Refresh helper-runtime prelink cache and gate release-readiness (#24302) - #24321
Merged
Conversation
The committed core_fingerprint no longer matched live inputs, so all 257 units were skipped on clean checkout and hello-world compile timed out. Re-emitted 293 units in the pinned Docker env and added a --strict gate to release-readiness so staleness is visible instead of silently costing users. Co-authored-by: Cursor <cursoragent@cursor.com>
This was referenced Jul 28, 2026
PurHur
added a commit
that referenced
this pull request
Jul 28, 2026
…24349) DOM and introspection helper units drifted after #24321; release-readiness --strict failed with 258 fresh / 35 stale. Re-emit via --prelink in the pinned Docker env so the committed cache matches live fingerprints again. Co-authored-by: PurHur <PurHur@users.noreply.github.com> Co-authored-by: Cursor <cursoragent@cursor.com>
PurHur
added a commit
that referenced
this pull request
Jul 28, 2026
…nt (#24302) (#24351) Compiling <?php echo "hi\n"; on a clean checkout took ~9 minutes. Second compile: 5 seconds. warmForUserAotBuild() decided whether to emit the entire helper corpus based on one thing — whether a marker file exists: $marker = self::coreMarkerPath(); // build/helper-runtime-cache/<prefix><fingerprint>.ok if (is_file($marker)) { return; } ... $rc = self::runWarmupCommand($cmd); // shells out to emit-helper-runtime-object.php coreMarkerPath() lives under cacheDir() = build/helper-runtime-cache, which is gitignored (.gitignore:29, 0 files tracked). A clean checkout can therefore NEVER have the marker, so the first user AOT build re-emitted the whole corpus regardless of whether prelinked/helper-runtime/ was present and current. That is also why #24321 (refreshing the committed cache) did not help: it fixes what the CONSUMPTION path can use, while the warmup decision never looked at it. Measured before 533s, after 517s — ~16s of ~530. Found by profiling, not by guessing: a SIGALRM sampler on the real bin/compile.php collected only 5 ticks in 441s, top path HelperRuntimeCache::runWarmupCommand < warmForUserAotBuild. The parent is blocked on a subprocess, which is why parent-side sampling sees nothing — the near-empty profile was itself the signal. Fix: skip the warmup when the committed per-arch cache exists, carries a core_fingerprint equal to the live one, and actually contains units. The consumption path already reads prelinkedUnitsDir() alongside unitsDir(), so in that state there is nothing to emit and warming is pure waste. first compile, clean checkout: 517s -> 5s committedCacheIsCurrent() is deliberately conservative: missing manifest, unreadable JSON, mismatched fingerprint or an empty units directory all fall through to the warmup, so a stale or partial committed cache is still emitted rather than silently producing a build with missing helpers. Gate (no CI on lib/, and AOT is informational in compiler-gate.yml), branch vs a clean master checkout, run sequentially, both with build/helper-runtime-cache removed first so the measurement is COLD: cold aot-smoke : 8/8 both sides VM --repeat 2 : 110/110 both sides, exit 0 AOT --repeat 3 : master 16 failing, branch 16 failing regressions (branch-only): NONE fixed (master-only) : NONE Note on method: my first attempt at that comparison read master's sweep while it was still running (73 of 110 cases) and showed four phantom regressions — j05, j06, k02, m02 — purely because those cases had not been reached yet. The numbers above are from the completed run. Co-authored-by: PurHur <tedyyyyy@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
3 tasks
PurHur
added a commit
that referenced
this pull request
Jul 28, 2026
…24349) DOM and introspection helper units drifted after #24321; release-readiness --strict failed with 258 fresh / 35 stale. Re-emit via --prelink in the pinned Docker env so the committed cache matches live fingerprints again. Co-authored-by: PurHur <PurHur@users.noreply.github.com> Co-authored-by: Cursor <cursoragent@cursor.com>
PurHur
added a commit
that referenced
this pull request
Jul 28, 2026
