Files
FemLite/todo.md
T
shenw 5e081806fb
FemLite CI / build-test (push) Has been cancelled
Release FemLite 0.6.1
2026-07-01 22:41:16 +08:00

402 lines
17 KiB
Markdown

# FemLite 0.6.1 TODO
Source plan: `docs/v6.1_development_plan.md`
Version goal:
```text
FemLite 0.6.1 = nonlinear/contact lifecycle diagnostics hardening
```
Boundary:
- 0.6.1 deepens the 0.6.0 lifecycle foundation; it is not a broad algorithm
rewrite release.
- Do not implement a complete industrial AL/mortar consistent tangent rewrite
in this task list.
- Do not move mixed-control continuation, shell preload `K_G`, high-order solid
`K_G`, cohesive contact mainline, nonzero harmonic cyclic, DEM subcycling or
production roll forming into the 0.6.1 must-complete list.
- Every new reason, status, counter or output field must be consumed by a
regression test, checker or report audit.
- Keep TODO item size uniform: each task should add one observable contract and
one focused verification path.
## Execution Order
1. NLHD-1: lifecycle increment summary.
2. NLHD-2: cutback reason taxonomy.
3. NLHD-3: lifecycle report output.
4. NLHD-4: contact state diagnostics.
5. NLHD-5: contact long-path diagnostic smoke.
6. NLHD-6: AL/friction diagnostics.
7. NLHD-7: tangent-audit report shell.
8. NLHD-8: Riks/arc diagnostics.
9. NLHD-9: restart-after-each-accepted audit.
10. NLHD-10: solver diagnostics coupling.
11. DOC-1: documentation and matrix synchronization.
12. REL-1: 0.6.1 regression/version close-out.
## Detailed Execution Map
| TODO | Code targets | Output/API contract | Test/checker target | Not in scope |
|---|---|---|---|---|
| 0.6.1-1 | lifecycle module and diagnostics API | Increment summary fields for lifecycle, convergence, residual, solver and audit state | `nonlinear_lifecycle`, `solver_implicit` | Change Newton algorithms |
| 0.6.1-2 | control/reason taxonomy | Stable cutback reason ids/names and last cutback reason query | `solver_implicit`, lifecycle test | New nonlinear strategy |
| 0.6.1-3 | output/report module | Compact lifecycle report or appended nonlinear report fields | output checker | Break old CSV columns |
| 0.6.1-4 | contact state/report modules | Stable contact state counters and gap/slip/friction fields | `contact`, contact reports | Full contact state-machine rewrite |
| 0.6.1-5 | contact benchmark checkers | Long-path diagnostic smoke for existing compact contact examples | contact benchmark checkers | New contact benchmark physics |
| 0.6.1-6 | multiplier/friction diagnostics | AL multiplier, augmentation and friction trend summaries | `contact_multiplier_report`, `restart` | Complete AL tangent |
| 0.6.1-7 | tangent audit reporting | Shared audit schema and coverage counters | tangent audit tests | Claim production consistent tangents |
| 0.6.1-8 | Riks/arc path diagnostics | Arc residual history, predictor and reversal indicators | `solver_implicit`, `implicit_inp` | Mixed-control continuation |
| 0.6.1-9 | restart writer/reader tests | Multi-increment restart lifecycle audit | `restart` | Large restart benchmark |
| 0.6.1-10 | linear solver diagnostics | Solver status fields linked into lifecycle/nonlinear rows | `linear_solver_cg`, `solver_implicit` | New solver backend |
| 0.6.1-11 | docs and matrices | Updated interface/user/developer docs and benchmark matrix | wording `rg` checks | Expand feature scope |
| 0.6.1-12 | version/report files | Close 0.6.1 after green targeted and V5 acceptance tests | build, targeted tests, acceptance, `git diff --check` | Version bump before tests pass |
## Implementation Rules
- Work in the execution order above.
- Do not mark a TODO done until its new field, reason or status is consumed by
a test, checker or report audit.
- Prefer additive diagnostics over solver-loop rewrites.
- Keep old CSV columns stable; append new fields.
