react-native 0.87.0-rc.1 → 0.87.0-rc.2

package/scripts/spm/spm-pbxproj.js

@@ -0,0 +1,654 @@
+/**
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ *
+ * This source code is licensed under the MIT license found in the
+ * LICENSE file in the root directory of this source tree.
+ *
+ * @flow strict-local
+ * @format
+ */
+
+'use strict';
+
+const crypto = require('crypto');
+
+/**
+ * Generate a deterministic 24-hex-character UUID from a seed string.
+ * SHA-256 truncated to 24 chars (standard Xcode pbxproj UUID length). Not a
+ * security use — the hash only provides stable, collision-unlikely IDs — but
+ * sha256 keeps static analysis (CodeQL weak-crypto) quiet.
+ */
+function generateUUID(seed /*: string */) /*: string */ {
+  return crypto
+    .createHash('sha256')
+    .update(seed)
+    .digest('hex')
+    .substring(0, 24)
+    .toUpperCase();
+}
+
+/**
+ * Escapes a string for OpenStep plist format if needed.
+ */
+function quoteIfNeeded(s /*: string */) /*: string */ {
+  if (/^[a-zA-Z0-9._/]+$/.test(s)) {
+    return s;
+  }
+  return `"${s.replace(/\\/g, '\\\\').replace(/"/g, '\\"').replace(/\n/g, '\\n')}"`;
+}
+
+/**
+ * Serialize a single pbxproj object entry to its OpenStep text form,
+ * including the leading `\t\t<uuid>` and trailing `};` but NO trailing
+ * newline. Short entries (≤3 scalar fields) collapse to one line, matching
+ * Xcode's own formatting. Used by the in-place injector to splice single
+ * entries into an existing project.
+ */
+function serializeEntry(
+  entry /*: {readonly uuid: string, readonly comment?: ?string, readonly fields: {readonly [string]: string}, ...} */,
+) /*: string */ {
+  const comment =
+    entry.comment != null && entry.comment !== ''
+      ? ` /* ${entry.comment} */`
+      : '';
+  let out = `\t\t${entry.uuid}${comment} = {`;
+  const fieldKeys = Object.keys(entry.fields);
+  if (
+    fieldKeys.length <= 3 &&
+    !fieldKeys.some(k => entry.fields[k].includes('\n'))
+  ) {
+    // Single-line format for short entries
+    out += fieldKeys.map(k => `${k} = ${entry.fields[k]};`).join(' ');
+    out += '};';
+  } else {
+    out += '\n';
+    for (const key of fieldKeys) {
+      out += `\t\t\t${key} = ${entry.fields[key]};\n`;
+    }
+    out += '\t\t};';
+  }
+  return out;
+}
+
+// ---------------------------------------------------------------------------
+// Surgical in-place pbxproj editing.
+//
+// To ADD SPM packages to a user's EXISTING project.pbxproj we splice new
+// objects and array members into the existing text by string anchors, leaving
+// every untouched byte identical (so the git diff is just the added lines).
+// These helpers operate on the raw OpenStep text — there is no AST. Quote-aware
+// delimiter matching lets them skip over field values (e.g. a shellScript
+// containing braces/parens) without miscounting.
+// ---------------------------------------------------------------------------
+
+/**
+ * Derive a deterministic UUID for an injected object, namespaced by the host
+ * project's root-object UUID so it is (a) stable across re-runs (idempotency)
+ * and (b) astronomically unlikely to collide with the user's existing
+ * randomly-assigned 24-hex IDs. `salt` lets the caller re-derive on the
+ * ~1-in-2^96 collision.
+ */
+function namespacedUUID(
+  rootUUID /*: string */,
+  section /*: string */,
+  id /*: string */,
+  salt /*: string */ = '',
+) /*: string */ {
+  return generateUUID(`${rootUUID}:spm${salt}:${section}:${id}`);
+}
+
+function escapeRegExp(s /*: string */) /*: string */ {
+  return s.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+}
+
+/**
+ * Given an index pointing at an opening `"`, return the index of the matching
+ * closing `"` (honoring backslash escapes).
