V2 - Increase frametime accuracy by davepl · Pull Request #126 · PlummersSoftwareLLC/NDSCPP

davepl · 2026-06-18T01:02:07Z

Was an integer number of millis before, now a proper timespan.
Adds the StockBanner scrolling ticker effect. Consumes the V2 stock API.

…tive drift - Refactored EffectsManager to use an absolute nanosecond timeline for rendering. - Updated interfaces.h and LEDEffectBase to use microsecond precision for delta time. - Updated all LED effects to utilize the high-precision delta time. - Standardized timing across SpeedTracker, monitor UI, and connection delays to microseconds. - Improved global millis() accuracy and added delayMicros utility.

- Fixed Schedule logic to correctly handle time ranges crossing midnight. - Updated Canvas and Controller deserialization to handle missing optional fields. - Switched Canvas feature mutex to recursive_mutex to prevent deadlocks.

- Refactored HSV2RGB to use hsv2rgb_rainbow for better speed and color accuracy. - Optimized blendWith using lerp8. - Added memcpy fast-path to PixelsToByteArray to reduce CPU overhead. - Defaulted logging level to debug.

davepl · 2026-06-18T01:03:22Z

+            : c;
+    }
+
+    static constexpr array<uint8_t, 7> GlyphRows(char c)


Took a while to get this constexpr - but it was worth it :-)

Copilot

Pull request overview

This PR improves frame timing accuracy by switching feature frame generation to use an explicit system_clock::time_point, expands socket diagnostics, and adds a new scrolling Stock Banner effect that consumes the V2 stock API.

Changes:

Update LED feature frame headers to use an explicit display timestamp and improve frametime scheduling in EffectsManager.
Add new Stock Banner scrolling ticker effect and expose it via effect serialization/catalog.
Extend socket channel diagnostics (failed connect count + last socket error) and improve server shutdown handling.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
webserver.h	Makes server wait/shutdown more robust by guarding invalid futures and logging exceptions.
main.cpp	Clarifies startup logging by separating dashboard vs API endpoints.
interfaces.h	Updates feature frame API to accept a display timestamp; adds new socket diagnostic getters.
ledfeature.h	Implements timestamped frame generation and fixes reversed multi-row pixel mapping.
effectsmanager.h	Improves frame pacing precision and passes computed display timestamps into features.
socketchannel.h	Adds connection failure diagnostics and adjusts frame send batching logic and response aging check.
apihelpers.h	Ensures sockets are restarted/stopped alongside canvas effect start/stop flows.
apihelpers.cpp	Registers StockBanner in the effect catalog for the web UI.
effects/stockbannereffect.h	Introduces the StockBanner effect with HTTP quote fetching and rendering.
secrets.h.example	Adds configuration placeholder for the realtime quote key used by StockBanner.
tests/tests.cpp	Adds a unit test validating reversed multi-row feature pixel mirroring; provides test logger/static defs.
tests/Makefile	Adds link libs needed by new/updated dependencies used in tests.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

+                        size_t tempCount = 0;
+                        size_t tempBytes = 0;
+                        auto queueCopy = _frameQueue;
+                        while (!queueCopy.empty() && tempCount < kMaxBatchSize)
+                        {
+                            tempBytes += queueCopy.front().size();
+                            tempCount++;
+                            queueCopy.pop();
+                        }
+
                        combinedBuffer.reserve(tempBytes);


+        addrinfo hints{};
+        hints.ai_family = AF_UNSPEC;
+        hints.ai_socktype = SOCK_STREAM;
+


+
+        try
+        {
+            const string path = "/?ticker=" + symbol + "&v=2&profile=hub75&points=1";


+        const string symbol = _quotes[*index].symbol;
+        _lastRequestStart = now;
+        _pendingFetch = async(launch::async, [this, symbol]()
+        {
+            return FetchQuote(symbol);
+        });


+#include <utility>
+#include <vector>
+
+class StockBanner : public LEDEffectBase