…nt (#24302) (#24351) Compiling <?php echo "hi\n"; on a clean checkout took ~9 minutes. Second compile: 5 seconds. warmForUserAotBuild() decided whether to emit the entire helper corpus based on one thing — whether a marker file exists: $marker = self::coreMarkerPath(); // build/helper-runtime-cache/<prefix><fingerprint>.ok if (is_file($marker)) { return; } ... $rc = self::runWarmupCommand($cmd); // shells out to emit-helper-runtime-object.php coreMarkerPath() lives under cacheDir() = build/helper-runtime-cache, which is gitignored (.gitignore:29, 0 files tracked). A clean checkout can therefore NEVER have the marker, so the first user AOT build re-emitted the whole corpus regardless of whether prelinked/helper-runtime/ was present and current. That is also why #24321 (refreshing the committed cache) did not help: it fixes what the CONSUMPTION path can use, while the warmup decision never looked at it. Measured before 533s, after 517s — ~16s of ~530. Found by profiling, not by guessing: a SIGALRM sampler on the real bin/compile.php collected only 5 ticks in 441s, top path HelperRuntimeCache::runWarmupCommand < warmForUserAotBuild. The parent is blocked on a subprocess, which is why parent-side sampling sees nothing — the near-empty profile was itself the signal. Fix: skip the warmup when the committed per-arch cache exists, carries a core_fingerprint equal to the live one, and actually contains units. The consumption path already reads prelinkedUnitsDir() alongside unitsDir(), so in that state there is nothing to emit and warming is pure waste. first compile, clean checkout: 517s -> 5s committedCacheIsCurrent() is deliberately conservative: missing manifest, unreadable JSON, mismatched fingerprint or an empty units directory all fall through to the warmup, so a stale or partial committed cache is still emitted rather than silently producing a build with missing helpers. Gate (no CI on lib/, and AOT is informational in compiler-gate.yml), branch vs a clean master checkout, run sequentially, both with build/helper-runtime-cache removed first so the measurement is COLD: cold aot-smoke : 8/8 both sides VM --repeat 2 : 110/110 both sides, exit 0 AOT --repeat 3 : master 16 failing, branch 16 failing regressions (branch-only): NONE fixed (master-only) : NONE Note on method: my first attempt at that comparison read master's sweep while it was still running (73 of 110 cases) and showed four phantom regressions — j05, j06, k02, m02 — purely because those cases had not been reached yet. The numbers above are from the completed run. Co-authored-by: PurHur <tedyyyyy@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
PurHur
added a commit
that referenced
this pull request
Jul 28, 2026
* json_decode: native HT runtime materialization for assoc arrays (#24137) Replace int-wire JsonDecodeJitHelper with tag-dispatched native __hashtable__* building via phpc_native_ht_* (ParseStr #13827 shape), wired through a single __compiler_json_decode bridge with ensureNativeHtInternalProxies before NestedJIT. AOT green for runtime literal JSON strings; json_encode→decode roundtrip still blocked (encode boxes __string__* in __value__* — separate issue path). Co-authored-by: Cursor <cursoragent@cursor.com> * json_decode: raise shrink line budget; add encode→decode AOT diag (#24137) StringJsonDecode grew to 409 lines with native HT materialization; bump the shrink ceiling. Add a small AOT diag repro for the NestedJIT heap-string param gap that still blocks json_encode→json_decode. Co-authored-by: Cursor <cursoragent@cursor.com> * AOT json_decode: box assoc HT in __value__* (#24137) Raw __hashtable__* cast to __value__* was misread as int/NULL after NestedJIT native HT fill. Peer #24298 boxHashtableValue shape. Co-authored-by: Cursor <cursoragent@cursor.com> * Foundation: refresh helper-runtime prelink (35 stale → 0) (#24336) (#24349) DOM and introspection helper units drifted after #24321; release-readiness --strict failed with 258 fresh / 35 stale. Re-emit via --prelink in the pinned Docker env so the committed cache matches live fingerprints again. Co-authored-by: PurHur <PurHur@users.noreply.github.com> Co-authored-by: Cursor <cursoragent@cursor.com> * DOM: getNamedItem matches Attr local name, not QName (#24332) (#24350) php-src namednodemap.c uses xmlHasProp (local name) for attribute maps. Wire JIT/AOT instance-method proxies for getNamedItem(NS) so :object attributes receivers bind the same path. Co-authored-by: PurHur <PurHur@users.noreply.github.com> Co-authored-by: Cursor <cursoragent@cursor.com> * AOT: skip the corpus warmup when the committed prelink cache is current (#24302) (#24351) Compiling <?php echo "hi\n"; on a clean checkout took ~9 minutes. Second compile: 5 seconds. warmForUserAotBuild() decided whether to emit the entire helper corpus based on one thing — whether a marker file exists: $marker = self::coreMarkerPath(); // build/helper-runtime-cache/<prefix><fingerprint>.ok if (is_file($marker)) { return; } ... $rc = self::runWarmupCommand($cmd); // shells out to emit-helper-runtime-object.php coreMarkerPath() lives under cacheDir() = build/helper-runtime-cache, which is gitignored (.gitignore:29, 0 files tracked). A clean checkout can therefore