| @@ -35,12 +35,181 @@ | ||
| 35 | 35 | |
| 36 | 36 | public const STATE_KEY = 'xspeed_preloader_state'; |
| 37 | 37 | public const STATE_TTL = 86400; // 24h — long enough for slow crawls. |
| 38 | 38 | public const CRON_HOOK = 'xspeed_preloader_tick'; |
| 39 | - public const USER_AGENT = 'xSpeed-Preloader/1.0 (+cache warmer; admin-initiated)'; | |
| 39 | + /** | |
| 40 | + * User-agent for every request the preloader makes. | |
| 41 | + * | |
| 42 | + * Deliberately contains no substring from the 7G/8G bad-bot lists. The | |
| 43 | + * previous value, "xSpeed-Preloader/1.0", matched the `loader` token in | |
| 44 | + * the alphabetical slice `(linkscan|linkwalker|loader|lwp-download|...)` | |
| 45 | + * — a match on "Pre*loader*" — so nginx ports of 8G answered every warm | |
| 46 | + * with 403 and newly published posts were never warmed. Upstream 8G | |
| 47 | + * v1.5 has since dropped `loader`, but forks and vendored copies (xCloud | |
| 48 | + * among them) still ship the older slice, so the name has to stay clear | |
| 49 | + * of it. "Warmer" matches nothing in either list. (#481) | |
| 50 | + * | |
| 51 | + * Read through user_agent() rather than using this constant directly, so | |
| 52 | + * the `xspeed_preloader_user_agent` filter applies. | |
| 53 | + */ | |
| 54 | + public const USER_AGENT = 'xSpeed-Warmer/1.0 (+cache warmer; admin-initiated)'; | |
| 40 | 55 | public const REQUEST_TIMEOUT = 8; |
| 41 | 56 | |
| 42 | 57 | /** |
| 58 | + * Default cap on NEW remote images resolved per warmed page. | |
| 59 | + * | |
| 60 | + * A crawl warms the cache; it is not a licence to hit third-party hosts | |
| 61 | + * hundreds of times for one page. | |
| 62 | + * | |
| 63 | + * The cap counts only images whose dimensions are not already known, and | |
| 64 | + * results persist between runs — so each crawl advances through a heavily | |
| 65 | + * embedded page rather than re-picking the same first N. That is what | |
| 66 | + * makes a cap safe here: without the skip it would strand everything past | |
| 67 | + * the limit permanently, because images appear in the same DOM order | |
| 68 | + * every time. | |
| 69 | + * | |
| 70 | + * 20 is a starting point, not a measurement. Sites that embed more can | |
| 71 | + * raise it via `xspeed_preloader_remote_dimension_limit`. | |
| 72 | + */ | |
| 73 | + private const REMOTE_DIMENSION_LIMIT = 20; | |
| 74 | + | |
| 75 | + /** | |
| 76 | + * The user-agent every preloader request sends. | |
| 77 | + * | |
| 78 | + * Filterable because the blocking rule lives on the server, not here: a | |
| 79 | + * host with its own bad-bot list can clear a warm without patching the | |
| 80 | + * plugin or waiting for a release. An empty filter return is ignored — | |
| 81 | + * sending no UA gets a request blocked at least as often. (#481) | |
| 82 | + */ | |
| 83 | + public static function user_agent(): string { | |
| 84 | + /** | |
| 85 | + * Filter the preloader's user-agent string. | |
| 86 | + * | |
| 87 | + * @param string $user_agent Default self::USER_AGENT. | |
| 88 | + */ | |
| 89 | + $ua = apply_filters( 'xspeed_preloader_user_agent', self::USER_AGENT ); | |
| 90 | + | |
| 91 | + return ( is_string( $ua ) && '' !== trim( $ua ) ) ? trim( $ua ) : self::USER_AGENT; | |
| 92 | + } | |
| 93 | + | |
| 94 | + /** | |
| 95 | + * Is this status code the signature of a firewall refusing our warmer? | |
| 96 | + * | |
| 97 | + * 403 and 406 are what bad-bot rules (7G/8G, mod_security, Wordfence) | |
| 98 | + * answer with. We only ever warm our OWN origin, and a page a visitor can | |
| 99 | + * load must be loadable by us too — so these codes mean the request was | |
