module as:
+ * - a name which must match one of the existing modules in browserlib
+ * - a relative path to an .mjs file which must exist
+ * - an object with an "href" property that is a relative path to an .mjs file
+ * which must exist
+ * 
+ * Relative paths provided by the user are interpreted as relative to the
+ * current working directory, and converted to be relative to the browserlib
+ * directory.
+ * 
+ * @function
+ * @public
+ * @return {Array(Object)} List of modules with an href, name and property keys
+ */
+function expandBrowserModules(modules) {
+    // Helper function to create a camelCase name out of a module path
+    function getCamelCaseName(href) {
+        const filename = href.replace(/([^\/\\]+)\.mjs$/, '$1');
+        const nameParts = filename.split('-');
+        let name;
+        let namePart;
+        while (namePart = nameParts.shift()) {
+            namePart = namePart.replace(/\W/g, '');
+            if (name) {
+                name += namePart.substring(0, 1).toUpperCase() + namePart.substring(1);
+            }
+            else {
+                name = namePart;
+            }
+        }
+        return name;
+    }
+
+    const browserlibPath = path.resolve(__dirname, '..', 'browserlib');
+    if (!modules) {
+        return reffyModules.map(mod => Object.assign({
+            name: getCamelCaseName(mod.href),
+            expanded: true
+        }, mod));
+    }
+
+    modules = modules.map(mod => {
+        if (typeof mod === 'string') {
+            if (mod.endsWith('.mjs')) {
+                const name = getCamelCaseName(mod);
+                return {
+                    href: path.relative(browserlibPath, path.join(process.cwd(), mod)).replace(/\\/g, '/'),
+                    name,
+                    property: name,
+                    expanded: true
+                };
+            }
+            else if (mod === 'core') {
+                return reffyModules.map(mod => Object.assign({
+                    name: getCamelCaseName(mod.href),
+                    expanded: true
+                }, mod));
+            }
+            else {
+                const res = reffyModules.find(m => m.href === mod ||
+                    getCamelCaseName(m.href) === mod || m.property === mod);
+                if (!res) {
+                    throw new Error(`Unknown browserlib module ${mod}`);
+                }
+                return Object.assign({
+                    name: getCamelCaseName(res.href),
+                    expanded: true
+                }, res);
+            }
+        }
+        else if (mod.expanded) {
+            return mod;
+        }
+        else {
+            if (!mod.href) {
+                throw new Error('Browserlib module does not have an "href" property');
+            }
+            mod.href = path.relative(browserlibPath, path.join(process.cwd(), mod.href)).replace(/\\/g, '/');
+            if (!mod.name) {
+                mod.name = getCamelCaseName(mod.href);
+            }
+            if (!mod.property) {
+                mod.property = mod.name;
+            }
+            mod.expanded = true;
+            return mod;
+        }
+    });
+
+    return modules.flat();
+}
+
+
+/**
+ * Prepare the browserlib script that will be loaded in every crawled page.
+ * 
+ * The script exposes a global reffy namespace with the requested modules.
+ * 
+ * The function must be called before any attempt to call `processSpecification`
+ * and should only be called once. The `setupBrowser` function takes care of it.
+ * 
+ * @function
+ * @private
+ */
+function setupBrowserlib(modules) {
+    modules = expandBrowserModules(modules);
+    browserlib = 'window.reffy = window.reffy ?? {};\n';
+
+    if (modules.find(module => module.needsIdToHeadingMap)) {
+        browserlib += `
+import mapIdsToHeadings from './map-ids-to-headings.mjs';
+window.reffy.mapIdsToHeadings = mapIdsToHeadings;\n`;
+    }
+
+    browserlib += modules.map(module => `
+import ${module.name} from '${module.href}';
+window.reffy.${module.name} = ${module.name};
+`).join('\n');
+}
+
+
+/**
+ * Setup and launch browser instance to use to load and process specifications.
+ *
+ * The function must be called before any attempt to call `processSpecification`
+ * and should only be called once.
+ *
+ * The function also generates the code that will inject the `reffy` namespace
+ * in each processed page.