+ */
+function scanString(text /*: string */, openIdx /*: number */) /*: number */ {
+  for (let i = openIdx + 1; i < text.length; i++) {
+    const c = text[i];
+    if (c === '\\') {
+      i++;
+      continue;
+    }
+    if (c === '"') {
+      return i;
+    }
+  }
+  throw new Error('pbxproj: unterminated string literal');
+}
+
+/**
+ * Given an index pointing at an opening `{` or `(`, return the index of the
+ * matching close delimiter. Nesting counts both brace and paren forms; quoted
+ * strings are skipped. Well-formed OpenStep never mismatches the two forms.
+ */
+function scanToClose(text /*: string */, openIdx /*: number */) /*: number */ {
+  let depth = 0;
+  for (let i = openIdx; i < text.length; i++) {
+    const c = text[i];
+    if (c === '"') {
+      i = scanString(text, i);
+      continue;
+    }
+    if (c === '{' || c === '(') {
+      depth++;
+    } else if (c === '}' || c === ')') {
+      depth--;
+      if (depth === 0) {
+        return i;
+      }
+    }
+  }
+  throw new Error('pbxproj: unbalanced delimiters');
+}
+
+/*::
+type ObjectRange = {uuid: string, bodyOpen: number, bodyClose: number};
+// Any object whose body range is known — field accessors only need the body
+// bounds, so they accept the richer shapes callers carry (e.g. app targets
+// with a name, or buildSettings dicts) inexactly.
+type BodyRange = {bodyOpen: number, bodyClose: number, ...};
+type FieldRange = {matchStart: number, valueStart: number, value: string, tokenEnd: number};
+*/
+
+/**
+ * Locate the object with the given 24-hex UUID. Returns the index of the body
+ * `{` and its matching `}`. Matches both single-line and multi-line entries.
+ */
+function findObjectByUuid(
+  text /*: string */,
+  uuid /*: string */,
+) /*: ObjectRange | null */ {
+  const m = new RegExp(`\\n\\t*${uuid}\\b[^\\n]*?= \\{`).exec(text);
+  if (m == null) {
+    return null;
+  }
+  const bodyOpen = text.indexOf('{', m.index);
+  const bodyClose = scanToClose(text, bodyOpen);
+  return {uuid, bodyOpen, bodyClose};
+}
+
+/**
+ * Find a field within a multi-line object body (`\n\t+key = value;`). Returns
+ * the value token range (value excludes the trailing `;`; `tokenEnd` points AT
+ * the `;`). Containers (`( … )` / `{ … }`) and quoted strings are matched as a
+ * whole. Returns null when the key is absent.
+ */
+function findField(
+  text /*: string */,
+  obj /*: BodyRange */,
+  key /*: string */,
+) /*: FieldRange | null */ {
+  const body = text.slice(obj.bodyOpen, obj.bodyClose);
+  const m = new RegExp(`\\n\\t+${escapeRegExp(key)} = `).exec(body);
+  if (m == null) {
+    return null;
+  }
+  const matchStart = obj.bodyOpen + m.index;
+  const valueStart = matchStart + m[0].length;
+  const fc = text[valueStart];
+  let tokenEnd;
+  if (fc === '(' || fc === '{') {
+    tokenEnd = scanToClose(text, valueStart) + 1;
+  } else if (fc === '"') {
+    tokenEnd = scanString(text, valueStart) + 1;
+  } else {
+    tokenEnd = text.indexOf(';', valueStart);
+  }
+  return {
+    matchStart,
+    valueStart,
+    value: text.slice(valueStart, tokenEnd),
+    tokenEnd,
+  };
+}
+
+/** Locate the `/* Begin X section *​/ … /* End X section *​/` byte range. */
+function findSection(
+  text /*: string */,
+  name /*: string */,
+) /*: {begin: number, contentStart: number, end: number} | null */ {
+  const beginTag = `/* Begin ${name} section */`;
+  const endTag = `/* End ${name} section */`;
+  const begin = text.indexOf(beginTag);
+  const end = text.indexOf(endTag);
+  if (begin < 0 || end < 0) {
+    return null;
+  }
+  return {begin, contentStart: begin + beginTag.length, end};
+}
+
+/** The PBXProject root object (via the trailing `rootObject = <uuid>;`). */
+function findProjectObject(text /*: string */) /*: ObjectRange | null */ {
+  const m = /\n\trootObject = ([0-9A-Fa-f]{24})/.exec(text);
+  if (m == null) {
+    return null;
+  }
+  return findObjectByUuid(text, m[1]);
+}
+
+/**
+ * Every PBXNativeTarget whose productType is an application. Returns uuid +
+ * name + body range for each. Used to pick the app target to inject into
+ * (and to refuse on ambiguity).