| 100 | + * judged by its user-agent, not that the page is missing or broken. (#481) | |
| 101 | + */ | |
| 102 | + private static function is_firewall_block( int $code ): bool { | |
| 103 | + return in_array( $code, array( 403, 406 ), true ); | |
| 104 | + } | |
| 105 | + | |
| 106 | + /** | |
| 107 | + * Explain a warm failure in terms the admin can act on. | |
| 108 | + * | |
| 109 | + * A bare "HTTP 403" sent people hunting a broken page; the page is fine, | |
| 110 | + * and the fix is a server rule, so the message has to name the cause and | |
| 111 | + * the exact UA to allow. (#481) | |
| 112 | + */ | |
| 113 | + private static function failure_detail( int $code ): string { | |
| 114 | + if ( ! self::is_firewall_block( $code ) ) { | |
| 115 | + return sprintf( 'HTTP %d', $code ); | |
| 116 | + } | |
| 117 | + | |
| 118 | + return sprintf( | |
| 119 | + 'HTTP %d — your server\'s firewall is blocking the xSpeed cache warmer by user-agent, so this page was not warmed. Allow the user-agent "%s" (on xCloud this is the 8G firewall\'s bad-bot rule), or change it with the xspeed_preloader_user_agent filter.', | |
| 120 | + $code, | |
| 121 | + self::user_agent() | |
| 122 | + ); | |
| 123 | + } | |
| 124 | + | |
| 125 | + /** Option holding the last firewall-shaped warm refusal. */ | |
| 126 | + public const FIREWALL_BLOCK_OPTION = 'xspeed_preloader_firewall_block'; | |
| 127 | + | |
| 128 | + /** | |
| 129 | + * Record that the origin refused a warm by user-agent, for ui_notices(). | |
| 130 | + * | |
| 131 | + * An option rather than a transient: the condition is a server rule that | |
| 132 | + * persists until someone changes it, and a notice that expired on its own | |
| 133 | + * would let a site go back to never warming, silently. Cleared by | |
| 134 | + * clear_firewall_block() on the first warm that succeeds. (#481) | |
| 135 | + */ | |
| 136 | + private static function remember_firewall_block( string $url, int $code ): void { | |
| 137 | + if ( ! function_exists( 'update_option' ) ) { | |
| 138 | + return; | |
| 139 | + } | |
| 140 | + update_option( | |
| 141 | + self::FIREWALL_BLOCK_OPTION, | |
| 142 | + array( | |
| 143 | + 'url' => $url, | |
| 144 | + 'code' => $code, | |
| 145 | + 'user_agent' => self::user_agent(), | |
| 146 | + 'ts' => time(), | |
| 147 | + ), | |
| 148 | + false | |
| 149 | + ); | |
| 150 | + } | |
| 151 | + | |
| 152 | + /** Forget the firewall block once a warm gets through. */ | |
| 153 | + public static function clear_firewall_block(): void { | |
| 154 | + if ( function_exists( 'delete_option' ) && self::firewall_block() ) { | |
| 155 | + delete_option( self::FIREWALL_BLOCK_OPTION ); | |
| 156 | + } | |
| 157 | + } | |
| 158 | + | |
| 159 | + /** The last firewall-shaped refusal, or null when there isn't one. */ | |
| 160 | + public static function firewall_block(): ?array { | |
| 161 | + if ( ! function_exists( 'get_option' ) ) { | |
| 162 | + return null; | |
| 163 | + } | |
| 164 | + $block = get_option( self::FIREWALL_BLOCK_OPTION, null ); | |
| 165 | + | |
| 166 | + return ( is_array( $block ) && ! empty( $block['code'] ) ) ? $block : null; | |
| 167 | + } | |
| 168 | + | |
| 169 | + /** | |
| 170 | + * How many new remote images one warmed page may resolve. | |
| 171 | + */ | |
| 172 | + private static function remote_dimension_limit(): int { | |
| 173 | + /** | |
| 174 | + * Filter the per-page cap on remote dimension lookups. | |
| 175 | + * | |
| 176 | + * @param int $limit Default 20. Values below 1 disable the lookup. | |
| 177 | + */ | |
| 178 | + return (int) apply_filters( 'xspeed_preloader_remote_dimension_limit', self::REMOTE_DIMENSION_LIMIT ); | |