+ *
+ * Note: Switch `headless` to `false` to access dev tools and debug processing
+ *
+ * @function
+ * @public
+ */
+async function setupBrowser(modules) {
+    // There can be only one crawl running at a time
+    await browserClosed;
+    browserClosed = new Promise(resolve => resolveBrowserClosed = resolve);
+
+    // Create browser instance
+    // Note: switch "headless" to "false" (and comment out the call to
+    // "browser.close()") to access dev tools in debug mode
+    browser = await puppeteer.launch({ headless: true });
+    setupBrowserlib(modules);
+}
+
+
+/**
+ * Close and destroy browser instance.
+ *
+ * The function should be called once at the end of the processing.
+ *
+ * @function
+ * @public
+ */
+async function teardownBrowser() {
+    if (browser) {
+        await browser.close();
+        browser = null;
+        resolveBrowserClosed();
+        resolveBrowserClosed = null;
+    }
+}
+
+
+/**
+ * Load and process the given specification.
+ *
+ * The method automatically exposes Reffy's library functions in a window.reffy
+ * namespace (see setupBrowserlib) so that the callback function can
+ * call them directly. Additional callback arguments that would need to be
+ * passed to the browser context can be provided through the "args" parameter.
+ *
+ * A crawl will typically fetch and render hundreds of specs, triggering a lot
+ * of network requests. Given that some of these requests (e.g. those on images)
+ * are of no interest for the processing, that it is wasteful to fetch the same
+ * resource again and again during a crawl, and that it is useful to have an
+ * offline mode for debugging purpose, the method will intercept network
+ * requests made by the browser, fail those that don't seem needed, and serve
+ * requests on resources that have already been fetched from a local file cache
+ * (the "cacheRefresh" setting in "config.json" allows to adjust this behavior).
+ *
+ * This triggers a few hiccups and needs for workarounds though:
+ * - Puppeteer's page.setRequestInterception does not play nicely with workers
+ * (which Respec typically uses) for the time being, so code uses the Chrome
+ * DevTools Protocol (CDP) directly, see:
+ * https://github.com/puppeteer/puppeteer/issues/4208
+ * - Tampering with network requests means that the loaded page gets
+ * automatically flagged as "non secure". That's mostly fine but means that
+ * "window.crypto.subtle" is not available and Respec needs that to generate
+ * hashes. The code re-creates that method manually.
+ * - A few specs send HTTP requests that return "streams". This does not work
+ * well with Puppeteer's "networkidle0" option (to detect when a spec is mostly
+ * done loading), and that does not work with a file cache approach either.
+ * These requests get intercepted.
+ *
+ * A couple of additional notes:
+ * - Requests to CSS stylesheets are not intercepted because Respec dynamically
+ * loads a few CSS resources, and intercepting them could perhaps impact the
+ * rest of the generation.
+ * - SVG images are not intercepted because a couple of specs have a PNG
+ * fallback mechanism that, when interception is on, make the browser spin
+ * forever, see discussion in: https://github.com/w3c/accelerometer/pull/55
+ *
+ * Strictly speaking, intercepting request is only needed to be able to use the
+ * "networkidle0" option. The whole interception logic could be dropped (and
+ * "networkidle2" could be used instead) if it proves too unstable.
+ *
+ * @function
+ * @public
+ * @param {Object|String} spec The spec to load. Must either be a URL string or
+ *   an object with a "url" property. If the object contains an "html" property,
+ *   the HTML content is loaded instead.
+ * @param {function} processFunction Processing function that will be evaluated
+ *   in the browser context where the spec gets loaded
+ * @param {Arrays} args List of arguments to pass to the processing function.
+ *   These arguments typically make it possible to pass contextual information
+ *   to the processing function (such as the spec object that describes the
+ *   spec being processed, or the list of processing modules to run)
+ * @param {Object} options Processing options. The "quiet" flag tells the
+ *   function not to report warnings to the console. The "forceLocalFetch"
+ *   flag tells the function that all network requests need to be only handled
+ *   by Node.js's "fetch" function (as opposed to falling back to Puppeteer's
+ *   network and caching logic), which is useful to keep full control of network
+ *   requests in tests.