NEVER have the marker, so the first user AOT build re-emitted the whole corpus regardless of whether prelinked/helper-runtime/ was present and current. That is also why #24321 (refreshing the committed cache) did not help: it fixes what the CONSUMPTION path can use, while the warmup decision never looked at it. Measured before 533s, after 517s — ~16s of ~530. Found by profiling, not by guessing: a SIGALRM sampler on the real bin/compile.php collected only 5 ticks in 441s, top path HelperRuntimeCache::runWarmupCommand < warmForUserAotBuild. The parent is blocked on a subprocess, which is why parent-side sampling sees nothing — the near-empty profile was itself the signal. Fix: skip the warmup when the committed per-arch cache exists, carries a core_fingerprint equal to the live one, and actually contains units. The consumption path already reads prelinkedUnitsDir() alongside unitsDir(), so in that state there is nothing to emit and warming is pure waste. first compile, clean checkout: 517s -> 5s committedCacheIsCurrent() is deliberately conservative: missing manifest, unreadable JSON, mismatched fingerprint or an empty units directory all fall through to the warmup, so a stale or partial committed cache is still emitted rather than silently producing a build with missing helpers. Gate (no CI on lib/, and AOT is informational in compiler-gate.yml), branch vs a clean master checkout, run sequentially, both with build/helper-runtime-cache removed first so the measurement is COLD: cold aot-smoke : 8/8 both sides VM --repeat 2 : 110/110 both sides, exit 0 AOT --repeat 3 : master 16 failing, branch 16 failing regressions (branch-only): NONE fixed (master-only) : NONE Note on method: my first attempt at that comparison read master's sweep while it was still running (73 of 110 cases) and showed four phantom regressions — j05, j06, k02, m02 — purely because those cases had not been reached yet. The numbers above are from the completed run. Co-authored-by: PurHur <tedyyyyy@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> * Stdlib: ftp_connect Reflection/named args use Zend hostname (#23644) (#24352) InternalArgInfo still advertises host; override BuiltinParamNames so Reflection and named hostname:/port:/timeout: match ext/ftp/ftp.stub.php. Co-authored-by: PurHur <PurHur@users.noreply.github.com> Co-authored-by: Cursor <cursoragent@cursor.com> * Language: UnhandledMatchError formats variable subjects (#24329) (#24353) Snapshot the match subject into a fresh temp before the message helper so ARG_SEND cannot bind the null match-result seed. Also refuse remapping unhandled-match helpers / named CVs onto that seed. Co-authored-by: PurHur <PurHur@users.noreply.github.com> Co-authored-by: Cursor <cursoragent@cursor.com> * Language: preserve bound closure calledClass across CFG edges (#24335) (#24354) Closure::bind scope was lost on ?? / if / try branch frames, so private static reads threw even when direct single-block access worked. Co-authored-by: PurHur <PurHur@users.noreply.github.com> Co-authored-by: Cursor <cursoragent@cursor.com> * AOT: persist disableRefcount typeinfo store (#24137) compile{} expand of `$ref.typeinfo = …` discarded insertValue (#24226 sibling), so disableRefcount was a no-op and NestedJIT heap-string delref still freed payloads. Hand-store the cleared REFCOUNTED bit; pin json_decode bridge copies. Co-authored-by: Cursor <cursoragent@cursor.com> * Stdlib: AOT json_decode assoc via decodeInto + LLVM tag peek (#24137) NestedJIT cannot use VmJsonFormat or return HT pointers; bridge allocates the HT and peeks the first JSON byte in LLVM (resultTag ARG_RECV emptied string formals). Skip NestedJIT string ARG_RECV rebind that UAF'd payload bytes. Parse into phpc_native_ht_* with ord-based ints and substr slices for nested arrays. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: PurHur <PurHur@users.noreply.github.com> Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: PurHur <tedyyyyy@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Closed
4 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
prelinked/helper-runtime/x86_64-linux/in the pinned Docker env: 293 units published (was 257 stale),core_fingerprintnow63e161964d00f30ff8c9check-helper-runtime-prelink.php --strictgate torelease-readiness.shso fingerprint staleness fails the bundle instead of silently forcing a full corpus emit on clean checkoutcheck-helper-runtime-prelink.phpnow reportscore_fingerprintmismatch early (Release blocker: committed helper-runtime cache is stale, so a clean checkout cannot compile hello world in 2 minutes #24302)Test plan
Refresh command used:
Closes #24302
Made with Cursor