| 179 | + } | |
| 180 | + | |
| 181 | + /** | |
| 182 | + * Why the top-level sitemap fetch failed on this request, or '' when it | |
| 183 | + * succeeded. Set by fetch_sitemap_urls(), read by resolve_queue() — the | |
| 184 | + * reason has to survive the return of an empty array, which is exactly | |
| 185 | + * what it could not do before. Request-scoped; never persisted. (#142) | |
| 186 | + * | |
| 187 | + * @var string | |
| 188 | + */ | |
| 189 | + private static $last_sitemap_error = ''; | |
| 190 | + | |
| 191 | + /** | |
| 192 | + * The sitemap URL the last error refers to. Kept beside the message so | |
| 193 | + * an error entry can carry a real `url` field like every other one, | |
| 194 | + * rather than repeating the URL already inside the message text. | |
| 195 | + */ | |
| 196 | + private static $last_sitemap_url = ''; | |
| 197 | + | |
| 198 | + /** | |
| 199 | + * How the queue for the current crawl was built — 'sitemap', 'fallback' | |
| 200 | + * (enumerated from the database because the sitemap was unreachable), or | |
| 201 | + * 'none'. Surfaced in the state so the panel, REST and CLI can each say | |
| 202 | + * what actually happened instead of reporting a bare zero. (#142) | |
| 203 | + * | |
| 204 | + * @var string | |
| 205 | + */ | |
| 206 | + private static $queue_source = 'none'; | |
| 207 | + | |
| 208 | + /** Largest number of URLs the database fallback will enumerate. */ | |
| 209 | + private const FALLBACK_LIMIT = 500; | |
| 210 | + | |
| 211 | + /** | |
| 43 | 212 | * Kick off a fresh crawl. Returns the initial state. |
| 44 | 213 | */ |
| 45 | 214 | public static function start(): array { |
| 46 | 215 | $opts = Settings_Manager::get( 'preloader' ); |
| @@ -45,25 +214,72 @@ | ||
| 45 | 214 | public static function start(): array { |
| 46 | 215 | $opts = Settings_Manager::get( 'preloader' ); |
| 47 | 216 | $urls = self::resolve_queue( $opts ); |
| 48 | 217 | |
| 218 | + // A crawl that queued nothing because the sitemap was unreachable is a | |
| 219 | + // FAILURE, and every layer above needs to be able to say so. It used | |
| 220 | + // to be indistinguishable from success: errors stayed empty, the REST | |
| 221 | + // route returned 200, and the CLI printed a green Success. (#142) | |
| 222 | + $sitemap_error = self::$last_sitemap_error; | |
| 223 | + $errors = array(); | |
| 224 | + if ( '' !== $sitemap_error && empty( $urls ) ) { | |
| 225 | + // Same {url, error, ts} shape every other entry uses. A bare | |
| 226 | + // string here fataled `wp xspeed preloader status`, which | |
| 227 | + // destructures `$e['url']` over the list — and took the MCP | |
| 228 | + // `get_preloader_status` tool down with it, so an agent asking | |
| 229 | + // why the preload failed got "Cannot access offset of type | |
| 230 | + // string on string" instead of the reason this code records. | |
| 231 | + // `url` is the sitemap because that is what failed. (QA F1) | |
| 232 | + $errors[] = array( | |
| 233 | + 'url' => self::$last_sitemap_url, | |
| 234 | + 'error' => $sitemap_error, | |
| 235 | + 'ts' => time(), | |
| 236 | + ); | |
| 237 | + } | |
| 238 | + | |
| 49 | 239 | $state = array( |
| 50 | - 'running' => ! empty( $urls ), | |
| 51 | - 'started_at' => time(), | |
| 52 | - 'finished_at' => 0, | |
| 53 | - 'queue' => array_values( $urls ), | |
| 54 | - 'processed' => 0, | |
| 55 | - 'total' => count( $urls ), | |
| 56 | - 'last_url' => '', | |
| 57 | - 'errors' => array(), | |
| 240 | + 'running' => ! empty( $urls ), | |
| 241 | + 'started_at' => time(), | |