+ * @return {Promise} The promise to get the results of the processing function
+ */
+async function processSpecification(spec, processFunction, args, options) {
+    spec = (typeof spec === 'string') ? { url: spec } : spec;
+    processFunction = processFunction || function () {};
+    args = args || [];
+    options = options || {};
+
+    if (!browser) {
+        throw new Error('Browser instance not initialized, setupBrowser() must be called before processSpecification().');
+    }
+
+    // Create an abort controller for network requests directly handled by the
+    // Node.js code (and not by Puppeteer)
+    const abortController = new AbortController();
+
+    // Inner function that returns a network interception method suitable for
+    // a given CDP session.
+    function interceptRequest(cdp, controller) {
+        return async function ({ requestId, request }) {
+            try {
+                // Abort network requests to common image formats
+                if (/\.(gif|ico|jpg|jpeg|png|ttf|woff)$/i.test(request.url)) {
+                    await cdp.send('Fetch.failRequest', { requestId, errorReason: 'Failed' });
+                    return;
+                }
+
+                // Abort network requests that return a "stream", they won't
+                // play well with Puppeteer's "networkidle0" option, and our
+                // custom "fetch" function does not handle streams in any case
+                if (request.url.startsWith('https://drafts.csswg.org/api/drafts/') ||
+                    request.url.startsWith('https://drafts.css-houdini.org/api/drafts/') ||
+                    request.url.startsWith('https://drafts.fxtf.org/api/drafts/') ||
+                    request.url.startsWith('https://api.csswg.org/shepherd/')) {
+                    await cdp.send('Fetch.failRequest', { requestId, errorReason: 'Failed' });
+                    return;
+                }
+
+                // The request needs to be intercepted, either because it
+                // targets one of the local script files, or because we would
+                // like to use our local cache to avoid sending network requests
+                // when possible.
+                //console.log(`intercept ${request.url}`);
+                const reffyPath = '/reffy/scripts/';
+                const webidl2Path = '/node_modules/webidl2/';
+                if (request.url.includes(reffyPath) || request.url.includes(webidl2Path)) {
+                    let body;
+                    if (request.url.endsWith('reffy.mjs')) {
+                        body = Buffer.from(browserlib);
+                    }
+                    else if (request.url.includes(webidl2Path)) {
+                        const file = path.resolve(modulesFolder, 'webidl2',
+                            request.url.substring(request.url.indexOf(webidl2Path) + webidl2Path.length));
+                        body = await fs.readFile(file);
+                    }
+                    else {
+                        // The "__" folders are just a means to resolve
+                        // relative paths that are higher than the "browserlib"
+                        // folder on the storage drive
+                        const requestPath = request.url.substring(request.url.indexOf(reffyPath) + reffyPath.length);
+                        let depth = requestPath.lastIndexOf('__/') / 3;
+                        const filename = requestPath.substring(requestPath.lastIndexOf('__/') + 3);
+                        let filePath = path.resolve(__dirname, '..', 'browserlib');
+                        while (depth < maxPathDepth - 1) {
+                            filePath = path.resolve(filePath, '..');
+                            depth += 1;
+                        }
+                        const file = path.resolve(filePath, filename);
+                        body = await fs.readFile(file);
+                    }
+                    await cdp.send('Fetch.fulfillRequest', {
+                        requestId,
+                        responseCode: 200,
+                        responseHeaders: [{ name: 'Content-Type', value: 'application/javascript' }],
+                        body: body.toString('base64')
+                    });
+                }
+                else {
+                    if ((request.method !== 'GET') ||
+                        (!request.url.startsWith('http:') && !request.url.startsWith('https:'))) {
+                        await cdp.send('Fetch.continueRequest', { requestId });
+                        return;
+                    }
+
+                    const response = await fetch(request.url, { signal: controller.signal });
+                    const body = await response.buffer();
+                    await cdp.send('Fetch.fulfillRequest', {
+                        requestId,
+                        responseCode: response.status,
+                        responseHeaders: Object.keys(response.headers.raw()).map(header => {