+ */
+function findApplicationTargets(
+  text /*: string */,
+) /*: Array<{uuid: string, name: string, bodyOpen: number, bodyClose: number}> */ {
+  const section = findSection(text, 'PBXNativeTarget');
+  if (section == null) {
+    return [];
+  }
+  const out = [];
+  const re = /\n\t\t([0-9A-Fa-f]{24})(?: \/\* (.*?) \*\/)? = \{/g;
+  re.lastIndex = section.contentStart;
+  for (;;) {
+    const m = re.exec(text);
+    if (m == null || m.index >= section.end) {
+      break;
+    }
+    const uuid = m[1];
+    const comment = m[2];
+    const bodyOpen = text.indexOf('{', m.index);
+    const bodyClose = scanToClose(text, bodyOpen);
+    const obj = {uuid, bodyOpen, bodyClose};
+    const productType = findField(text, obj, 'productType');
+    if (
+      productType != null &&
+      /com\.apple\.product-type\.application/.test(productType.value)
+    ) {
+      const nameField = findField(text, obj, 'name');
+      const name =
+        nameField != null
+          ? nameField.value.replace(/^"|"$/g, '')
+          : (comment ?? uuid);
+      out.push({uuid, name, bodyOpen, bodyClose});
+    }
+    re.lastIndex = bodyClose;
+  }
+  return out;
+}
+
+/** UUIDs already referenced inside a `( … )` array field value. */
+function uuidsInArray(value /*: string */) /*: Set<string> */ {
+  const found = new Set /*:: <string> */();
+  const re = /\b([0-9A-Fa-f]{24})\b/g;
+  for (;;) {
+    const m = re.exec(value);
+    if (m == null) {
+      break;
+    }
+    found.add(m[1]);
+  }
+  return found;
+}
+
+/**
+ * The leading-tab indent of fields inside an object body (e.g. `\t\t\t` for a
+ * top-level object, `\t\t\t\t` for a nested dict like buildSettings). Used so
+ * inserted fields/members match the surrounding depth at any nesting level.
+ */
+function detectFieldIndent(
+  text /*: string */,
+  obj /*: BodyRange */,
+) /*: string */ {
+  const m = /\n(\t+)\S/.exec(text.slice(obj.bodyOpen, obj.bodyClose));
+  return m != null ? m[1] : '\t\t\t';
+}
+
+/**
+ * Insert one or more already-serialized object entries (text produced by
+ * serializeEntry, no surrounding newlines) into the named section — created
+ * just before the close of the `objects` dict if the section is absent.
+ */
+function insertObjectsIntoSection(
+  text /*: string */,
+  sectionName /*: string */,
+  entriesText /*: string */,
+) /*: string */ {
+  const section = findSection(text, sectionName);
+  if (section != null) {
+    return (
+      text.slice(0, section.end) + entriesText + '\n' + text.slice(section.end)
+    );
+  }
+  // No such section yet — create it just before the `objects` dict closes.
+  const anchor = '\n\t};\n\trootObject = ';
+  const at = text.indexOf(anchor);
+  if (at < 0) {
+    throw new Error('pbxproj: could not find end of objects dict');
+  }
+  const block =
+    `/* Begin ${sectionName} section */\n${entriesText}\n` +
+    `/* End ${sectionName} section */\n\n`;
+  return text.slice(0, at + 1) + block + text.slice(at + 1);
+}
+
+/**
+ * Append members to a `( … )` array field, deduping by UUID. Creates the field
+ * (with a `$(inherited)`-free literal list) after the object's opening `{` when
+ * absent. `members` are `{uuid, comment}`. Indentation is derived from the
+ * object so it works for top-level fields and nested dicts alike.