| 242 | + 'finished_at' => empty( $urls ) ? time() : 0, | |
| 243 | + 'queue' => array_values( $urls ), | |
| 244 | + 'processed' => 0, | |
| 245 | + 'total' => count( $urls ), | |
| 246 | + 'last_url' => '', | |
| 247 | + 'errors' => $errors, | |
| 248 | + // Consumers render on these: the panel needs to distinguish | |
| 249 | + // "not started" from "ran and found nothing", and to tell the | |
| 250 | + // user when the queue came from the fallback rather than the | |
| 251 | + // sitemap they configured. | |
| 252 | + 'source' => self::$queue_source, | |
| 253 | + 'sitemap_error' => $sitemap_error, | |
| 58 | 254 | ); |
| 59 | 255 | set_transient( self::STATE_KEY, $state, self::STATE_TTL ); |
| 60 | 256 | |
| 61 | - Activity_Log::record( | |
| 62 | - 'preloader_started', | |
| 63 | - sprintf( 'Preloader queued %d URL%s for warming.', $state['total'], 1 === $state['total'] ? '' : 's' ), | |
| 64 | - $state['total'] > 0 ? Activity_Log::INFO : Activity_Log::WARN | |
| 65 | - ); | |
| 257 | + if ( '' !== $sitemap_error && 'fallback' === self::$queue_source ) { | |
| 258 | + $message = sprintf( | |
| 259 | + /* translators: 1: number of URLs, 2: the sitemap failure detail. */ | |
| 260 | + __( 'Preloader queued %1$d URLs from the site content — %2$s', 'xspeed' ), | |
| 261 | + $state['total'], | |
| 262 | + $sitemap_error | |
| 263 | + ); | |
| 264 | + $severity = Activity_Log::WARN; | |
| 265 | + } elseif ( '' !== $sitemap_error ) { | |
| 266 | + $message = sprintf( | |
| 267 | + /* translators: %s: the sitemap failure detail. */ | |
| 268 | + __( 'Preloader could not start — %s', 'xspeed' ), | |
| 269 | + $sitemap_error | |
| 270 | + ); | |
| 271 | + $severity = Activity_Log::WARN; | |
| 272 | + } else { | |
| 273 | + $message = sprintf( | |
| 274 | + /* translators: 1: number of URLs, 2: plural suffix. */ | |
| 275 | + __( 'Preloader queued %1$d URL%2$s for warming.', 'xspeed' ), | |
| 276 | + $state['total'], | |
| 277 | + 1 === $state['total'] ? '' : 's' | |
| 278 | + ); | |
| 279 | + $severity = $state['total'] > 0 ? Activity_Log::INFO : Activity_Log::WARN; | |
| 280 | + } | |
| 281 | + Activity_Log::record( 'preloader_started', $message, $severity ); | |
| 66 | 282 | |
| 67 | 283 | // Schedule the first tick ~5 seconds out so the kick-off REST call |
| 68 | 284 | // returns instantly; wp_schedule_single_event covers the |
| 69 | 285 | // "process the queue ASAP" path without a heavy synchronous loop. |
| @@ -173,9 +389,9 @@ | ||
| 173 | 389 | $url, |
| 174 | 390 | array( |
| 175 | 391 | 'timeout' => self::REQUEST_TIMEOUT, |
| 176 | 392 | 'sslverify' => false, |
| 177 | - 'user-agent' => self::USER_AGENT, | |
| 393 | + 'user-agent' => self::user_agent(), | |
| 178 | 394 | 'blocking' => true, |
| 179 | 395 | ) |
| 180 | 396 | ); |
| 181 | 397 | if ( is_wp_error( $response ) ) { |
| @@ -189,13 +405,19 @@ | ||
| 189 | 405 | $code = (int) wp_remote_retrieve_response_code( $response ); |
| 190 | 406 | if ( $code >= 400 ) { |
| 191 | 407 | Activity_Log::record( |
| 192 | 408 | 'preloader_warm_failed', |
| 193 | - sprintf( 'Warm %s failed (%s): HTTP %d', $cause, $url, $code ), | |
| 409 | + sprintf( 'Warm %s failed (%s): %s', $cause, $url, self::failure_detail( $code ) ), | |
| 194 | 410 | Activity_Log::WARN |
| 195 | 411 | ); |
| 412 | + if ( self::is_firewall_block( $code ) ) { | |
| 413 | + self::remember_firewall_block( $url, $code ); | |
| 414 | + } | |
| 196 | 415 | return false; |
| 197 | 416 | } |
| 417 | + // A warm that got through proves the firewall is no longer refusing us, | |
| 418 | + // so the notice must go — otherwise it outlives the problem. (#481) | |
| 419 | + self::clear_firewall_block(); | |
