Remove ML-powered queries

This commit is contained in:
Henry Mercer 2023-10-02 17:20:50 +01:00
parent a370ce344f
commit ebbadee09e
33 changed files with 60 additions and 836 deletions

97
lib/config-utils.js generated
View file

@ -23,7 +23,7 @@ var __importStar = (this && this.__importStar) || function (mod) {
return result;
};
Object.defineProperty(exports, "__esModule", { value: true });
exports.wrapEnvironment = exports.generateRegistries = exports.downloadPacks = exports.getConfig = exports.getPathToParsedConfigFile = exports.initConfig = exports.getMlPoweredJsQueriesStatus = exports.parsePacks = exports.validatePackSpecification = exports.parsePacksSpecification = exports.parsePacksFromConfig = exports.calculateAugmentation = exports.getDefaultConfig = exports.getRawLanguages = exports.getLanguageAliases = exports.getLanguages = exports.getLanguagesInRepo = exports.getUnknownLanguagesError = exports.getNoLanguagesError = exports.getConfigFileDirectoryGivenMessage = exports.getConfigFileFormatInvalidMessage = exports.getConfigFileRepoFormatInvalidMessage = exports.getConfigFileDoesNotExistErrorMessage = exports.getConfigFileOutsideWorkspaceErrorMessage = exports.getLocalPathDoesNotExist = exports.getLocalPathOutsideOfRepository = exports.getPacksStrInvalid = exports.getPacksInvalid = exports.getPacksInvalidSplit = exports.getPathsInvalid = exports.getPathsIgnoreInvalid = exports.getQueryUsesInvalid = exports.getQueriesMissingUses = exports.getQueriesInvalid = exports.getDisableDefaultQueriesInvalid = exports.getNameInvalid = exports.validateAndSanitisePath = exports.defaultAugmentationProperties = void 0;
exports.wrapEnvironment = exports.generateRegistries = exports.downloadPacks = exports.getConfig = exports.getPathToParsedConfigFile = exports.initConfig = exports.parsePacks = exports.validatePackSpecification = exports.parsePacksSpecification = exports.parsePacksFromConfig = exports.calculateAugmentation = exports.getDefaultConfig = exports.getRawLanguages = exports.getLanguageAliases = exports.getLanguages = exports.getLanguagesInRepo = exports.getUnknownLanguagesError = exports.getNoLanguagesError = exports.getConfigFileDirectoryGivenMessage = exports.getConfigFileFormatInvalidMessage = exports.getConfigFileRepoFormatInvalidMessage = exports.getConfigFileDoesNotExistErrorMessage = exports.getConfigFileOutsideWorkspaceErrorMessage = exports.getLocalPathDoesNotExist = exports.getLocalPathOutsideOfRepository = exports.getPacksStrInvalid = exports.getPacksInvalid = exports.getPacksInvalidSplit = exports.getPathsInvalid = exports.getPathsIgnoreInvalid = exports.getQueryUsesInvalid = exports.getQueriesMissingUses = exports.getQueriesInvalid = exports.getDisableDefaultQueriesInvalid = exports.getNameInvalid = exports.validateAndSanitisePath = exports.defaultAugmentationProperties = void 0;
const fs = __importStar(require("fs"));
const path = __importStar(require("path"));
const perf_hooks_1 = require("perf_hooks");
@ -45,13 +45,12 @@ const PATHS_IGNORE_PROPERTY = "paths-ignore";
const PATHS_PROPERTY = "paths";
const PACKS_PROPERTY = "packs";
/**
* The default, empty augmentation properties. This is most useeful
* The default, empty augmentation properties. This is most useful
* for tests.
*/
exports.defaultAugmentationProperties = {
queriesInputCombines: false,
packsInputCombines: false,
injectedMlQueries: false,
packsInput: undefined,
queriesInput: undefined,
};
@ -139,10 +138,8 @@ const builtinSuites = [
/**
* Determine the set of queries associated with suiteName's suites and add them to resultMap.
* Throws an error if suiteName is not a valid builtin suite.
* May inject ML queries, and the return value will declare if this was done.
