From 55d1f848062ef0ae28c1f5fe691e2f2e9a3eb68d Mon Sep 17 00:00:00 2001 From: Claude Date: Fri, 4 Sep 2026 17:01:30 +0000 Subject: [PATCH 1/2] Say ##NO_HANDLER on app 484's deliberately empty handler The TRY around the two component-parameter reads is there because the app runs with or without them - a missing PRODUCT/QUANTITY leaves the fields initial and the app carries on. SLIN reports an empty CATCH on every real system, and the pragma is how ABAP says the handler is meant to be empty; `src/00/00`'s context class already writes it that way in four places. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01CD8XM2xmi9NHb17JHP67nZ --- src/09/z2ui5_cl_smps_app_484.clas.abap | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/09/z2ui5_cl_smps_app_484.clas.abap b/src/09/z2ui5_cl_smps_app_484.clas.abap index 39d7762..69ccb3d 100644 --- a/src/09/z2ui5_cl_smps_app_484.clas.abap +++ b/src/09/z2ui5_cl_smps_app_484.clas.abap @@ -25,7 +25,7 @@ CLASS z2ui5_cl_smps_app_484 IMPLEMENTATION. TRY. product = t_params[ n = `PRODUCT` ]-v. quantity = t_params[ n = `QUANTITY` ]-v. - CATCH cx_root. + CATCH cx_root ##NO_HANDLER. ENDTRY. IF client->check_on_navigated( ). From 19e8cf89dd7c1d4e38a347e530667884dfe0287d Mon Sep 17 00:00:00 2001 From: Claude Date: Fri, 4 Sep 2026 21:41:26 +0000 Subject: [PATCH 2/2] Re-measure the thumbnail gaps: same thirteen classes, bigger corpus AGENTS.md carried "18 of 31 app views render" from the 2026-08 measurement, and the corpus has grown since. Re-run with the same harness the deploy uses (`screenshotFiles( )` over `src` at the pinned runtime): 19 of the 32 app classes photograph, 24 of the 37 documents they build. The interesting half is that the FAILURE set did not move at all - the same thirteen classes, in the same three categories the first measurement named, and every one of them still fails for the reason it named: seven sap.ui.comp Smart Controls (SmartFilterBar 313/477, SmartForm 475/476, SmartVariantManagement 478/493, SmartChart 479) that the harness's OpenUI5 runtime does not carry, three z2ui5.cc controls that do not load headless (Websocket 489, MIME-audio 487, Smart Multi Input 319), and the three RAP samples whose ObjectStatus gets an empty `state` from the mock model (005/006/010). The categories are written out per class now, so the next re-measurement can tell "the same gaps" from "different gaps that happen to count the same". Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01CD8XM2xmi9NHb17JHP67nZ --- AGENTS.md | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff --git a/AGENTS.md b/AGENTS.md index 7966067..0461045 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -387,12 +387,17 @@ pass over the classes on `main` — `stats.types` and the `*-too-new` findings. harness. The script went with the page and the playground's deploy takes them now, from the same harness against the same `main`, so what a card shows is still what the render gate checks. Expect gaps, and expect the same ones: - measured over this corpus (2026-08) **18 of 31 app views render**, and the 13 - skips are three stable categories — `sap.ui.comp` controls (SAPUI5-only, - absent from the harness's OpenUI5 runtime: 7 of the 9 Smart Controls - samples), `z2ui5.cc` custom controls that do not load headless (the + re-measured over this corpus (2026-09, `screenshotFiles( )` over `src` at the + pinned runtime) **19 of the 32 app classes photograph, 24 of the 37 documents + they build**, and the 13 that do not are the same three stable categories the + first measurement (2026-08, 18 of 31) found — `sap.ui.comp` controls + (SAPUI5-only, absent from the harness's OpenUI5 runtime: 7 of the 9 Smart + Controls samples — SmartFilterBar, SmartForm, SmartVariantManagement, + SmartChart), `z2ui5.cc` custom controls that do not load headless (the WebSocket, MIME-audio and Smart Multi Input samples), and three RAP samples - whose `ObjectStatus` gets an empty `state` from the mock model. A card + whose `ObjectStatus` gets an empty `state` from the mock model. The count + moved because the corpus grew, not the failure set: it is the same thirteen + classes. A card without a picture is normal here, not broken, and a change to the harness rather than to those classes is what would fix it.