| 198 | 420 | Activity_Log::record( |
| 199 | 421 | 'preloader_warmed_one', |
| 200 | 422 | sprintf( 'Warmed %s (%s)', $url, $cause ), |
| 201 | 423 | Activity_Log::INFO |
| @@ -208,9 +430,9 @@ | ||
| 208 | 430 | $url, |
| 209 | 431 | array( |
| 210 | 432 | 'timeout' => self::REQUEST_TIMEOUT, |
| 211 | 433 | 'sslverify' => false, |
| 212 | - 'user-agent' => self::USER_AGENT, | |
| 434 | + 'user-agent' => self::user_agent(), | |
| 213 | 435 | 'headers' => array( |
| 214 | 436 | 'Accept' => 'text/html,application/xhtml+xml', |
| 215 | 437 | ), |
| 216 | 438 | 'blocking' => true, |
| @@ -230,15 +452,110 @@ | ||
| 230 | 452 | $code = (int) wp_remote_retrieve_response_code( $response ); |
| 231 | 453 | if ( $code >= 400 ) { |
| 232 | 454 | $state['errors'][] = array( |
| 233 | 455 | 'url' => $url, |
| 234 | - 'error' => sprintf( 'HTTP %d', $code ), | |
| 456 | + 'error' => self::failure_detail( $code ), | |
| 235 | 457 | 'ts' => time(), |
| 236 | 458 | ); |
| 237 | 459 | $state['errors'] = array_slice( $state['errors'], -20 ); |
| 460 | + if ( self::is_firewall_block( $code ) ) { | |
| 461 | + self::remember_firewall_block( $url, $code ); | |
| 462 | + } | |
| 463 | + return; | |
| 238 | 464 | } |
| 465 | + | |
| 466 | + self::clear_firewall_block(); | |
| 467 | + self::warm_remote_dimensions( (string) wp_remote_retrieve_body( $response ) ); | |
| 239 | 468 | } |
| 240 | 469 | |
| 470 | + /** | |
| 471 | + * Resolve dimensions for externally hosted images found on a warmed page. | |
| 472 | + * | |
| 473 | + * The crawl already has the HTML in hand, so harvesting image URLs from it | |
| 474 | + * costs nothing extra — and this is the one place where paying for a | |
| 475 | + * remote lookup is free of consequence, because no visitor is waiting. | |
| 476 | + * | |
| 477 | + * An image on another domain has no local file to measure, so the front | |
| 478 | + * end skips it and the page ships without width/height — which is layout | |
| 479 | + * shift, on precisely the sites least able to fix it by hand (a CDN, a | |
| 480 | + * sister site, a shared asset host). Warming here means the NEXT render | |
| 481 | + * finds the dimensions in cache and stamps them, with the visitor paying | |
| 482 | + * nothing. | |
| 483 | + * | |
| 484 | + * Deliberately bounded per page: a crawl should not turn into a scraper | |
| 485 | + * for a page embedding hundreds of third-party images. | |
| 486 | + * | |
| 487 | + * @param string $html The warmed page's HTML. | |
| 488 | + */ | |
| 489 | + private static function warm_remote_dimensions( string $html ): void { | |
| 490 | + if ( '' === $html || ! class_exists( '\XSpeed\Lazy_Loader' ) ) { | |
| 491 | + return; | |
| 492 | + } | |
| 493 | + | |
| 494 | + $opts = Settings_Manager::get( 'lazy' ); | |
| 495 | + if ( empty( $opts['add_missing_dimensions'] ) ) { | |
| 496 | + return; | |
| 497 | + } | |
| 498 | + | |
| 499 | + // Match any <img>, not only one carrying `src`. The URL worth warming | |
| 500 | + // may live in a lazy attribute instead — which is the whole point of | |
| 501 | + // #328 — and resolvable_image_url() below is what knows where to look. | |
| 502 | + if ( ! preg_match_all( '#<img\b[^>]*>#i', $html, $m, PREG_SET_ORDER ) ) { | |
| 503 | + return; | |
| 504 | + } | |
| 505 | + | |
| 506 | + $home = wp_parse_url( home_url(), PHP_URL_HOST ); | |
| 507 | + $targets = array(); | |
| 508 | + foreach ( $m as $tag ) { | |
| 509 | + // Only tags MISSING a dimension are worth resolving — one that | |
| 510 | + // already declares both needs nothing. | |
| 511 | + // Same lookbehind as Lazy_Loader::ensure_dimensions(): a bare | |