*/
async function addBuiltinSuiteQueries(languages, codeQL, resultMap, packs, suiteName, features, configFile) {
let injectedMlQueries = false;
async function addBuiltinSuiteQueries(languages, codeQL, resultMap, suiteName, configFile) {
const found = builtinSuites.find((suite) => suite === suiteName);
if (!found) {
throw new util_1.UserError(getQueryUsesInvalid(configFile, suiteName));
@ -153,27 +150,8 @@ async function addBuiltinSuiteQueries(languages, codeQL, resultMap, packs, suite
${codeql_1.CODEQL_VERSION_SECURITY_EXPERIMENTAL_SUITE}. Please upgrade to CodeQL CLI version
${codeql_1.CODEQL_VERSION_SECURITY_EXPERIMENTAL_SUITE} or later.`);
}
// If we're running the JavaScript security-extended analysis (or a superset of it), the repo is
// opted into the ML-powered queries beta, and a user hasn't already added the ML-powered query
// pack, then add the ML-powered query pack so that we run ML-powered queries.
if (languages.includes("javascript") &&
(found === "security-experimental" ||
found === "security-extended" ||
found === "security-and-quality") &&
!packs.javascript?.some(isMlPoweredJsQueriesPack) &&
(await features.getValue(feature_flags_1.Feature.MlPoweredQueriesEnabled, codeQL))) {
if (!packs.javascript) {
packs.javascript = [];
}
packs.javascript.push(await (0, util_1.getMlPoweredJsQueriesPack)(codeQL));
injectedMlQueries = true;
}
const suites = languages.map((l) => `${l}-${suiteName}.qls`);
await runResolveQueries(codeQL, resultMap, suites, undefined);
return injectedMlQueries;
}
function isMlPoweredJsQueriesPack(pack) {
return parsePacksSpecification(pack).name === util_1.ML_POWERED_JS_QUERIES_PACK_NAME;
}
/**
* Retrieve the set of queries at localQueryPath and add them to resultMap.
@ -230,13 +208,8 @@ async function addRemoteQueries(codeQL, resultMap, queryUses, tempDir, apiDetail
* parsing the 'uses' actions in the workflow file. So it can handle
* local paths starting with './', or references to remote repos, or
* a finite set of hardcoded terms for builtin suites.
*
* This may inject ML queries into the packs to use, and the return value will
* declare if this was done.
*
* @returns whether or not we injected ML queries into the packs
*/
async function parseQueryUses(languages, codeQL, resultMap, packs, queryUses, tempDir, workspacePath, apiDetails, features, logger, configFile) {
async function parseQueryUses(languages, codeQL, resultMap, queryUses, tempDir, workspacePath, apiDetails, features, logger, configFile) {
queryUses = queryUses.trim();
if (queryUses === "") {
throw new util_1.UserError(getQueryUsesInvalid(configFile));
@ -244,11 +217,12 @@ async function parseQueryUses(languages, codeQL, resultMap, packs, queryUses, te
// Check for the local path case before we start trying to parse the repository name
if (queryUses.startsWith("./")) {
await addLocalQueries(codeQL, resultMap, queryUses.slice(2), workspacePath, configFile);
return false;
return;
}
// Check for one of the builtin suites
if (queryUses.indexOf("/") === -1 && queryUses.indexOf("@") === -1) {
return await addBuiltinSuiteQueries(languages, codeQL, resultMap, packs, queryUses, features, configFile);
await addBuiltinSuiteQueries(languages, codeQL, resultMap, queryUses, configFile);
return;
}
// Otherwise, must be a reference to another repo.
// If config parsing is handled in CLI, then this repo will be downloaded
@ -256,7 +230,6 @@ async function parseQueryUses(languages, codeQL, resultMap, packs, queryUses, te
if (!(await (0, feature_flags_1.useCodeScanningConfigInCli)(codeQL, features))) {
await addRemoteQueries(codeQL, resultMap, queryUses, tempDir, apiDetails, logger, configFile);
}
return false;
}
// Regex validating stars in paths or paths-ignore entries.
// The intention is to only allow ** to appear when immediately
@ -516,16 +489,13 @@ async function getRawLanguages(languagesInput, repository, logger) {
return { rawLanguages, autodetected };
}
exports.getRawLanguages = getRawLanguages;
async function addQueriesAndPacksFromWorkflow(codeQL, queriesInput, languages, resultMap, packs, tempDir, workspacePath, apiDetails, features, logger) {
let injectedMlQueries = false;
async function addQueriesAndPacksFromWorkflow(codeQL, queriesInput, languages, resultMap, tempDir, workspacePath, apiDetails, features, logger) {
queriesInput = queriesInput.trim();
// "+" means "don't override config file" - see shouldAddConfigFileQueries
queriesInput = queriesInput.replace(/^\+/, "");
for (const query of queriesInput.split(",")) {
const didInject = await parseQueryUses(languages, codeQL, resultMap, packs, query, tempDir, workspacePath, apiDetails, features, logger);
injectedMlQueries = injectedMlQueries || didInject;
await parseQueryUses(languages, codeQL, resultMap, query, tempDir, workspacePath, apiDetails, features, logger);
}
return injectedMlQueries;
}
// Returns true if either no queries were provided in the workflow.