+                            return {
+                                name: header,
+                                value: response.headers.raw()[header].join(',')
+                            };
+                        }),
+                        body: body.toString('base64')
+                    });
+                }
+                //console.log(`intercept ${request.url} - done`);
+            }
+            catch (err) {
+                if (controller.signal.aborted) {
+                    // All is normal, processing was over, page and CDP session
+                    // have been closed, and network requests have been aborted
+                    // console.log(`intercept ${request.url} - aborted`);
+                    return;
+                }
+
+                // Fetch from file cache failed somehow
+                // Let Puppeteer handle the request as fallback unless
+                // calling function asked us not to do that
+                if (options.forceLocalFetch) {
+                    options.quiet ?? console.warn(`[warn] Network request for ${request.url} failed`, err);
+                    await cdp.send('Fetch.failRequest', { requestId, errorReason: 'Failed' });
+                }
+                else {
+                    options.quiet ?? console.warn(`[warn] Fall back to regular network request for ${request.url}`, err);
+                    try {
+                        await cdp.send('Fetch.continueRequest', { requestId });
+                    }
+                    catch (err) {
+                        if (!controller.signal.aborted) {
+                            options.quiet ?? console.warn(`[warn] Fall back to regular network request for ${request.url} failed`, err);
+                        }
+                    }
+                }
+            }
+        }
+    }
+
+    try {
+        const page = await browser.newPage();
+
+        // Disable cache if caller wants to handle all network requests
+        await page.setCacheEnabled(!options.forceLocalFetch);
+
+        // Intercept all network requests to use our own version of "fetch"
+        // that makes use of the local file cache.
+        const cdp = await page.target().createCDPSession();
+        await cdp.send('Fetch.enable');
+        cdp.on('Fetch.requestPaused', interceptRequest(cdp, abortController));
+
+        // Quick and dirty workaround to re-create the "window.crypto.digest"
+        // function that Respec needs (context is seen as unsecure because we're
+        // tampering with network requests)
+        await page.exposeFunction('hashdigest', (algorithm, buffer) => {
+            return crypto.createHash(algorithm).update(Buffer.from(Object.values(buffer))).digest();
+        });
+        await page.evaluateOnNewDocument(() => {
+            window.crypto.subtle = {
+                digest: function (algorithm, buffer) {
+                    const res = window.hashdigest('sha1', buffer);
+                    return res.then(buf => {
+                        return Uint8Array.from(buf.data);
+                    });
+                }
+            };
+        });
+
+        // Common loading option to give the browser enough time to load large
+        // specs, and to consider navigation done when there haven't been
+        // network connections in the past 500ms. This should be enough to
+        // handle "redirection" through JS or meta refresh (which would not
+        // have time to run if we used "load").
+        const loadOptions = {
+            timeout: 120000,
+            waitUntil: 'networkidle0'
+        };
+
+        // Load the page
+        // (note HTTP status is 0 when `file://` URLs are loaded)
+        if (spec.html) {
+            await page.setContent(spec.html, loadOptions);
+        }
+        else {
+            const result = await page.goto(spec.url, loadOptions);
+            if ((result.status() !== 200) && (!spec.url.startsWith('file://') || (result.status() !== 0))) {
+                throw new Error(`Loading ${spec.url} triggered HTTP status ${result.status()}`);
+            }
+        }
+
+        // Handle multi-page specs
+        const pageUrls = spec.pages || [];
+
+        if (pageUrls.length > 0) {
+            const pages = [];
+            for (const url of pageUrls) {
+                const subAbort = new AbortController();
+                const subPage = await browser.newPage();
+                await subPage.setCacheEnabled(!options.forceLocalFetch);
+                const subCdp = await subPage.target().createCDPSession();
+                await subCdp.send('Fetch.enable');
+                subCdp.on('Fetch.requestPaused', interceptRequest(subCdp, subAbort));
+                try {
+                    // (Note HTTP status is 0 when `file://` URLs are loaded)
+                    const subresult = await subPage.goto(url, loadOptions);