| 512 | + // `\bwidth=` also matches `data-width=`, so a slider carrying its | |
| 513 | + // own metadata looked already-sized and was skipped from warming. | |
| 514 | + // The two must agree, or the collector skips exactly the tags the | |
| 515 | + // renderer still needs measured. (#333 review round 3, issue 2) | |
| 516 | + if ( preg_match( '#(?<![-\w])width\s*=#i', $tag[0] ) && preg_match( '#(?<![-\w])height\s*=#i', $tag[0] ) ) { | |
| 517 | + continue; | |
| 518 | + } | |
| 519 | + // Ask the same resolver the render path uses, rather than reading | |
| 520 | + // `src` directly. A slider parks a spacer in `src` and the real | |
| 521 | + // URL in `data-lazy`/`data-src`/`data-original`, so a collector | |
| 522 | + // looking only at `src` warmed the SPACER and never the image — | |
| 523 | + // leaving remotely-hosted slider images unresolvable at render | |
| 524 | + // time, the exact markup #328 is about. (#333 review round 2, | |
| 525 | + // issue 3) | |
| 526 | + // `false`: do not let the resolver settle a name-refused URL by | |
| 527 | + // MEASURING it. That is circular here — remote measurement is | |
| 528 | + // gated until warm_dimensions() sets $warming, and this collector | |
| 529 | + // is what feeds warm_dimensions(). Take the URL the tag offers and | |
| 530 | + // let the warm pass decide. (#333 review round 3, issue 3) | |
| 531 | + $src = Lazy_Loader::resolvable_image_url( $tag[0], false ); | |
| 532 | + if ( '' === $src || ! preg_match( '#^https?://#i', $src ) ) { | |
| 533 | + continue; | |
| 534 | + } | |
| 535 | + $host = wp_parse_url( $src, PHP_URL_HOST ); | |
| 536 | + if ( ! $host || $host === $home ) { | |
| 537 | + continue; // Local images already resolve from disk. | |
| 538 | + } | |
| 539 | + // Already resolved (or already known unresolvable) — looking it up | |
| 540 | + // again costs a request and teaches us nothing. Skipping it is | |
| 541 | + // also what makes the cap below advance: images appear in the same | |
| 542 | + // DOM order every crawl, so a collector that did not skip would | |
| 543 | + // re-pick the same first N for ever and never reach the rest. | |
| 544 | + if ( Lazy_Loader::dimensions_known( $src ) ) { | |
| 545 | + continue; | |
| 546 | + } | |
| 547 | + $targets[ $src ] = true; | |
| 548 | + if ( count( $targets ) >= self::remote_dimension_limit() ) { | |
| 549 | + break; | |
| 550 | + } | |
| 551 | + } | |
| 552 | + | |
| 553 | + if ( $targets ) { | |
| 554 | + Lazy_Loader::warm_dimensions( array_keys( $targets ) ); | |
| 555 | + } | |
| 556 | + } | |
| 557 | + | |
| 241 | 558 | private static function mark_complete( array $state ): void { |
| 242 | 559 | $state['running'] = false; |
| 243 | 560 | $state['finished_at'] = time(); |
| 244 | 561 | $state['queue'] = array(); |
| @@ -263,8 +580,14 @@ | ||
| 263 | 580 | * |
| 264 | 581 | * @return string[] |
| 265 | 582 | */ |
| 266 | 583 | private static function resolve_queue( array $opts ): array { |
| 584 | + // Request-scoped statics: reset so a previous crawl in the same | |
| 585 | + // process can't leak its verdict into this one. | |
| 586 | + self::$last_sitemap_error = ''; | |
| 587 | + self::$last_sitemap_url = ''; | |
| 588 | + self::$queue_source = 'none'; | |
| 589 | + | |
| 267 | 590 | $sitemap = trim( (string) ( $opts['sitemap_url'] ?? '' ) ); |
| 268 | 591 | if ( '' === $sitemap ) { |
| 269 | 592 | $sitemap = home_url( '/wp-sitemap.xml' ); |
| 270 | 593 | } |
| @@ -269,9 +592,28 @@ | ||
| 269 | 592 | $sitemap = home_url( '/wp-sitemap.xml' ); |
| 270 | 593 | } |
| 271 | 594 | |
| 272 | 595 | $urls = self::fetch_sitemap_urls( $sitemap, 0 ); |