+ */
+function addArrayMembers(
+  text /*: string */,
+  obj /*: BodyRange */,
+  key /*: string */,
+  members /*: ReadonlyArray<{readonly uuid: string, readonly comment?: ?string, ...}> */,
+  options /*: {prepend?: boolean} */ = {},
+) /*: string */ {
+  const fieldIndent = detectFieldIndent(text, obj);
+  const memberIndent = fieldIndent + '\t';
+  const line = (
+    m /*: {readonly uuid: string, readonly comment?: ?string, ...} */,
+  ) =>
+    `${memberIndent}${m.uuid}${m.comment != null && m.comment !== '' ? ` /* ${m.comment} */` : ''},\n`;
+
+  const field = findField(text, obj, key);
+  if (field != null) {
+    const existing = uuidsInArray(field.value);
+    const fresh = members.filter(m => !existing.has(m.uuid));
+    if (fresh.length === 0) {
+      return text;
+    }
+    // Prepend: insert right after the array's opening `(\n` so the new members
+    // run first (used for the sync phase, which must precede Sources).
+    const insertAt =
+      options.prepend === true
+        ? text.indexOf('\n', field.valueStart) + 1
+        : text.lastIndexOf('\n', field.tokenEnd - 1) + 1;
+    return (
+      text.slice(0, insertAt) + fresh.map(line).join('') + text.slice(insertAt)
+    );
+  }
+  const block = `\n${fieldIndent}${key} = (\n${members.map(line).join('')}${fieldIndent});`;
+  return text.slice(0, obj.bodyOpen + 1) + block + text.slice(obj.bodyOpen + 1);
+}
+
+/**
+ * Append raw string values to a `( … )` array build-setting (e.g.
+ * OTHER_LDFLAGS), deduping by exact token. Creates the setting seeded with
+ * `"$(inherited)"` when absent. Values must already be plist-quoted by caller.
+ */
+function addArrayStringValues(
+  text /*: string */,
+  obj /*: BodyRange */,
+  key /*: string */,
+  values /*: Array<string> */,
+) /*: string */ {
+  const fieldIndent = detectFieldIndent(text, obj);
+  const memberIndent = fieldIndent + '\t';
+  const arrayBlock = (members /*: Array<string> */) =>
+    `(\n${members.map(v => `${memberIndent}${v},\n`).join('')}${fieldIndent})`;
+
+  const field = findField(text, obj, key);
+  if (field != null) {
+    // Dedup by EXACT existing member, not substring — a substring check would
+    // treat `"-ObjC"` as already present when only `"-ObjCFoo"` is there (and
+    // vice-versa). Parse the current members (array `( … )` or bare scalar).
+    const existingMembers = new Set(
+      field.value
+        .replace(/^\s*\(/, '')
+        .replace(/\)\s*$/, '')
+        .split(',')
+        .map(s => s.trim())
+        .filter(s => s.length > 0),
+    );
+    const fresh = values.filter(v => !existingMembers.has(v));
+    if (fresh.length === 0) {
+      return text;
+    }
+    if (field.value.trimStart().startsWith('(')) {
+      // Existing array — splice fresh members before the closing `)`.
+      const lineStart = text.lastIndexOf('\n', field.tokenEnd - 1) + 1;
+      const lines = fresh.map(v => `${memberIndent}${v},\n`).join('');
+      return text.slice(0, lineStart) + lines + text.slice(lineStart);
+    }
+    // Existing scalar — promote to an array preserving the prior value.
+    const replacement = arrayBlock([
+      '"$(inherited)"',
+      field.value.trim(),
+      ...fresh,
+    ]);
+    return (
+      text.slice(0, field.valueStart) + replacement + text.slice(field.tokenEnd)
+    );
+  }
+  const block = `\n${fieldIndent}${key} = ${arrayBlock(['"$(inherited)"', ...values])};`;
+  return text.slice(0, obj.bodyOpen + 1) + block + text.slice(obj.bodyOpen + 1);
+}
+
+/**
+ * Add a scalar field after the object's `{` only when ABSENT (never clobbers a
+ * value the user already set). Returns text unchanged if the key exists.
+ */
+function ensureScalarField(
+  text /*: string */,
+  obj /*: BodyRange */,
+  key /*: string */,
+  value /*: string */,
+) /*: string */ {
+  if (findField(text, obj, key) != null) {
+    return text;
+  }
+  const fieldIndent = detectFieldIndent(text, obj);
+  const block = `\n${fieldIndent}${key} = ${value};`;
+  return text.slice(0, obj.bodyOpen + 1) + block + text.slice(obj.bodyOpen + 1);
+}
+
+/**
+ * Set a scalar field's value, UNLIKE ensureScalarField this overwrites an
+ * existing value in place rather than leaving it alone — used by fields the
+ * injector itself owns (e.g. the generated `shellScript`) that must be kept
+ * in sync on re-injection. When the field is present, only the value token
+ * (`findField`'s `valueStart..tokenEnd` range — the trailing `;` is NOT part
+ * of that range and is preserved untouched) is replaced in place, so field
+ * order never shifts and passing the same `value` again yields
+ * byte-identical output. Falls back to `ensureScalarField`'s append-after-`{`
+ * behavior when the field is absent.
+ */
+function setScalarField(
+  text /*: string */,
+  obj /*: BodyRange */,
+  key /*: string */,
+  value /*: string */,
+) /*: string */ {
+  const field = findField(text, obj, key);
+  if (field != null) {
+    return text.slice(0, field.valueStart) + value + text.slice(field.tokenEnd);
+  }
+  return ensureScalarField(text, obj, key, value);
+}
+// ---------------------------------------------------------------------------
+// Surgical removal — the inverse of the additive helpers above. `deinit` uses
+// these to undo exactly what injection added, leaving every other byte (incl.
+// user edits made after injection) untouched. All are pure string transforms.
+// ---------------------------------------------------------------------------
+
+/**
+ * Remove the object whose UUID is `uuid` (its whole `\t\t<uuid> … = { … };`
+ * entry, single- or multi-line). No-op when the object is absent.
+ */
+function removeObjectByUuid(
+  text /*: string */,
+  uuid /*: string */,
+) /*: string */ {
+  const obj = findObjectByUuid(text, uuid);
+  if (obj == null) {
+    return text;
+  }
+  // Start at the newline preceding the entry's line; end just past its `;`.
+  // Leaving the trailing newline in place preserves it as the next entry's
+  // separator (byte-identical to never having inserted the line).
+  const start = text.lastIndexOf('\n', obj.bodyOpen);
+  let end = obj.bodyClose + 1; // past `}`
+  if (text[end] === ';') {
+    end++;
+  }
+  return text.slice(0, start) + text.slice(end);
+}
+
+/**
+ * Remove array-member lines (`\n\t+<uuid> /* … *​/,`) referencing any of
+ * `uuids` from every `( … )` list in the file (packageReferences,
+ * packageProductDependencies, a Frameworks phase's `files`, buildPhases, …).
+ * Only matches member lines (trailing comma), never the object-definition line
+ * (which ends in `= {`), so it composes safely with removeObjectByUuid.
+ */
+function removeArrayMembersByUuid(
+  text /*: string */,
+  uuids /*: ReadonlyArray<string> */,
+) /*: string */ {
+  let out = text;
+  for (const uuid of uuids) {
+    out = out.replace(
+      new RegExp(`\\n[\\t ]*${escapeRegExp(uuid)}\\b[^\\n]*,`, 'g'),
+      '',
+    );
+  }
+  return out;
+}
+
+/** Remove a whole `\n\t+key = value;` field from `obj`. No-op when absent. */
+function removeField(
+  text /*: string */,
+  obj /*: BodyRange */,
+  key /*: string */,
+) /*: string */ {
+  const f = findField(text, obj, key);
+  if (f == null) {
+    return text;
+  }
+  // f.matchStart points at the leading `\n`; f.tokenEnd points AT the `;`.
+  return text.slice(0, f.matchStart) + text.slice(f.tokenEnd + 1);
+}
+
+/**
+ * Remove specific raw string members from an existing `( … )` array field
+ * (inverse of addArrayStringValues' append branch). Leaves the field and any
+ * other members in place. No-op when the field or a value is absent.