- Keep restart readers backward compatible; optional diagnostic blocks are
preferred over format breaks.
- If a task exposes incompatible designs for contact, Riks, material or restart
ownership, pause and ask for a decision before editing that route.
- If a later task needs a field created by an earlier task, finish and test the
earlier task first.
## 0.6.1 Task List
### [x] TODO 0.6.1-1: Add lifecycle increment summary
- Source:
- `docs/v6.1_development_plan.md` sections 4.1 and 5, NLHD-1.
- Implementation:
- [x] Add a compact lifecycle increment summary structure or query path.
- [x] Include step, increment, final event id/name and reason id/name.
- [x] Include convergence status, residual breakdown, linear solver status
and audit status fields already available in 0.6.0.
- [x] Keep existing lifecycle snapshot APIs compatible.
- Tests:
- [x] Extend `nonlinear_lifecycle` to verify empty and populated summary
fields.
- [x] Extend `solver_implicit` to consume the summary after one accepted and
one rejected/cutback path.
- Acceptance:
- [x] A test can explain the latest nonlinear increment from one summary
query without reading console output.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^(nonlinear_lifecycle|solver_implicit)$" --output-on-failure
```
### [x] TODO 0.6.1-2: Add cutback reason taxonomy
- Source:
- `docs/v6.1_development_plan.md` sections 4.2 and 5, NLHD-2.
- Implementation:
- [x] Add stable cutback reason ids/names for common nonlinear failure modes.
- [x] Track the last cutback reason separately from the final convergence
failure reason.
- [x] Map existing min-step, nonconvergence and linear-solver failure paths
into the new taxonomy.
- [x] Leave unavailable reasons as reserved/documented ids until a real path
consumes them.
- Tests:
- [x] Extend `nonlinear_lifecycle` to verify reason-name stability.
- [x] Extend `solver_implicit` to verify at least min-step and
nonconvergence cutback reasons.
- Acceptance:
- [x] Every implemented 0.6.1 cutback reason is consumed by a regression
assertion.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^(nonlinear_lifecycle|solver_implicit)$" --output-on-failure
```
### [x] TODO 0.6.1-3: Write lifecycle report output
- Source:
- `docs/v6.1_development_plan.md` sections 4.3 and 5, NLHD-3.
- Implementation:
- [x] Add `nonlinear_lifecycle.csv` or append equivalent lifecycle summary
fields to an existing nonlinear report.
- [x] Cover initialized/no-solve, accepted increment and rejected/cutback
increment rows.
- [x] Include lifecycle summary, cutback reason, residual breakdown, solver
status and audit status columns.
- [x] Preserve old `continuation_path.csv` column order if that file is
extended again.
- Tests:
- [x] Add or extend an output checker to validate required lifecycle report
headers.
- [x] Verify at least one accepted row and one rejected/cutback row.
- Acceptance:
- [x] Missing lifecycle report fields cause a regression failure.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^(nonlinear_lifecycle|solver_implicit)$" --output-on-failure
```
### [x] TODO 0.6.1-4: Stabilize contact state diagnostics
- Source:
- `docs/v6.1_development_plan.md` sections 4.4 and 5, NLHD-4.
- Implementation:
- [x] Define stable contact state names: `open`, `closed-stick`,
`closed-slip`, `closed-restick`, `separating` and `failed-projection`.
- [x] Report state counters through existing contact diagnostics or a compact
helper.
- [x] Include gap maximum/mean/L2, slip increment/residual, friction residual
and active-set flip count where available.
- [x] Keep this diagnostic-only; do not rewrite contact solve state
ownership.
- Tests:
- [x] Extend `contact` or contact report tests to verify state names and
counters.
- [x] Verify unsupported/unavailable state fields default consistently.
- Acceptance:
- [x] Contact diagnostics can distinguish open, stick/slip and failed
projection states without parsing logs.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^contact$" --output-on-failure
```
### [x] TODO 0.6.1-5: Add compact contact long-path diagnostic smoke
- Source:
- `docs/v6.1_development_plan.md` sections 4.5 and 5, NLHD-5.
- Implementation:
- [x] Select existing compact examples for two-block compression, block
sliding, shell-shell contact and shell-solid contact.