+                    if ((subresult.status() !== 200) && (!url.startsWith('file://') || (subresult.status() !== 0))) {
+                        throw new Error(`Loading ${spec.url} triggered HTTP status ${result.status()}`);
+                    }
+                    const html = await subPage.evaluate(() => {
+                        return document.body.outerHTML
+                            .replace(/<body/, '<section')
+                            .replace(/<\/body/, '</section');
+                    });
+                    pages.push({ url, html });
+                }
+                finally {
+                    subAbort.abort();
+                    await subCdp.detach();
+                    await subPage.close();
+                }
+            }
+            await page.evaluate(pages => {
+                for (const subPage of pages) {
+                    const section = document.createElement('section');
+                    section.setAttribute('data-reffy-page', subPage.url);
+                    section.innerHTML = subPage.html;
+                    document.body.appendChild(section);
+                }
+            }, pages);
+        }
+
+        // Wait until the generation of the spec is completely over
+        await page.evaluate(async () => {
+            // Detect draft CSS server hiccups as done in browser-specs:
+            // https://github.com/w3c/browser-specs/blob/b31fc0b03ba67a19162883afc30e01fcec3c600d/src/fetch-info.js#L292
+            const title = (window.document.querySelector('h1')?.textContent || '')
+                .replace(/\n/g, '').trim();
+            if (title.startsWith('Index of ')) {
+                throw new Error(`CSS server issue detected`);
+            }
+
+            const usesRespec = (window.respecConfig || window.eval('typeof respecConfig !== "undefined"')) &&
+                window.document.head.querySelector("script[src*='respec']");
+
+            function sleep(ms) {
+                return new Promise(resolve => setTimeout(resolve, ms));
+            }
+
+            async function isReady(counter) {
+                counter = counter || 0;
+                if (counter > 60) {
+                    throw new Error('Respec generation took too long');
+                }
+                if (window.document.respec?.ready) {
+                    await window.document.respec.ready;
+                }
+                else if (usesRespec) {
+                    await sleep(1000);
+                    await isReady(counter + 1);
+                }
+            }
+
+            await isReady();
+        });
+
+        // Capture and report Reffy's browserlib warnings
+        page.on('console', msg => {
+            const text = msg.text();
+            if (text.startsWith('[reffy] ')) {
+                options.quiet ?? console.warn(spec.url, `[${msg.type()}]`, msg.text().substr('[reffy] '.length));
+            }
+        });
+
+        // Capture and report when page throws an error
+        page.on('pageerror', err => {
+            options.quiet ?? console.warn(err);
+        });
+
+        // Expose additional functions defined in src/browserlib/ to the
+        // browser context, under a window.reffy namespace, so that processing
+        // script may call them. The script is an ES6 module and needs to be
+        // loaded as such.
+        // Note that we're using a fake relative URL on purpose. In practice,
+        // the request will be processed by "interceptRequest", which will
+        // respond with the contents of the script file. Also, there are
+        // multiple path levels in that fake URL on purpose as well, because
+        // scripts import the WebIDL2.js library with a URL like
+        // "../../node_modules/[...]" and may import other scripts that are
+        // higher in the folder tree.
+        await page.addScriptTag({
+            url: `reffy/scripts/${range(maxPathDepth).map(n => '__').join('/')}/reffy.mjs`,
+            type: 'module'
+        });
+
+        // Run the processFunction method in the browser context
+        const results = await page.evaluate(processFunction, ...args);
+
+        // Pending network requests may still be in the queue, flag the page
+        // as closed not to send commands on a CDP session that's no longer
+        // attached to anything
+        abortController.abort();
+
+        // Close CDP session and page
+        // Note that gets done no matter what when browser.close() gets called.
+        await cdp.detach();
+        await page.close();
+
+        return results;
+    }
+    finally {
+        // Signal abortion again (in case an exception was thrown)
+        abortController.abort();
+    }
+}
+
+
+/**
+ * Enrich the spec description with alternative URLs (versions and equivalents)
+ *
+ * TODO: The list used to contain published versions of TR specs retrieved from
+ * the W3C API. They are useful to improve the relevance of reported anomalies.