+ */
+function removeArrayStringValues(
+  text /*: string */,
+  obj /*: BodyRange */,
+  key /*: string */,
+  values /*: ReadonlyArray<string> */,
+) /*: string */ {
+  const f = findField(text, obj, key);
+  if (f == null) {
+    return text;
+  }
+  let region = text.slice(f.valueStart, f.tokenEnd);
+  for (const val of values) {
+    region = region.replace(new RegExp(`\\n[\\t ]*${escapeRegExp(val)},`), '');
+  }
+  return text.slice(0, f.valueStart) + region + text.slice(f.tokenEnd);
+}
+
+/**
+ * Remove the empty `Pods` PBXGroup that `pod deintegrate` can leave behind in
+ * the navigator (build integration is already gone — xcconfigs/[CP] phases/
+ * linking — but the group lingers). Removes the group object AND its membership
+ * in any parent group. Only acts when the group is EMPTY (`children = ()`), so a
+ * still-integrated project (non-empty Pods group) is never touched. No-op when
+ * absent. PBXGroup bodies contain no nested braces, so `[^{}]` body matching is
+ * safe.
+ */
+function removeEmptyPodsGroup(text /*: string */) /*: string */ {
+  const m = /\n[\t ]*([0-9A-Fa-f]{24}) \/\* Pods \*\/ = \{[^{}]*?\};/.exec(
+    text,
+  );
+  if (m == null) {
+    return text;
+  }
+  const block = m[0];
+  if (
+    !/isa = PBXGroup;/.test(block) ||
+    !/children = \(\s*\);/.test(block) ||
+    !/\b(?:path|name) = Pods;/.test(block)
+  ) {
+    return text;
+  }
+  const uuid = m[1];
+  // Drop the parent group's child reference first, then the group object.
+  return removeObjectByUuid(removeArrayMembersByUuid(text, [uuid]), uuid);
+}
+
+/**
+ * Remove the dangling `JavaScriptCore.framework` PBXFileReference the
+ * community template has carried since RN 0.60 (an SDK framework reference
+ * left over from the pre-Hermes JSC-via-CocoaPods era). It's navigator-only —
+ * never wired into any PBXBuildFile/build phase — so it's meaningless (and
+ * confusing) now that React Native uses Hermes. Only removes references that
+ * are truly unlinked: if a PBXBuildFile still references the UUID (e.g. an
+ * app that deliberately links JSC, such as via react-native-javascriptcore),
+ * that reference is left completely untouched. Removes the file-reference
+ * object AND its membership in any parent group. No-op when absent. Like
+ * `removeEmptyPodsGroup`, `deinit` does not restore this removal — git is the
+ * safety net.
+ */
+function removeDanglingJavaScriptCoreRef(text /*: string */) /*: string */ {
+  const re =
+    /\n[\t ]*([0-9A-Fa-f]{24}) \/\* JavaScriptCore\.framework \*\/ = \{[^{}]*?\};/g;
+  const uuidsToRemove = [];
+  for (const m of text.matchAll(re)) {
+    const block = m[0];
+    if (
+      !/isa = PBXFileReference;/.test(block) ||
+      !/path = System\/Library\/Frameworks\/JavaScriptCore\.framework;/.test(
+        block,
+      ) ||
+      !/sourceTree = SDKROOT;/.test(block)
+    ) {
+      continue;
+    }
+    const uuid = m[1];
+    const linked = new RegExp(`\\bfileRef = ${escapeRegExp(uuid)}\\b`).test(
+      text,
+    );
+    if (!linked) {
+      uuidsToRemove.push(uuid);
+    }
+  }
+  if (uuidsToRemove.length === 0) {
+    return text;
+  }
+  // Drop the parent group's child reference(s) first, then the file objects.
+  let out = removeArrayMembersByUuid(text, uuidsToRemove);
+  for (const uuid of uuidsToRemove) {
+    out = removeObjectByUuid(out, uuid);
+  }
+  return out;
+}
+
+module.exports = {
+  generateUUID,
+  namespacedUUID,
+  serializeEntry,
+  quoteIfNeeded,
+  // Surgical-edit toolkit (in-place injection):
+  scanString,
+  scanToClose,
+  findObjectByUuid,
+  findField,
+  findSection,
+  findProjectObject,
+  findApplicationTargets,
+  uuidsInArray,
+  detectFieldIndent,
+  insertObjectsIntoSection,
+  addArrayMembers,
+  addArrayStringValues,
+  ensureScalarField,
+  setScalarField,
+  escapeRegExp,
+  // Surgical removal (deinit):
+  removeObjectByUuid,
+  removeArrayMembersByUuid,
+  removeField,
+  removeArrayStringValues,
+  removeEmptyPodsGroup,
+  removeDanglingJavaScriptCoreRef,
+};