Ten upstream OCCT PRs now sit on a pre-8.0.1 base, and patch 0020 was never filed at all

[gsdali]

Fallout from the OCCT 8.0.1 absorb (#654). Housekeeping, deliberately not done as part of that
change so each PR update stays reviewable on its own.

The ten open PRs need rebasing

V8_0_1 absorbed ten of our carried patches, which means every OCCT PR still open was authored
against a base that has now moved. Verified open as of 2026-08-03:

Our patch	Upstream PR	Fixes
0010	OCCT#1386	Intf_Interference O(1) tangent zone + breaker (#319)
0011	OCCT#1388	XCAFDoc_ShapeTool::OwnAutoNamingScope (#341/#363)
0012	OCCT#1390	GetApplication / CDF_Directory races (#344)
0014	OCCT#1394	PCDM driver reentrancy (#349)
0015	OCCT#1397	CDM_Application metadata table (#353)
0016	OCCT#1399	Resource_Manager / Storage_Schema (#374)
0017	OCCT#1410	Null ReShape context (#484)
0018	OCCT#1417	GCPnts count contract (#555)
0019	OCCT#1418	AdvApp2Var Jacobi workspace slot (#522)
0021	OCCT#1420	CPnts adaptive arc-length integration (#603)

All eleven surviving patches apply cleanly to V8_0_1 with zero file-level collisions against
the 74 files 8.0.1 touched, so a rebase should be mechanical. That is not the same as the PRs being
mergeable: GitHub compares against master, which has moved further than the tag.

Two are worth a closer look than a plain rebase:

• 0016 / OCCT#1399 still carries the mutex version of the Storage_Schema half. Revisit #374's Storage_Schema::ICurrentData fix: per-instance field instead of a mutex (upstream review, OCCT#1399) #518 records
  that maintainer review asked for the per-instance field instead, and our patch was redesigned;
  the PR was never updated to match.
• 0017 / OCCT#1410 is in ShapeFix, the area 8.0.1 changed most heavily (fix(kernel): make TopoDS_TShape::myState atomic (Issue #1154) #1323, chore(#819): extend derive-swift-file-split.py to any file, reassess #687's four #1377, Split the remaining .mm bridge implementation files by OCCT subsystem (#396 follow-on) #1380,
  test: share the bsplineSurface(interiorMultiplicityU:) continuity fixture (#1253) #1331, Spike: gate the Scripts/patches/ count and the gate-scripts step count against their own stated prose (recurred at v2.0.0-kernel.1-3, #1032, #1157/#1402, and still-open #1066) #1408 all landed there). Re-verify the defect still reproduces on master before pushing.

Patch 0020 has no upstream filing

0020-BRepFeat_MakeCylindricalHole-select-tool-parts-532 fixes a silent wrong answer:
BRepFeat_NoError with no material removed whenever the drill crosses two bodies or severs one
(#532). Scripts/patches/README.md and okf/references/carried-occt-patches.md both record it as
"not yet filed". It is the only carried patch with no upstream PR, so it is the only one that can
never retire.

The reproducer and before/after measurements already exist at
Scripts/repro/532-cylindrical-hole-part-selection/. Per upstream's guidance on
OCCT#1409 it should
go straight to a PR with no companion repro issue.

Note the writeup also records a second defect in the same heuristic that we did not fix:
PerformThruNext's closest-interval fallback nests its // parbar > Last branch inside
if (parbar < First), making that case unreachable as written. Worth deciding whether it rides
along or stays reported-only.

[gsdali]

Policy: prove the test fails

Whatever test or --self-test case this issue adds, run it once with its subject broken. Inject
the defect it exists to catch, confirm it fails, restore, confirm it passes, and report both results
in the PR rather than the green one.

Adding a self-test is not the rule. Watching it fail is. The policy has the reasoning and the four
occasions this repo has been caught by it.

(Supersedes my earlier comment on this issue, which inlined the rationale instead of pointing at
the policy. The policy is the canonical text.)

[gsdali]

Two of the ten are now carved out with their own issues, because they need more than a rebase.

• Address the review on OCCT#1417 and re-cut carried patch 0018 to match #755 covers OCCT#1417 (patch 0018,
  GCPnts_UniformAbscissa/QuasiUniformAbscissa: two unpatched kernel defects behind #501's buffer-overflow fix (unbounded NbPoints(), degenerate-count SIGSEGV) #555). gkv311 asked for three changes on 2026-08-03, one of which is a design decision about the
  No_Exception guard that lands differently in our build than in a stock one.
• Address the review on OCCT#1418: record the 2021 regression provenance and add the Draw test #756 covers OCCT#1418 (patch 0019,
  GeomConvert_ApproxSurface at GeomAbs_C0 returns a degree-1 collapse while reporting IsDone and a maxError five orders of magnitude too small #522). Not a code change: he identified it as a regression from a 2021 UBSan cleanup and supplied
  a Draw script with the assertions commented out.

Both are dispatched.

A finding from #756 that changes what a rebase of these two is worth. gkv311 attributed #1418 to
3016a390.
Verified, and it is more precise than "a regression since 7.6.0": that commit re-based mma2ce1_'s
workspace indices from ipt1 = iofwr onto ipt1 = isz1, decrementing every slot number by one, and
one of two adjacent mma2jmx_ calls did not get the decrement. Both then wrote the V slot. Our
one-character patch is that missed decrement, so it restores 7.5.x behaviour rather than proposing
anything.

That is worth carrying into the rebases of the other eight: the PR description a maintainer can act
on is the one that names the commit that broke it, and we have not been writing them that way.

Still in scope here: rebasing the remaining eight onto post-8.0.1 master, and filing patch 0020
(#532, BRepFeat_MakeCylindricalHole part selection), which has never been filed upstream at all
and is the only carried patch with no upstream presence.

[gsdali]

Closing. Rebasing the eight remaining upstream PRs is upstream's call to ask for, and they have not
asked.

All eight are open against a pre-8.0.1 base and GitHub reports them mergeable. A maintainer who wants
a rebase says so, and gkv311 has been responsive on this family: he reviewed
#1417 and
#1418 with substantive feedback and asked for
neither. Rebasing eight PRs unprompted rewrites their history, invalidates the review anchors on the
ones that have review, and buys nothing a maintainer requested.

The two that did have unanswered feedback were carved out and are done: #755 for OCCT#1417 and
#756 for OCCT#1418, both merged, both with their replies staged and awaiting authorisation to post.

Patch 0020 (#532, BRepFeat_MakeCylindricalHole part selection), the only carried patch with no
upstream presence at all, is now being filed as its own piece of work. That was the part of this
issue with any real urgency, and it does not need the other eight to move first.

Reopen if upstream asks for a rebase.