+ *
+ * @function
+ * @param {Object} spec Spec description structure (only the URL is useful)
+ * @return {Object} The same structure, enriched with the URL of the editor's
+ *   draft when one is found
+ */
+function completeWithAlternativeUrls(spec) {
+    spec.versions = new Set();
+    spec.versions.add(spec.url);
+    if (spec.release) {
+        spec.versions.add(spec.release.url);
+    }
+    if (spec.nightly) {
+        spec.versions.add(spec.nightly.url);
+    }
+    if (specEquivalents[spec.url]) {
+        spec.versions = new Set([
+            ...spec.versions,
+            ...specEquivalents[spec.url]
+        ]);
+    }
+    spec.versions = [...spec.versions];
+    return spec;
+}
+
+
+/**
+ * Returns true when the given spec is the latest "fullest" level of that spec
+ * in the given list of specs that passes the given predicate.
+ *
+ * "Fullest" means "not a delta spec, unless that is the only level that passes
+ * the predicate".
+ *
+ * @function
+ * @public
+ * @param {Object} spec Spec to check
+ * @param {Array(Object)} list List of specs (must include the spec to check)
+ * @param {function} predicate Predicate function that the spec must pass. Must
+ *   be a function that takes a spec as argument and returns a boolean.
+ * @return {Boolean} true if the spec is the latest "fullest" level in the list
+ *   that passes the predicate.
+ */
+function isLatestLevelThatPasses(spec, list, predicate) {
+    predicate = predicate || (_ => true);
+    if (!predicate(spec)) {
+        return false;
+    }
+    if (spec.seriesComposition === 'delta') {
+        while (spec.seriesPrevious) {
+            spec = list.find(s => s.shortname === spec.seriesPrevious);
+            if (!spec) {
+                break;
+            }
+            if ((spec.seriesComposition === 'full') && predicate(spec)) {
+                return false;
+            }
+        }
+        return true;
+    }
+    while (spec.seriesNext) {
+        if (!spec) {
+            break;
+        }
+        spec = list.find(s => s.shortname === spec.seriesNext);
+        if ((spec.seriesComposition === 'full') && predicate(spec)) {
+            return false;
+        }
+    }
+    return true;
+}
+
+
+/**
+ * Takes the results of a crawl for a given spec and expands it to include the
+ * contents of referenced files.
+ *
+ * The function handles both files and HTTPS resources, using either filesystem
+ * functions (for files) or fetch (for HTTPS resources).
+ *
+ * Note the spec object is expanded in place.
+ *
+ * @function
+ * @public
+ * @param {Object} spec Spec crawl result that needs to be expanded
+ * @param {string} baseFolder The base folder that contains the crawl file, or
+ *   the base HTTPS URI to resolve relative links in the crawl object.
+ * @param {Array(string)} properties An explicit list of properties to expand
+ *   (no value means "expand all possible properties")
+ * @return {Promise(object)} The promise to get an expanded crawl object that
+ *   contains the contents of referenced files and no longer references external
+ *   files (for the requested properties)
+ */
+async function expandSpecResult(spec, baseFolder, properties) {
+    baseFolder = baseFolder || '';
+    await Promise.all(Object.keys(spec).map(async property => {
+        // Only consider properties explicitly requested
+        if (properties && !properties.includes(property)) {
+            return;
+        }
+
+        // Only consider properties that link to an extract, i.e. an IDL
+        // or JSON file in subfolder.
+        if (!spec[property] ||
+                (typeof spec[property] !== 'string') ||
+                !spec[property].match(/^[^\/]+\/[^\/]+\.(json|idl)$/)) {
+            return;
+        }
+        let contents = null;
+        if (baseFolder.startsWith('https:')) {
+            const url = (new URL(spec[property], baseFolder)).toString();
+            const response = await fetch(url, { nolog: true });
+            contents = await response.text();
+        }
+        else {
+            const filename = path.join(baseFolder, spec[property]);
+            contents = await fs.readFile(filename, 'utf8');
+        }
+        if (spec[property].endsWith('.json')) {
+            contents = JSON.parse(contents);
+        }
+        if (property === 'css') {
+            // Special case for CSS where the "css" level does not exist
+            // in the generated files
+            const css = Object.assign({}, contents);
+            delete css.spec;
+            spec[property] = css;
+        }
+        else if (property === 'idl') {
+            // Special case for raw IDL extracts, which are text extracts.