// or if the queries in the workflow were provided in "additive" mode,
@ -557,8 +527,7 @@ async function getDefaultConfig(languagesInput, rawQueriesInput, rawPacksInput,
}
: {};
if (rawQueriesInput) {
augmentationProperties.injectedMlQueries =
await addQueriesAndPacksFromWorkflow(codeQL, rawQueriesInput, languages, queries, packs, tempDir, workspacePath, apiDetails, features, logger);
await addQueriesAndPacksFromWorkflow(codeQL, rawQueriesInput, languages, queries, tempDir, workspacePath, apiDetails, features, logger);
}
const { trapCaches, trapCacheDownloadTime } = await downloadCacheWithTime(trapCachingEnabled, codeQL, languages, logger);
return {
@ -641,8 +610,7 @@ async function loadConfig(languagesInput, rawQueriesInput, rawPacksInput, config
// unless they're prefixed with "+", in which case they supplement those
// in the config file.
if (rawQueriesInput) {
augmentationProperties.injectedMlQueries =
await addQueriesAndPacksFromWorkflow(codeQL, rawQueriesInput, languages, queries, packs, tempDir, workspacePath, apiDetails, features, logger);
await addQueriesAndPacksFromWorkflow(codeQL, rawQueriesInput, languages, queries, tempDir, workspacePath, apiDetails, features, logger);
}
if (shouldAddConfigFileQueries(rawQueriesInput) &&
QUERIES_PROPERTY in parsedYAML) {
@ -654,7 +622,7 @@ async function loadConfig(languagesInput, rawQueriesInput, rawPacksInput, config
if (typeof query[QUERIES_USES_PROPERTY] !== "string") {
throw new util_1.UserError(getQueriesMissingUses(configFile));
}
await parseQueryUses(languages, codeQL, queries, packs, query[QUERIES_USES_PROPERTY], tempDir, workspacePath, apiDetails, features, logger, configFile);
await parseQueryUses(languages, codeQL, queries, query[QUERIES_USES_PROPERTY], tempDir, workspacePath, apiDetails, features, logger, configFile);
}
}
if (PATHS_IGNORE_PROPERTY in parsedYAML) {
@ -724,7 +692,6 @@ function calculateAugmentation(rawPacksInput, rawQueriesInput, languages) {
const queriesInputCombines = shouldCombine(rawQueriesInput);
const queriesInput = parseQueriesFromInput(rawQueriesInput, queriesInputCombines);
return {
injectedMlQueries: false,
packsInputCombines,
packsInput: packsInput?.[languages[0]],
queriesInput,
@ -932,46 +899,6 @@ function combinePacks(packs1, packs2) {
}
return packs;
}
/**
* Get information about ML-powered JS queries to populate status reports with.
*
* This will be:
*
* - The version string if the analysis is using a single version of the ML-powered query pack.
* - "latest" if the version string of the ML-powered query pack is undefined. This is unlikely to
* occur in practice (see comment below).
* - "false" if the analysis won't run any ML-powered JS queries.
* - "other" in all other cases.
*
* Our goal of the status report here is to allow us to compare the occurrence of timeouts and other
* errors with ML-powered queries turned on and off. We also want to be able to compare minor
* version bumps caused by us bumping the version range of `ML_POWERED_JS_QUERIES_PACK` in a new
* version of the CodeQL Action. For instance, we might want to compare the `~0.1.0` and `~0.0.2`
* version strings.
*
* This function lives here rather than in `init-action.ts` so it's easier to test, since tests for
* `init-action.ts` would each need to live in their own file. See `analyze-action-env.ts` for an
* explanation as to why this is.
*/
function getMlPoweredJsQueriesStatus(config) {
const mlPoweredJsQueryPacks = (config.packs.javascript || [])
.map((p) => parsePacksSpecification(p))
.filter((pack) => pack.name === util_1.ML_POWERED_JS_QUERIES_PACK_NAME && !pack.path);
switch (mlPoweredJsQueryPacks.length) {
case 1:
// We should always specify an explicit version string in `getMlPoweredJsQueriesPack`,
// otherwise we won't be able to make changes to the pack unless those changes are compatible
// with each version of the CodeQL Action. Therefore in practice we should only hit the
// `latest` case here when customers have explicitly added the ML-powered query pack to their
// CodeQL config.
return mlPoweredJsQueryPacks[0].version || "latest";
case 0:
return "false";
default:
return "other";
}
}
exports.getMlPoweredJsQueriesStatus = getMlPoweredJsQueriesStatus;
function dbLocationOrDefault(dbLocation, tempDir) {
return dbLocation || path.resolve(tempDir, "codeql_databases");
}