Resolved conflicts keeping robertlipe's sub-millisecond timing changes (GetDataFrame with targetTime, microseconds deltaTime for effects) while taking upstream bug fixes (overflow-safe buffer size cast, resilient invalid-byte handling in socketchannel, SetId preservation in from_json, clientBufferCount default 500). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

timing precision & adds aurora borealis effect

performance and color

Follow-up fixes to #102

schedule and robustness

rbergen

Code review findings from reading the full diff.

rbergen · 2026-06-21T17:39:34Z

+    uint32_t _stripHeight = 0;
+    bool _stripDirty = true;
+
+    future<FetchResult> _pendingFetch;


Blocking destructor on std::async future. When StockBanner is destroyed while a fetch is in progress (e.g. because the effect is stopped), the destructor of a future obtained from std::async(launch::async, ...) blocks until the async task completes — up to _requestTimeout (1500 ms). Since Stop() can be called from the effects manager thread, this stalls the whole render loop. Consider using a detached thread with a shared cancellation flag, or storing the future in an optional and resetting it under a condition that allows early exit.

rbergen · 2026-06-21T17:39:35Z

+
+        try
+        {
+            const string path = "/?ticker=" + symbol + "&v=2&profile=hub75&points=1";


No URL-encoding on user-supplied symbols. Ticker symbols from the JSON config are concatenated directly into the query string. A value like FOO&bar=1 would silently corrupt the request. Stock tickers are normally alphanumeric, but since these come from user config it's worth at least validating they match [A-Z0-9.\-]+ before building the URL.

rbergen · 2026-06-21T17:39:35Z

+            const int bottomScale = LargestTextScale(priceText + (changeText.empty() ? "" : " " + changeText),
+                                                     static_cast<int>(_compactQuoteWidth) - 8,
+                                                     bottomHeight - 2);
+            const int width = static_cast<int>(max<uint32_t>({


Tile width calculated twice. RebuildStrip computes each tile's width here to sum up the total strip width, and then RenderQuoteTile (line 576) recomputes it independently. If those two calculations ever diverge — e.g. due to a future edit — tiles will misalign. Consider extracting a ComputeTileWidth(const Quote&) helper that both callers use.

rbergen · 2026-06-21T17:39:35Z

+    {
+        (void) deltaTime;
+
+        EnsureQuoteList();


EnsureQuoteList() runs O(n²) work every frame. At the end of EnsureQuoteList, _symbols is always repopulated from _quotes, so it is never empty after the first call. That means the reconciliation loop (with its inner find_if over _quotes) runs on every Update() call. A _symbolsDirty flag that from_json sets and EnsureQuoteList clears would make this a no-op on the hot path.

rbergen · 2026-06-21T17:39:35Z

-            auto nextFrameTime = steady_clock::now();
+            const auto effectiveFps = max<uint16_t>(1, _fps);
+            const auto frameDuration = duration_cast<steady_clock::duration>(duration<double>(1.0 / effectiveFps));
+            const auto effectDelta = duration_cast<milliseconds>(frameDuration + 500us);


Magic 500 µs offset is undocumented. effectDelta is passed as deltaTime to every effect's Update(). It is now a fixed value rather than the actual elapsed time, and the 500 µs padding has no explanation. Effects that use deltaTime for physics or animation will receive a slightly inflated constant every frame. Please document why 500 µs was chosen (scheduler wake-up margin? rounding buffer?).

rbergen · 2026-06-21T17:39:35Z

-                    if (tempBytes > 0) 
+                        size_t tempCount = 0;
+                        size_t tempBytes = 0;
+                        auto queueCopy = _frameQueue;


queueCopy allocation is vestigial with kMaxBatchSize = 1. The copy of the whole queue exists only to compute tempBytes for combinedBuffer.reserve(). With batch size 1, this copies at most one item to determine one frame's size — a combinedBuffer.reserve(_frameQueue.front().size()) would do the same without the copy.

Bumps [actions/checkout](https://github.com/actions/checkout) from 6 to 7. - [Release notes](https://github.com/actions/checkout/releases) - [Changelog](https://github.com/actions/checkout/blob/main/CHANGELOG.md) - [Commits](actions/checkout@v6...v7) --- updated-dependencies: - dependency-name: actions/checkout dependency-version: '7' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <support@github.com>

…tions/actions/checkout-7 Bump actions/checkout from 6 to 7

rbergen

This PR suffers from a more-involved-than-usual merge conflict, almost certainly caused by a number of PRs from Robert's hand that I recently merged.

I can take a shot at resolving them if you'd like me to - I went through the merged PRs fairly recently so I still have some memories about which changes were made why.

davepl · 2026-06-28T16:15:36Z

If you could, I’d appreciate it! I’m working on the 899 PR in the NDS folder… - Dave

…

On Jun 28, 2026, at 9:14 AM, Rutger van Bergen ***@***.***> wrote: @rbergen commented on this pull request. This PR suffers from a more-involved-than-usual merge conflict, almost certainly caused by a number of PRs from Robert's hand that I recently merged. I can take a shot at resolving them if you'd like me to - I went through the merged PRs fairly recently so I still have some memories about which changes were made why. — Reply to this email directly, view it on GitHub <#126?email_source=notifications&email_token=AA4HCFZQUXS2T5FUBT4HGPT5CE76BA5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTINJYG43TQNJYGYZ2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#pullrequestreview-4587785863>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA4HCFZMDZIKQ643PPEBW3T5CE76BAVCNFSNUABFKJSXA33TNF2G64TZHM4DQOJRGQYDAMRYHNEXG43VMU5TINRYHAZDINZZHEZ2C5QC>. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/AA4HCF33SPQKRPHVPNBXYUD5CE76BA5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTINJYG43TQNJYGYZ2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG> and Android <https://github.com/notifications/mobile/android/AA4HCF6P5GWU5N2FSCFI7FL5CE76BA5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTINJYG43TQNJYGYZ2M4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>. Download it today! You are receiving this because you authored the thread.

rbergen · 2026-07-04T08:33:08Z

@davepl I've been looking into this merge conflict, and I think I have a plan to bring main and v2 together in a way that combines the best of both. One question came up during this that I would like your input on.

In the v2 branch, EffectsManager::Start() now computes one displayTime per frame by taking the maximum TimeOffset() across all of a canvas's features, and applies that single shared value to every feature's frame.
Previously (and still on main), each ILEDFeature::GetDataFrame() applied its own TimeOffset() individually, so two features with different timeOffset values on the same canvas would each get their own offset, rather than both inheriting the largest one present.

Was that change to a shared/max offset intentional (e.g. to keep all features on a canvas frame-synchronized), or an incidental side effect of restructuring the timing loop for accuracy? If it's intentional, is it meant to replace per-feature offsetting everywhere, or only in specific cases?

As you will understand, your answer to this question determines whether I will go one way with the conflict resolution, or the other.

Documents the file-by-file conflict analysis between v2 and main, and tracks progress resolving them incrementally on this branch. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Resolves the 4 files that conflict between main and v2 by keeping main's side as-is for now: effectsmanager.h, interfaces.h, ledfeature.h, socketchannel.h. This establishes a compilable baseline that includes v2's cleanly-auto-merged additions (StockBanner effect, secrets key, tests) but NOT yet v2's intended changes to the 4 placeholder files. Each placeholder will be replaced with its real, hand-analyzed resolution in a follow-up commit — see pr126-merge-plan.md for the analysis and progress tracking. effectsmanager.h in particular is deferred to a separate pass since it requires blending both branches' logic rather than picking a side. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Three issues, none part of the actual conflict set, that only showed up once v2's cleanly-auto-merged additions were compiled against main's (placeholder-resolved) interfaces: - effects/stockbannereffect.h: StockBanner::Update() declared its unused deltaTime parameter as milliseconds, but ILEDEffect::Update() (from main) takes microseconds - hid the virtual instead of overriding it. Parameter value is unused, so retyping it is behavior-neutral. - tests/tests.cpp: DummyEffectsManager::UpdateCurrentEffect() had the same milliseconds/microseconds mismatch against IEffectsManager, matching v2's own (now-superseded) interface signature instead of main's. - socketchannel.h: pre-existing -Wreorder issue on main itself (ctor initializer list order didn't match member declaration order for _totalQueuedBytes/_lastInvalidByteWarning) - invisible on the main Makefile (no -Werror) but breaks the -Werror tests build. Unrelated to the v2 merge; fixed here only because it blocked verifying every step per the plan in pr126-merge-plan.md. With these, `make` and `make -C tests` both succeed and 13/14 tests pass. The one expected failure, APITest.ReversedMatrixFeatureMirrorsRowsForHardware, exercises the multi-row-reversed pixel bug in ledfeature.h that this branch's ledfeature.h placeholder (main's original, buggy-for-this-case version) hasn't been replaced with its real resolution yet - tracked as Step B in pr126-merge-plan.md. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

These two files turned out to be compilation-coupled and are resolved together: interfaces.h's ISocketChannel gains v2's GetFailedConnectCount()/ GetLastSocketError() pure virtuals, which socketchannel.h must implement or the class becomes abstract. interfaces.h (Step A): the one real conflict (GetDataFrame's parameter name, targetTime vs displayTime) resolves to main's targetTime, since main's per-feature TimeOffset() semantics are being kept (see ledfeature.h, Step B, and pr126-merge-plan.md). The non-conflicting GetFailedConnectCount/ GetLastSocketError declarations from v2 are carried over as-is. socketchannel.h (Step C): WorkerLoop() keeps main's structure wholesale (persistent combinedBuffer/reserve, throttled response polling via ReadSocketResponse()) rather than v2's per-iteration buffer allocation and inline per-send response reads. v2's RecordConnectFailure()/ RecordConnectSuccess() diagnostics helpers and their call sites are kept, as is v2's _lastSocketError/_failedConnectCount tracking. The one real conflict this creates - the ConnectSocket() success-logging tail - is resolved by calling RecordConnectSuccess() (mutex-guarded bookkeeping) while keeping main's tiered log level (info once, debug on reconnects) rather than v2's uniform info-level log on every reconnect. Note: the initial placeholder-merge commit (5de9659) resolved conflicts via `git checkout --ours`, which discards a file's entire merge result, not just the conflicting hunks - so it silently dropped v2's non-conflicting RecordConnectFailure/Success feature and the matching interfaces.h declarations entirely. This commit restores them by reconstructing the true 3-way merge (git merge-file against the branches' actual tips) rather than hand-editing from the placeholder state. `make` and `make -C tests` pass; 13/14 tests pass (the remaining failure, ReversedMatrixFeatureMirrorsRowsForHardware, is expected until ledfeature.h is resolved in the next commit). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Two real conflicts, both resolved by combining rather than picking a side: - GetPixelData()'s fast-path condition keeps main's bounds check but adds v2's extra `&& (!_reversed || _height == 1)` guard. Without it, a full-canvas multi-row reversed feature would hit Utilities::ConvertPixelsToByteArray(), which reverses the whole flattened pixel buffer - correct for a single row, but wrong across rows (it reverses row order too, not just each row's pixel order). v2 added this guard specifically to route multi-row reversed features to the slow path, which handles per-row reversal correctly. - GetDataFrame() keeps main's targetTime parameter and per-feature TimeOffset() application, rejecting v2's change to a shared canvas-wide max offset (see the "Question to send to the PR author" in pr126-merge-plan.md - this preserves the existing per-feature contract pending the author's confirmation). As with the previous commit, this reconstructs the true 3-way merge rather than hand-editing from the placeholder state, which also restores v2's non-conflicting per-row reversal fix in the slow path and the failedConnectCount/lastSocketError fields in to_json (both silently dropped by the initial `git checkout --ours` placeholder). `make` and `make -C tests` pass; all 14 tests pass, including ReversedMatrixFeatureMirrorsRowsForHardware which exercises the fast-path guard fix above. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Records what actually happened while executing Steps 1-3, A, B, C, notably the git checkout --ours pitfall (it discards a file's whole merge result, not just conflicting hunks) and the A/C compilation coupling, so a future session picking up Step D (effectsmanager.h) doesn't repeat either. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Author confirmed (2026-07-05) the shared/max time-offset change in v2 was not intentional - likely an incidental side effect of vibe-coded visual work on the StockBanner effect. Confirms the Step B resolution already applied in ledfeature.h (keep main's per-feature TimeOffset()) was correct. Author's guidance: take whichever path resolves the merge fastest. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

The last of the 4 conflicting files, and the only one needing genuine hand-blending rather than picking a side: main's Start() (schedule-based effect switching, heartbeat/blackout frames, drift detection, >1s sane reset) and v2's Start() (this PR's actual purpose - frametime accuracy) independently rewrote the same worker-thread loop since v2 branched off before the scheduling work landed on main. Kept main's outer control flow as the skeleton (schedule check -> active/ inactive branch -> heartbeat/blackout -> drift detection -> sane reset), and spliced in v2's timing-precision improvements: - frameDuration now computed via a double-precision intermediate (duration_cast<steady_clock::duration>(duration<double>(1.0/fps))) instead of truncating nanoseconds(1'000'000'000LL / fps) up front - the actual fix this PR is titled for. - FPS clamped to a minimum of 1 (max<uint16_t>(1, _fps)). - Catch-up avoidance: if the loop falls more than one frame behind, jump the frame counter forward to match real elapsed time instead of sending a rapid-fire burst of queued frames. This sits alongside (not instead of) main's existing >1s sane-reset, which remains as a backstop for more severe stalls (e.g. system suspend) since it also re-centers the epoch and avoids unbounded frame-counter growth. - Frames are now compressed via Socket()->CompressFrame() before being enqueued, in both the active-effect and heartbeat send paths - a v2 addition that doesn't conflict with anything on main. - to_from_json_map gets v2's StockBanner registration alongside main's AuroraEffect one (both additive, both kept). Explicitly NOT carried over: v2's change to compute a single canvas-wide maximum TimeOffset() and bake it into one shared displayTime for every feature. main's (and this branch's, per ledfeature.h/Step B) per-feature TimeOffset() application inside GetDataFrame() is kept instead. The PR author confirmed this wasn't an intentional design change (see pr126-merge-plan.md). As with the earlier files, this reconstructs the true 3-way merge (git merge-file against the branches' actual tips) rather than hand-editing from the placeholder state, which also restores v2's non-conflicting #include "effects/stockbannereffect.h" and #include <algorithm>. This is the last of the 4 conflicting files. `make` and `make -C tests` pass; all 14 tests pass. The merge is otherwise complete pending a final review pass and manual smoke test per pr126-merge-plan.md. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

All 4 conflicting files (interfaces.h, ledfeature.h, socketchannel.h, effectsmanager.h) are resolved and committed on this branch. Full build and test suite pass. Remaining work before landing: manual smoke test and a final diff read-through, per the execution plan. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Every effect subclassing LEDEffectBase now passes its own class name as the type string to the base constructor (static TypeName member), instead of only AuroraEffect doing so while everything else fell back to the default "LEDEffectBase" string. jsonPair<T>() keys to_from_json_map by T::TypeName instead of typeid(T).name(), so the AuroraEffect-only special case (make_pair("AuroraEffect", jsonPair<AuroraEffect>().second)) is no longer needed - every entry is now a plain jsonPair<T>(). This makes to_json's mangled-typeid fallback dead code (effect.Type() now always matches on the first lookup), so it's removed. from_json instead gains a small fallback: if the JSON "type" string isn't found directly, strip any leading digits (the Itanium ABI name-length prefix baked into old typeid-mangled saves, e.g. "15StarfieldEffect") and retry. This keeps config files saved before this change - and the previously-shipped config.led.example, which had the same old mangled names - deserializing correctly instead of silently replacing those effects with magenta fill. config.led.example itself is also updated to the new clean type strings, so a fresh checkout no longer needs the fallback at all. `make` and `make -C tests` pass; all 14 tests pass. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Working doc used to track the merge conflict resolution as it progressed; no longer needed now that all conflicts are resolved and verified. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

rbergen

LGTM, with the explicit note that you (@davepl) should take another look to apply the 4-eye principle to my conflict resolution changes.

robertlipe and others added 18 commits April 13, 2026 21:49

Don't recompile if only config.led changes.

1bb2390

Fix overnight schedules and improve JSON robustness

61ac41f

- Fixed Schedule logic to correctly handle time ranges crossing midnight. - Updated Canvas and Controller deserialization to handle missing optional fields. - Switched Canvas feature mutex to recursive_mutex to prevent deadlocks.

Optimize color math and enhance performance

01ae287

- Refactored HSV2RGB to use hsv2rgb_rainbow for better speed and color accuracy. - Optimized blendWith using lerp8. - Added memcpy fast-path to PixelsToByteArray to reduce CPU overhead. - Defaulted logging level to debug.

Merge remote-tracking branch 'upstream/main' into pr/robertlipe/101

e36aa96

Merge remote-tracking branch 'upstream/main' into pr/robertlipe/102

2694896

Merge remote-tracking branch 'upstream/main' into pr/robertlipe/103

b0d1289

fix(api): restore backward compatibility for color JSON keys

d0fbb34

fix(api): restore backward compatibility for color JSON keys

5080a19

fix(api): restore backward compatibility for color JSON keys

e96f8bc

Give ndscpp more time to settle down

a28da38

make config.led for CI test run

c14332d

Merge branch 'PlummersSoftwareLLC:main' into performance-and-color

57b4f57

Merge branch 'PlummersSoftwareLLC:main' into schedule-and-robustness

78ae894

make config.led for CI test run

e459162

make config.led for CI test run

5e79b3b

Add StockBanner effect, use V2 quote interface

afb174d

Increase frametime accuracy

ba3f726

davepl commented Jun 18, 2026

View reviewed changes

davepl requested review from Copilot and rbergen June 18, 2026 01:06

Copilot started reviewing on behalf of davepl June 18, 2026 01:06 View session

Copilot AI reviewed Jun 18, 2026

View reviewed changes

rbergen and others added 7 commits June 21, 2026 11:13

Merge pull request #101 from robertlipe/timing-precision

9f80181

timing precision & adds aurora borealis effect

Clean up dead code

2955f80

Merge remote-tracking branch 'upstream/main' into pr/robertlipe/102

a2ad900

Merge pull request #102 from robertlipe/performance-and-color

ad3fd31

performance and color

Follow-up fixes to #102

2bca341

Fix merge conflicts with upstream/main

82d03d7

rbergen added 4 commits June 21, 2026 12:43

Catch invalid hex color pattern

02d00e0

Merge pull request #127 from rbergen/102-follow-up

f1b25b0

Follow-up fixes to #102

Merge remote-tracking branch 'upstream/main' into pr/robertlipe/103

9287572

Merge pull request #103 from robertlipe/schedule-and-robustness

2d42e3c

schedule and robustness

rbergen reviewed Jun 21, 2026

View reviewed changes

dependabot Bot and others added 2 commits June 23, 2026 12:53

Merge pull request #128 from PlummersSoftwareLLC/dependabot/github_ac…

77b2bf7

…tions/actions/checkout-7 Bump actions/checkout from 6 to 7

rbergen reviewed Jun 28, 2026

View reviewed changes

rbergen and others added 11 commits July 4, 2026 10:41

Add PR #126 merge conflict analysis and resolution plan

6d2e498

Documents the file-by-file conflict analysis between v2 and main, and tracks progress resolving them incrementally on this branch. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Remove PR #126 merge planning doc

617d886

Working doc used to track the merge conflict resolution as it progressed; no longer needed now that all conflicts are resolved and verified. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

rbergen approved these changes Jul 5, 2026

View reviewed changes

davepl merged commit 8dea122 into main Jul 14, 2026
8 checks passed

Uh oh!

Conversation

davepl commented Jun 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

davepl Jun 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

rbergen left a comment

Choose a reason for hiding this comment

Uh oh!

rbergen Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

rbergen Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

rbergen Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

rbergen Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

rbergen Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

rbergen Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

rbergen left a comment

Choose a reason for hiding this comment

Uh oh!

davepl commented Jun 28, 2026 via email

Uh oh!

rbergen commented Jul 4, 2026

Uh oh!

rbergen left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

davepl commented Jun 18, 2026 •

edited

Loading