+            // Also drop header that may have been added when extract was
+            // serialized.
+            if (contents.startsWith('// GENERATED CONTENT - DO NOT EDIT')) {
+                const endOfHeader = contents.indexOf('\n\n');
+                contents = contents.substring(endOfHeader + 2);
+            }
+            spec.idl = contents;
+        }
+        else {
+            spec[property] = contents[property];
+        }
+    }));
+    return spec;
+}
+
+
+/**
+ * Takes the results of a crawl (typically the contents of the index.json file)
+ * and expands it to include the contents of all referenced files.
+ *
+ * The function handles both files and HTTPS resources, using either filesystem
+ * functions (for files) or fetch (for HTTPS resources).
+ *
+ * Note the crawl object is expanded in place.
+ *
+ * @function
+ * @public
+ * @param {Object} crawl Crawl index object that needs to be expanded
+ * @param {string} baseFolder The base folder that contains the crawl file, or
+ *   the base HTTPS URI to resolve relative links in the crawl object.
+ * @param {Array(string)} properties An explicit list of properties to expand
+ *   (no value means "expand all possible properties")
+ * @return {Promise(object)} The promise to get an expanded crawl object that
+ *   contains the entire crawl report (and no longer references external files)
+ */
+async function expandCrawlResult(crawl, baseFolder, properties) {
+    baseFolder = baseFolder || '';
+    crawl.results = await Promise.all(
+        crawl.results.map(spec => expandSpecResult(spec, baseFolder, properties))
+    );
+    return crawl;
+}
+
+
+/**
+ * Retrieves the list of IDL attribute names that the CSS property generates
+ * per the CSSOM spec, see:
+ * https://drafts.csswg.org/cssom/#ref-for-css-property-to-idl-attribute
+ *
+ * @function
+ * @param {String} property CSS property name
+ * @return {Array(String)} An array of IDL attribute names, dashed attribute
+ *   first, then camel-cased attribute if different, then webkit-cased attribute
+ *   name if needed
+ */
+function getGeneratedIDLNamesByCSSProperty(property) {
+    // Converts a CSS property to an IDL attribute name per the CSSOM spec:
+    // https://drafts.csswg.org/cssom/#css-property-to-idl-attribute
+    function cssPropertyToIDLAttribute(property, lowercaseFirst) {
+        let output = '';
+        let uppercaseNext = false;
+        if (lowercaseFirst) {
+            property = property.substr(1);
+        }
+        for (const c of property) {
+            if (c === '-') {
+                uppercaseNext = true;
+            } else if (uppercaseNext) {
+                uppercaseNext = false;
+                output += c.toUpperCase();
+            } else {
+                output += c;
+            }
+        }
+        return output;
+    }
+
+    // Start with dashed attribute
+    const res = [property];
+
+    // Add camel-cased attribute if different
+    const camelCased = cssPropertyToIDLAttribute(property, false);
+    if (camelCased !== property) {
+        res.push(camelCased);
+    }
+
+    // Add webkit-cased attribute if needed
+    if (property.startsWith('-webkit-')) {
+        res.push(cssPropertyToIDLAttribute(property, true));
+    }
+
+    return res;
+};
+
+
+/**
+ * Creates the given folder if it does not exist yet.
+ *
+ * @function
+ * @public
+ * @param {String} folder Path to folder to create
+ *   (from current working directory)
+ */
+async function createFolderIfNeeded(folder) {
+    try {
+        await fs.mkdir(folder);
+    }
+    catch (err) {
+        if (err.code !== 'EEXIST') {
+            throw err;
+        }
+    }
+}
+
+
+module.exports = {
+    fetch,
+    requireFromWorkingDirectory,
+    expandBrowserModules,
+    setupBrowser,
+    teardownBrowser,
+    processSpecification,
+    completeWithAlternativeUrls,
+    isLatestLevelThatPasses,
+    expandCrawlResult,
+    expandSpecResult,
+    getGeneratedIDLNamesByCSSProperty,
+    createFolderIfNeeded
+};