- [x] Add checker expectations for required diagnostic columns and finite
state counters.
- [x] Check gap/slip/friction/multiplier diagnostics across path steps.
- [x] Keep benchmark physics and tolerances bounded to existing examples.
- Tests:
- [x] Extend existing contact benchmark checkers or their CTest assertions.
- [x] Verify at least two path steps for one compact long-path case.
- Acceptance:
- [x] A compact contact path can fail diagnostics loudly if state or residual
fields disappear.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^contact$" --output-on-failure
ctest --test-dir build-codex-gcc -R "^example_(contact_two_blocks_compression|contact_block_sliding|shell_shell_contact|shell_solid_contact)_benchmark$" --output-on-failure
```
### [x] TODO 0.6.1-6: Add AL multiplier and friction diagnostic summaries
- Source:
- `docs/v6.1_development_plan.md` sections 4.6 and 5, NLHD-6.
- Implementation:
- [x] Add multiplier residual trend and maximum augmented multiplier summary
fields.
- [x] Add augmentation update count and accepted-step penalty/augmentation
factor fields.
- [x] Add slip/friction residual summary fields where existing friction state
supports them.
- [x] Include restart before/after consistency checks for supported fields.
- Tests:
- [x] Extend `contact_multiplier_report` to consume new summary fields.
- [x] Extend `restart` or `contact` to compare supported multiplier/friction
summaries before and after restart.
- Acceptance:
- [x] AL/friction diagnostic summaries are visible in tests and survive the
covered restart path.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^(contact_multiplier_report|restart)$" --output-on-failure
```
### [x] TODO 0.6.1-7: Add shared tangent-audit report shell
- Source:
- `docs/v6.1_development_plan.md` sections 4.7 and 5, NLHD-7.
- Implementation:
- [x] Define a shared tangent-audit report schema with audit name, branch,
sample count, relative error and status fields.
- [x] Route existing material/contact/shell/Riks audit summaries into the
shared vocabulary where already available.
- [x] Add coverage counters for missing or not-applicable audit branches.
- [x] Document that this is an audit report shell, not production consistent
tangent delivery.
- Tests:
- [x] Extend existing tangent audit tests or add a schema checker.
- [x] Verify at least normal contact and one non-contact audit category in
the shared report vocabulary.
- Acceptance:
- [x] Tangent audit output has a uniform schema without changing production
tangent behavior.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^contact_mortar_tangent$" --output-on-failure
```
### [x] TODO 0.6.1-8: Harden Riks and arc diagnostics
- Source:
- `docs/v6.1_development_plan.md` sections 4.8 and 5, NLHD-8.
- Implementation:
- [x] Record arc residual history for accepted Riks/arc increments.
- [x] Add predictor direction consistency and load-factor reversal indicators.
- [x] Add a simple turning-point indicator where current path data supports
it.
- [x] Add path/report fields that distinguish load-control Riks from direct
displacement control.
- Tests:
- [x] Extend `solver_implicit` to verify Riks/arc diagnostic fields.
- [x] Extend `implicit_inp` only if an existing deterministic input path can
consume the fields cheaply.
- Acceptance:
- [x] Riks/arc diagnostics expose residual and predictor/path direction state
without implementing mixed-control continuation.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^(solver_implicit|implicit_inp)$" --output-on-failure
```
### [x] TODO 0.6.1-9: Add restart-after-each-accepted audit
- Source:
- `docs/v6.1_development_plan.md` sections 4.9 and 5, NLHD-9.
- Implementation:
- [x] Add a compact multi-increment restart audit flow.
- [x] After each accepted increment, write restart, read into a fresh model
and compare lifecycle diagnostics.
- [x] Compare supported contact/material/path diagnostic summaries before and
after restart.
- [x] Continue one increment after restart to prove the restored state is
usable.
- Tests:
- [x] Extend `restart` or add a compact restart-lifecycle test.
- [x] Keep the model small and deterministic.
