Repository navigation
Reconcile JFace viewer queue and iterative tree search without losing merged M3 array lanes - #36
Merged
Conversation
This was referenced Oct 22, 2026
hsoliwal
marked this pull request as ready for review
October 22, 2026 04:36
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Actual Eclipse/JFace original-source integration
This is a master-based four-file receiver replacing neither the original Eclipse viewer nor the integrated M3JDK21 SDK. Synexia remains the canonical Maven/OpenRewrite recipe owner, Eclipse original sources own the public APIs, UI thread, widget lifecycle, selection, and model-delta behavior.
Why a new master-based reconciliation is necessary
UI #28 is merged to master and contributes
System.nanoTimefixed-first-postdeadlineNanos, stale timer batch isolation, and fast bulk five-arrayM3ViewerDelta.append. The still-open #27, including its merged-on-feature-branch #30, has worker-safe UI dispatch, coalescing, 9 concrete SWT/JFace viewer tests and iterative TreeViewer A5, but still retains an older mutable wall-clock deadline and per-item delta append. Simply merging that source would regress #28.Exact original-owner changes
M3Viewers.java: keep all master queue/batch/nanosecond code andapply; admit only M3: repair shared JFace viewer delta dispatch and iterative tree lookup #27 UI-thread-safepost(...), public additive explicit-owning-Displayoverload andflush; and Fix shared viewer refresh coalescing and stale batch timers #30's adjacent same-object-target single REFRESH coalescing (no crossing UPDATE/ADD/REMOVE barriers). Empty posts are ignored before SWT access.M3ViewersTest.java: union of the M3: repair shared JFace viewer delta dispatch and iterative tree lookup #27/Fix shared viewer refresh coalescing and stale batch timers #30 concrete viewer/worker/callback tests plus both master fix(JFace M3): bounded delta batching and isolated viewer timer generations #28 long-delay and stale-batch tests. 16 JUnit @test methods authored, existing main sources preserved.M3ViewerPaths.java: first-party M3: repair shared JFace viewer delta dispatch and iterative tree lookup #27 opt-in iterative preorderItem[][]+ primitiveint[]stack, same callback/first-match order; no per-node Entry/Node owner.AbstractTreeViewer.java: 97-character opt-in A5 dispatch in the originalinternalFindItem; normal recursive path remains when feature flag off.Intentionally untouched:
M3ViewerDelta.javablobb68f93b3854e49ad0571905ad274823c9c502baeandM3ViewerDeltaTest.javablobffbd64e05cb9c65e2837ff102d8bda83c93d5002, exactly matching current master. Original Eclipse license/header retained on each file; no JNI method or class name changes. The new explicit Display overload is the only additive public helper; established descriptors remain.Recipe-first and source identity
Canonical Synexia #12917 committed first, with four named existing
M3HashPinnedJavaSnapshotRecipeJava21 FILE stages, complete before/postimage resources and SHA-256 manifests, actual OpenRewrite SDK test source, Maven crate, categorized donor comparison, improvement memory, andCANONICAL_RECIPE_HOME.tsventries.This branch starts from
master@70dba12b4b976f6396c7e678e5646f674d10bb4c, no direct master writes/rebase/force push. All four receivers were applied serially from those precommitted canonical postimage templates and independently read back byte-for-byte identical. Original preimage blob identities are in the recipe README/target Git history.Executed evidence, limits, and acceptance
Executed: standalone Java21 strict javac+runtime 460,010 differential/control assertions, covering 20,000 randomized recursive-vs-iterative tree result/trace comparisons, 100,001-depth bounded stack, first-post monotonic fixed deadline, adjacent refresh barriers, stale batch timers and 60,000 queue batches. Source SHA-256
aa5198ef38156bf9adf52a24a6762b71d4cc5bf17a2d273392fce8a98ed8b0cf. This is an independent algorithm model, not a compiled exact SWT/Eclipse test.Prior head evidence: #30's 18 GTK-native SWT/JFace tests and 8 actual OpenRewrite SDK checks passed on its exact historical source; these do not establish pass/fail on this new integrated head.
NOT RUN / HOLD: this exact-head Maven/OpenRewrite and generated Results/fixed-point/refusal, 16 combined actual SWT/JFace tests, full Eclipse Java21/25 Tycho/JFace/PDE build, native GTK/Win32/Cocoa matrix, lazy providers, SWT pixels, Project Explorer #29 and JDT/PDE/Debug/GEF/Nebula consumer deltas, heap/native memory and CPU performance. GitHub check startup failures need separate investigation. Keep draft and do not change the default
viewportmode or claim all-viewers qualification.Catalogue LeetCode/HackerRank/GfG is problem-category only; actual code is reused from reviewed first-party Eclipse PRs with original
Apache-2.0 OR EPL-2.0rights intact. The implementation uses M3 arrays/offsets and existing UI ownership; not M3 + JDK as another layer.