| 596 | + $from = empty( $urls ) ? 'none' : 'sitemap'; | |
| 273 | 597 | |
| 598 | + /* | |
| 599 | + * A missing sitemap must not disable the feature. Two very common | |
| 600 | + * setups produce one with no misconfiguration by the user: | |
| 601 | + * `blog_public = 0` (WordPress disables /wp-sitemap.xml outright, | |
| 602 | + * standard on staging and pre-launch sites), and an SEO plugin | |
| 603 | + * filtering `wp_sitemaps_enabled` to false while serving its own | |
| 604 | + * sitemap at a path we were never told about. | |
| 605 | + * | |
| 606 | + * Enumerate warmable URLs straight from the database instead. Only | |
| 607 | + * on a genuine fetch FAILURE — a sitemap that is reachable and | |
| 608 | + * legitimately empty is a real answer, and silently crawling | |
| 609 | + * something else would be worse than doing nothing. (#142) | |
| 610 | + */ | |
| 611 | + if ( empty( $urls ) && '' !== self::$last_sitemap_error ) { | |
| 612 | + $urls = self::fallback_urls(); | |
| 613 | + $from = empty( $urls ) ? 'none' : 'fallback'; | |
| 614 | + } | |
| 615 | + | |
| 274 | 616 | $cache_opts = Settings_Manager::get( 'cache' ); |
| 275 | 617 | $excluded = is_array( $cache_opts['excluded_urls'] ?? null ) ? $cache_opts['excluded_urls'] : array(); |
| 276 | 618 | if ( ! empty( $excluded ) ) { |
| 277 | 619 | $urls = array_filter( |
| @@ -289,12 +631,90 @@ | ||
| 289 | 631 | } |
| 290 | 632 | |
| 291 | 633 | // Dedup + cap at 5000 to bound the transient size on huge sites. |
| 292 | 634 | $urls = array_values( array_unique( $urls ) ); |
| 293 | - return array_slice( $urls, 0, 5000 ); | |
| 635 | + $urls = array_slice( $urls, 0, 5000 ); | |
| 636 | + | |
| 637 | + /* | |
| 638 | + * Commit the verdict only now, AFTER the exclusion filter — the | |
| 639 | + * source describes what we ACTUALLY queued, not what we hoped to. | |
| 640 | + * Setting it earlier let a queue that the exclusions stripped to | |
| 641 | + * nothing still claim `fallback`, so the panel announced "Warmed | |
| 642 | + * from site content" over 0 URLs, and `crawlFailed` (which needs | |
| 643 | + * source !== 'fallback' at total 0) could never become true. | |
| 644 | + * One assignment fixes both. (QA F2/F3 on #155) | |
| 645 | + */ | |
| 646 | + self::$queue_source = empty( $urls ) ? 'none' : $from; | |
| 647 | + | |
| 648 | + return $urls; | |
| 294 | 649 | } |
| 295 | 650 | |
| 296 | 651 | /** |
| 652 | + * Enumerate warmable URLs from the database, for sites whose sitemap | |
| 653 | + * can't be fetched. Deliberately modest in scope: the home page, then | |
| 654 | + * the most recently modified public posts across every public post type. | |
| 655 | + * Newest-first is the right bias — those are the URLs most likely to be | |
| 656 | + * requested and least likely to be warm already. | |
| 657 | + * | |
| 658 | + * Uses WP_Query rather than SQL so post-type registration, status | |
| 659 | + * handling and multisite switching all behave the way the rest of | |
| 660 | + * WordPress does. | |
| 661 | + * | |
| 662 | + * @return string[] | |
| 663 | + */ | |
| 664 | + private static function fallback_urls(): array { | |
| 665 | + $urls = array(); | |
| 666 | + $home = (string) home_url( '/' ); | |
| 667 | + if ( '' !== trim( $home, '/' ) ) { | |
| 668 | + $urls[] = $home; | |
| 669 | + } | |
| 670 | + | |
| 671 | + $types = get_post_types( | |
| 672 | + array( | |
| 673 | + 'public' => true, | |
| 674 | + 'publicly_queryable' => true, | |
| 675 | + ) | |
| 676 | + ); | |
| 677 | + // `page` is public but not publicly_queryable, so the query above | |
| 678 | + // misses it — and pages are exactly what a warm cache wants most. | |
| 679 | + // Only add it when the site has post types at all: an empty list | |
| 680 | + // means there is nothing to enumerate, and constructing a WP_Query | |
| 681 | + // for it would be wasted work. | |
| 682 | + if ( ! empty( $types ) ) { | |
| 683 | + $types['page'] = 'page'; | |
| 684 | + unset( $types['attachment'] ); | |
| 685 | + } | |
| 686 | + | |
| 687 | + if ( empty( $types ) || ! class_exists( '\WP_Query' ) ) { | |
| 688 | + return $urls; | |
| 689 | + } | |
| 690 | + | |
| 691 | + $query = new \WP_Query( | |
| 692 | + array( | |
| 693 | + 'post_type' => array_values( $types ), | |
| 694 | + 'post_status' => 'publish', | |
| 695 | + 'posts_per_page' => self::FALLBACK_LIMIT, | |
| 696 | + 'orderby' => 'modified', | |
| 697 | + 'order' => 'DESC', | |
| 698 | + 'ignore_sticky_posts' => true, | |
| 699 | + 'no_found_rows' => true, | |
| 700 | + 'update_post_meta_cache' => false, | |
| 701 | + 'update_post_term_cache' => false, | |
| 702 | + 'fields' => 'ids', | |
| 703 | + ) | |
| 704 | + ); | |
| 705 | + | |
| 706 | + foreach ( $query->posts as $post_id ) { | |
| 707 | + $permalink = get_permalink( (int) $post_id ); | |
| 708 | + if ( is_string( $permalink ) && '' !== $permalink ) { | |
| 709 | + $urls[] = $permalink; | |
| 710 | + } | |
| 711 | + } | |
| 712 | + | |
| 713 | + return array_values( array_unique( $urls ) ); | |
| 714 | + } | |
| 715 | + | |
| 716 | + /** | |
| 297 | 717 | * Recursive sitemap parser. Depth-limited to 3 so a maliciously |
| 298 | 718 | * deep index can't stack-overflow. |
| 299 | 719 | */ |
| 300 | 720 | private static function fetch_sitemap_urls( string $sitemap_url, int $depth ): array { |
| @@ -305,12 +725,39 @@ | ||
| 305 | 725 | $sitemap_url, |
| 306 | 726 | array( |
| 307 | 727 | 'timeout' => self::REQUEST_TIMEOUT, |
| 308 | 728 | 'sslverify' => false, |
| 309 | - 'user-agent' => self::USER_AGENT, | |
| 729 | + 'user-agent' => self::user_agent(), | |
| 310 | 730 | ) |
| 311 | 731 | ); |
| 312 | - if ( is_wp_error( $res ) || (int) wp_remote_retrieve_response_code( $res ) >= 400 ) { | |
| 732 | + if ( is_wp_error( $res ) ) { | |
| 733 | + // Record WHY, don't just vanish. "Unreachable" and "valid but | |
| 734 | + // empty" both used to collapse into an empty array here, which is | |
| 735 | + // what made a sitemap-less site look like a successful crawl of | |
| 736 | + // zero URLs. Only the top-level fetch is recorded: a nested index | |
| 737 | + // failing is a partial result, not a dead crawl. (#142) | |
| 738 | + if ( 0 === $depth ) { | |
| 739 | + self::$last_sitemap_error = sprintf( | |
| 740 | + /* translators: 1: sitemap URL, 2: error detail. */ | |
| 741 | + __( 'Could not fetch the sitemap at %1$s — %2$s', 'xspeed' ), | |
| 742 | + $sitemap_url, | |
| 743 | + $res->get_error_message() | |
| 744 | + ); | |
| 745 | + self::$last_sitemap_url = (string) $sitemap_url; | |
| 746 | + } | |
| 747 | + return array(); | |
| 748 | + } | |
| 749 | + $code = (int) wp_remote_retrieve_response_code( $res ); | |
| 750 | + if ( $code >= 400 ) { | |
| 751 | + if ( 0 === $depth ) { | |
| 752 | + self::$last_sitemap_error = sprintf( | |
| 753 | + /* translators: 1: sitemap URL, 2: HTTP status code. */ | |
| 754 | + __( 'Could not fetch the sitemap at %1$s — the server returned HTTP %2$d.', 'xspeed' ), | |
| 755 | + $sitemap_url, | |
| 756 | + $code | |
| 757 | + ); | |
| 758 | + self::$last_sitemap_url = (string) $sitemap_url; | |
| 759 | + } | |
| 313 | 760 | return array(); |
| 314 | 761 | } |
| 315 | 762 | $body = (string) wp_remote_retrieve_body( $res ); |
| 316 | 763 | if ( '' === $body ) { |