- Acceptance:
- [x] Restart audit covers more than one accepted increment without relying
on a large benchmark.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^restart$" --output-on-failure
```
### [x] TODO 0.6.1-10: Couple solver diagnostics to nonlinear rows
- Source:
- `docs/v6.1_development_plan.md` sections 4.10 and 5, NLHD-10.
- Implementation:
- [x] Link matrix class, solver method, preconditioner, iteration count and
solver status into lifecycle or nonlinear report rows.
- [x] Include initial/final/relative residual where current solver diagnostics
already provide them.
- [x] Add a fallback marker for dense pivoting or method fallback paths if
available.
- [x] Keep solver selection behavior unchanged.
- Tests:
- [x] Extend `linear_solver_cg` for field availability and naming where
needed.
- [x] Extend `solver_implicit` to verify nonlinear rows include solver
diagnostics.
- Acceptance:
- [x] A nonlinear failure can be distinguished from a linear solver failure
through report/API fields.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^(solver_implicit|linear_solver_cg)$" --output-on-failure
```
### [x] TODO 0.6.1-11: Synchronize documentation and benchmark matrix
- Source:
- `docs/v6.1_development_plan.md` sections 4.11 and 5, DOC-1.
- Implementation:
- [x] Update `docs/v6_benchmark_matrix.md` with 0.6.1 diagnostic tests and
evidence placeholders.
- [x] Update interface docs with new lifecycle, cutback, contact and solver
diagnostic fields.
- [x] Update user/developer manuals with the 0.6.1 boundary and deferred
algorithm rewrites.
- [x] Add or update a 0.6.1 acceptance report section.
- Tests:
- [x] Run wording `rg` checks so docs do not claim 0.6.1 delivers complete
AL/mortar tangent, mixed-control continuation or cohesive mainline.
- Acceptance:
- [x] Documentation states that 0.6.1 delivers diagnostic hardening, while
larger solver rewrites remain follow-up work.
Verified:
```powershell
rg -n "0\.6\.1.*(complete|完整|industrial|生产一致切线|mixed-control continuation|AL/mortar consistent tangent).*交付|0\.6\.1.*delivers complete|0\.6\.1.*完整 AL|0\.6\.1.*mixed-control" docs
rg -n "0\.6\.1|nonlinear_lifecycle.csv|arc_mode|fem_tangent_audit_report|closed-stick|linear_fail_class" docs/v6_benchmark_matrix.md docs/v6_acceptance_report.md docs/07_user_manual.md docs/08_developer_manual.md
```
### [x] TODO 0.6.1-12: Update version, report and release evidence
- Source:
- `docs/v6.1_development_plan.md` section 6, REL-1.
- Implementation:
- [x] Update benchmark/acceptance matrix with final 0.6.1 test evidence.
- [x] Update acceptance report with 0.6.1 scope, commands and limitations.
- [x] Raise project version only when 0.6.1 implementation is ready to close.
- [x] Keep deferred boundaries explicit in release documentation.
- Tests:
- [x] Run the planned 0.6.1 targeted regression command.
- [x] Run `ctest --test-dir build-codex-gcc -L v5_acceptance --output-on-failure`.
- [x] Run `git diff --check`.
- Acceptance:
- [x] 0.6.1 closes with green targeted tests, green acceptance subset and
documented deferred boundaries.
Verified:
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^(smoke|perf_contact_pair_p1)$" --output-on-failure
ctest --test-dir build-codex-gcc -R "^(nonlinear_lifecycle|solver_implicit|implicit_inp|restart|contact|contact_mortar_report|contact_multiplier_report|cohesive_contact|particle_acceptance|linear_solver_cg)$" --output-on-failure
ctest --test-dir build-codex-gcc -L v5_acceptance --output-on-failure
git diff --check
```
## Suggested 0.6.1 Regression Commands
```powershell
cmake --build build-codex-gcc --parallel
ctest --test-dir build-codex-gcc -R "^(nonlinear_lifecycle|solver_implicit|implicit_inp|restart|contact|contact_mortar_report|contact_multiplier_report|cohesive_contact|particle_acceptance|linear_solver_cg)$" --output-on-failure
ctest --test-dir build-codex-gcc -L v5_acceptance --output-on-failure
git diff --check
```