[pull] master from GaijinEntertainment:master#1024
Merged
Conversation
Three deep-link bugs surfaced by the daslang.io § 01 "try <test> on the playground →" jumps: 1. `?example=<slug>` was silently overridden by localStorage autosave — visitors who'd used the playground before landed on their stale buffer instead of the workload they clicked through for. The playground-tabs.js tryInit autosave-restore now treats `?example=` the same way it already treats `#code=` / `#z=` hashes: a URL-borne intent signal that skips autosave so main.js's pageInit can apply the URL param. 2. As a consequence of #1, the JIT radio also stayed disabled — when the deep-link silently fell through to Hello world (no .wasm) instead of selecting SHA-256 (has .wasm), updateEngineAvailability HEAD-probed the wrong artifact. Fixed by #1. 3. When JIT was opted into for a sample with .wasm and the user switched to a sample without one, `updateEngineAvailability` greyed out the JIT radio but left it `checked = true`. `selectedEngine()` kept returning 'jit' and runCode() 404'd on the missing artifact. Now disabling JIT also unchecks it and re-checks interpreter. Coverage: - persistence.spec.js gains an `autosave does not override ?example=` test, mirroring the existing `#code=` test. - New engine-toggle.spec.js covers both "JIT enables when .wasm exists" and the auto-revert. HEAD probe is stubbed via page.route so the test doesn't depend on `cmake --build … all_wasm` running locally. Verified end-to-end against a local serve with my own Chrome driver: `?example=sha256` after planting stale autosave lands on the SHA-256 source; opt into JIT on SHA-256 → switch to Macros → JIT radio is now disabled+unchecked, interpreter checked. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Earlier if JIT failed it could be missed, because return code was still 0. Now we will panic in such cases.
Return code should be non zero if something leaked. Return 1 in alloc_tracked, by calling exit().
Three out of four threads engaged; the helper-extraction nit is left
in place (out of scope for this PR — when a third spec needs the same
waitDropdownsPopulated body, hoist it then).
1. updateEngineAvailability: gate the HEAD response on `name ===
currentJitName` so a late-arriving response for a stale sample
doesn't clobber the radio. Repro under HTTP/2 (or any async
ordering): pick sha256 (HEAD slow), pick tree (HEAD slow); without
the gate, whichever .then resolves last wins.
2. playground-tabs.js: replace `URLSearchParams.has('example')` with a
truthy check on `.get('example')`. The pre-fix `has` form was true
for `?example=` (empty) and `?example` (no value) — autosave would
be skipped, then main.js's `params.get` would null-check, fall
through to the default sample, and the user's autosave was clobbered
for nothing. Now we only skip autosave when there's a real slug.
3. engine-toggle.spec.js: narrow `page.route` to fulfill 200 for
sha256.wasm only, 404 for every other `**/samples/examples/*.wasm`.
The previous blanket-200 stub also satisfied the startup probe for
the default Hello world sample, so the JIT-enables test passed even
when the sample-switch path was broken. Test body now explicitly
selects Functions first (asserts disabled) → SHA-256 (asserts
enabled), so the transition is what's verified, not a coincidentally-
correct end state.
Verified under both directions of the race (sha256→tree and
tree→sha256, each with an 800ms HEAD-delay monkey-patch). 32/32
no-wasm playwright tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Usage of self-> inside class method is not advised. Simply write method name, compiler will deduce self.
JIT fails to link without sanitizer flag passed to linker, when sanitizer enabled.
`autosave does not override ?example=` test could race: waitTabsReady returns as soon as pgState exists (initial empty doc), but main.js's `?example=` branch still has to async-fetch data.json + sha256.das before pgLoadFiles swaps the buffer content. The bare evaluate on window.pgState.files['main.das'].getValue() could catch the pre-load empty state and fail the regex assertion. Localhost was always fast enough to race in our favor; CI under load is the realistic flake risk. Swap the bare evaluate for an expect.poll that retries until the sha256 marker appears (5s timeout). The "not stale autosave" assertion runs after the poll resolves on the final content, so it's unaffected. Verified with --repeat-each 5: my test passed 5/5 (and the 5/5 was already true pre-fix on localhost; the change is for CI hardening). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
waitForTimeout(400) over the 250ms autosave debounce leaves only a 150ms margin. Localhost always made it; CI under load (worker concurrency + system stress) consistently lost the race — saw 2/5 fail with --repeat-each 5, intermittent fail in the full-suite runs. Swap for an expect.poll on the localStorage payload, polling until both the file-set AND the active-tab pointer have been persisted by the autosave writer. Deterministic — we proceed exactly when the state has actually landed, not on a fixed wall-clock fudge. Verified with --repeat-each 10: 30/30 pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
… sweep) `qmacro` / `qmacro_block` / `qmacro_expr` / `qmacro_block_to_array` go through `apply_template` (templates_boost.das:251), which calls `clone_expression` on every `$e(...)` substitution input. So `var X = clone_expression(E); ... $e(X) ...` clones twice. Drop the pre-clone and splice the source directly — `apply_template` clones at substitution time. PERF023 detection (post-expansion form the lint sees): `$e(X)` becomes `add_ptr_ref(X)` inside an `ExprMakeBlock`. The visitor tracks a splice-wrapper depth counter via preVisitExprCall/visitExprCall on `add_ptr_ref`, and classifies each candidate `ExprVar` reference as "safe" when depth > 0, "disqualifying" otherwise. Fires only when ALL uses are safe AND ≥1 is observed. Multi-clone-of-same-source is flagged too — apply_template clones each substitution independently, so inlining `$e(E)` at N splice sites preserves N-independent-clones semantics. `clone_type` is out of scope (no matching `add_type_ptr_ref` wrapper in user-facing qmacro grammar). Sweep removes 76 redundant pre-clones across the codebase: - daslib/linq_fold.das — 68 sites (-68 LOC net) - modules/dasSQLITE/daslib/sqlite_linq.das — 7 sites - daslib/ast_match.das — 1 site Validation: tests/linq (1360), tests/decs (245), tests/dasSQLITE (782), tests/ast_match + tests/template + utils/lint/tests all green. utils/lint/tests/perf023_clone_into_qmacro.das covers 4 positive shapes + 4 negative shapes for the rule. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
lint: add self lint check
…deeplink-jit-fixes playground: ?example= overrides autosave + JIT radio auto-reverts
…undant-clone-before-qmacro perf_lint: add PERF023 — drop pre-clone before qmacro splice (76-site sweep)
Linker in CI for JIT should match with linker for libDaScript.
UBSAN complains if memset called on nullptr, even if size is 0 as well.
This reverts commit 82e1404.
This reverts commit 870416e.
…n-failure-on-error JIT: return failure on error
…cache Aleksisch/revert sccache
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 participants
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.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )