Repository navigation
Native mixed-route QA shows stale channel/well labels and omits raw in combined views #551
Description
Live installed target08 witness from BBBC013 public analysis. Diagnosis owner: Planck; parent owns integration and live acceptance. Existing #451 exposed composite controls but is closed; #445 physical-source discovery and #548 linked-selection reduction are distinct.
Evidence root: /home/ts/wt/openhcs-issue-batch-20260929/next-public039013-after08-20261003/BBBC013_PUBLIC94/author-workspace/output. Original runtime/mcp.stdin and stdout remain unchanged. Parent personally opened the saved native MCP bitmaps.
- qa/finalqa_a01_dna_raw/20261003T203000749737Z_napari_6013_OpenHCS_Napari_Visualization.png displays the selected DNA.bmp image with nuclear morphology, but the footer says ChGFP:gfp. The raw GFP capture at 203020089104 displays GFP.bmp/cytoplasm but says ChDNA:dna. Result/combined footers match their roles.
- qa/finalqa_d06_gfp_raw/20261003T203212935754Z_napari_6013_OpenHCS_Napari_Visualization.png displays the D06 manual image requested by its source-bearing route but the footer says Well A01. Result-only at 203219151450 reports D06, aggregate well41. Combined at 203225047739 visibly contains the result only, with no raw underneath.
- Original requests isolate source-specific manual routes with site/z/time local axes; result routes select aggregate well41. Atomic visibility success is therefore not sufficient evidence of a matched combined view. The exact cause still needs source and original reply analysis. No assertion that analytical input pixels were swapped, no replay, and no biological hint sent to the author.
Trace the existing semantic route presentation, physical source identity, dimension projection and dimension-label overlay owners. Derive the visible identity from the actual mounted route/plane and apply existing compatible route projection for simultaneous raw/result review. Do not relabel pixels, duplicate component inventory, add another coordinate/frame store, or special-case this assay. Check Root394 shared-source ownership before editing.
Acceptance: a small controlled fixture with independent DNA/GFP raw routes and multiwell result routes can isolate and capture raw-only/result-only/combined for A01 and D06. Footer and returned source/component identity must agree with actual displayed pixels, contrast remains route-specific, and combined truly shows the same physical field underneath its result. Incompatible requests must report an accurate structured error rather than successful misleading QA. Exercise the installed public MCP entrypoint and open native captures. Preserve all prior journals/UNKNOWN inputs and continue unrelated scientific runs; no hosted CI hold.
Activity
Merge pull request #554 from OpenHCSDev/fix/semantic-viewer-frame-551
trissim commentedon Oct 5, 2026
New frozen reserve witness for the existing presentation-identity family (no pixel-swap claim).
Root:
/home/ts/wt/openhcs-issue-batch-20260929/next-bbbc01388-p00196-fresh13-20261005/BBBC013_FRESH13_88/author-workspace/output/receipts, exactreserveB01DNAdetail01{win,rawiso,resultiso,combinediso,rawcap,resultcap,combinedcap}.json.I personally opened the three original render-complete native bitmaps: raw
20261005T145749928171Z, result145756671799Z, combined145804587846Z(Napari6021). Raw shows the native layer title DNA B01. Result and combined show a second raw-layer title GFP G06. The combined underlying nuclear field visually matches the raw B01 witness; this review does not claim pixelwise equality or analytical input substitution. Capture replies have identical native axis order/displayed YX/point/canvas dimensions.Determining discrepancy:
winresolves pathB01_s001_wDNA_z001_t001.tif, components DNA/B01/site1/Z1/T1, numeric limits0..100.combinedisoreturns the visible raw route endingchannel_DNA_well_B01and titleselected_images channel DNA well B01, alongside the B01 DNA ROI route. Thus even the returned control title agrees with DNA/B01 while the actual native layer-list title says GFP G06. This is a native-title versus typed route/readback discrepancy; the cause (last payload versus current component, or another naming/update boundary) is not yet established.Planck is the diagnosis owner named in this issue; I requested current owner/implementation disposition directly, not a competing viewer patch. #522 selected-survivor receiving remains on Dewey's sole successor15 lease. No scientific rerun, client/viewer mutation, author feedback or original receipt rewrite.
Focused acceptance extension: after selecting a physical field/channel and isolating raw/result/combined, the visible native layer name and returned title must identify that exact mounted physical component. Moving to a different field/channel and back must update both together. Preserve source/payload/domain and strict source guards; no relabelled pixels or mirrored title/component store.
trissim commentedon Oct 5, 2026
Correction to my preceding comment5998200338: this B01 witness does NOT establish a native layer-name/current-component defect.
Planck independently opened the same three original bitmaps and traced the existing Qt layer-list projection. The GFP G06 second row has a SLASHED/HIDDEN eye in both result and combined captures. Selecting the B01 Shapes layer changes the scrolled list position; only two adjacent rows fit in that viewport. QtLayerList ReverseProxyModel presents native rows in descending order, and reads each original layer.name. The unrelated hidden GFP G06 row therefore cannot be equated with the visible B01 raw route from the isolation receipt.
My original observation of the text was correct but its association with the visible raw layer was not supported. The typed combined receipt selects B01 DNA raw plus its B01 DNA Shapes, and the visible combined field matches the raw witness. There is no proven name-write or pixel-substitution defect here. All original PNGs/receipts and my earlier comment are preserved; no scientific replay, source patch, viewer action or new capture was performed. Original #551/merged #554 footer-family evidence and acceptance remain a separate historical scope.
trissim commentedon Oct 5, 2026
Correction to the new B01 native-title allegation in comment5998200338: I personally reopened all3 originals at full resolution. The GFP G06 second row has a SLASHED/HIDDEN eye in BOTH result and combined; it is not the visible B01 raw route. Exact acknowledgements show raw=B01 DNA, result=B01 DNA Shapes, combined=those two, allerrors[]. The50-entry order has B01raw26/H01Shapes25 and B01Shapes38/G06raw37. Napari QtLayerList uses ReverseProxyModel descending; selection changes the current index/scroll position, so the two shown rows are adjacent list items, not all visible layers. The three screenshots show exactly those neighbouring pairs. QtLayerListModel derives name and eye from the original native layer; no secondary title renderer was found. This witness therefore does NOT establish native name/pixel substitution and warrants no patch or new live run. Original554 footer/layout acceptance remains distinct. Exact source/receipt/PNG SHA record: engineering551/NATIVE-LAYER-ROW-DIAGNOSIS15.rst under the existing issue-batch root. Original captures/journals untouched, no author feedback or viewer mutation.
trissim commentedon Oct 5, 2026
Merged PR #554 qualifies shared slot/footer ownership through actual installed native receiving in both loading orders and six personally inspected raw/result/combined captures. The later B01 layer-name allegation was explicitly retracted in this issue after inspecting the hidden layer row; it establishes no remaining in-scope defect. All five reviewed merge commits are ancestors of current main (GitHub compare reports behind_by=0). The existing PR body now contains Closes #551. Closing this covered focused issue; no new runtime or scientific operation was performed.