GeezerVison

…ed ones

Clarify, that the scheduler callback must not be used to modify
the internal state of stackless.

Fix demo/tracing.py:
Don't use flextype features. Unfortunately it is no longer possible
to change the __repr__ method of class tasklet.

Fix tracing.py, fakechannel.py and fakechannel2.py and update
the copy of tracing.py in the documentation.

Remove Stackless/demo/paelike/PyOpenSteer and
Stackless/demo/stephan/stacklessness.

https://bitbucket.org/stackless-dev/stackless/issue/83
(grafted from 410ed1df77dfae9b48519a06a4c232d65caa9745, c5cac8a05d68 and
fb182bb2be4c)

…ing module. Ref python#83.

…ixes python#83.

Alternate fix for ImportError on Python 3.5.{0-3} with flake8-typing-imports

Closes python#83

See merge request python-devs/importlib_resources!91

update readme

(cherry picked from commit 75c2b66)

update readme

(cherry picked from commit 75c2b66)

update readme

(cherry picked from commit 75c2b66)

update readme

(cherry picked from commit 75c2b66)

update readme

(cherry picked from commit 75c2b66)

…(6) script-driven counts

Per supervisor [chat L2323] discipline-mechanization sprint:
- (5) Refuting build-class claims (filesystem-first): any agent disputing
  BUILD PASS MUST cite filesystem evidence (stat -c %Y, .o, grep build
  stdout) before warning. Session-log-only is insufficient (Bash-tool
  builds not visible in nbs-ts). Per medic 2026-04-22 L2226 false-
  positive (escalated to HALT + Alex-threat before filesystem check).
- (6) Counts in commit messages and chat reports (script-driven): any
  quantitative claim MUST come from script output (count_w25_b1b_tus.sh,
  count_emit_methods.sh, etc.) — not memory. Quote script cmd + output.
  Per 4+ /N-class lapses despite feedback memory.

Closes 2 of 4 outstanding discipline extensions per pythia python#83 +
theologian L2322. Other 2 (transitive-verify, mid-flight-build) remain
contextual / operational guidance per supervisor judgment.

§5.3 Signature-drift reproducer:
- Empirical PASS end-to-end recorded (PRE-STEP-B BUILD_EXIT=0 + POST-STEP-B
  BUILD_EXIT=2 with 6 compile errors at licm_c.c + pass_output_type_c.c +
  simplify_c.c expected sites)
- Acceptance per spec §5 satisfied for struct-typed handle args
- Caveat: PASS is for HirBasicBlock*/HirCFG* args ONLY; void*-aliased
  handles (HirInstr/HirRegister/HirFunction) still drift-prone

§7 What this does NOT solve — expanded per pythia python#83:
- HirInstr/HirRegister/HirFunction void* aliases — drift remains possible
  for these handle types because type-checker can't distinguish void* from
  void*. W25b promotion to canonical struct pointers required to close.
- Function-scoped extern decls — ^extern lint pattern (line-anchored)
  misses indented intra-function externs. Caught empirically in B-5
  (2 hir_bb_* externs surfaced via compile error not lint). Step C lint
  pattern revision proposed.

Per supervisor [chat L2328] task assignment + testkeeper L2327 execution
+ spec §5 acceptance gate. End-to-end §5.3 falsification COMPLETE.

EMPIRICAL RESULT (testkeeper executed):
  Mutation: hir_c_insert_before +3rd arg in BOTH hir_c_api.h:740 + hir_c_api.cpp:2349
  Build cmd: scripts/build_phoenix.sh
  HEAD: ab07641 (post-Step-B substantive close)
  BUILD_EXIT=2 ← DRIFT CAUGHT at compile time

  Compile errors (6 call sites across 3 §1b TUs):
    licm_c.c:244:50
    pass_output_type_c.c:981:53, 989:48, 1054:59
    simplify_c.c:40:57, 2829:54
  All 'too few arguments to function call, expected 3, have 2'.

§5.3 ACCEPTANCE PER SPEC §5: PASS END-TO-END
  PRE-STEP-B baseline (push 67 commit 911e173): BUILD_EXIT=0 = drift undetected
  POST-STEP-B re-run (HERE): BUILD_EXIT=2 = drift caught at compile time
  Step B closed §1b drift surface as designed.

CHANGES (+22/0):
  Appended POST-STEP-B section to docs/w25-step-b-mutation-baseline.txt
  (PRE-STEP-B section landed in push 67 commit 911e173).

INTERPRETATION:
  Pre-Step-B: §1b TUs each held a local 2-arg extern shadowing canonical
    decl. 2-arg call sites compiled fine; linker silently bound to impl
    by name. Result: silent runtime UB at runtime if call sites executed
    with mismatched arg counts.
  Post-Step-B: local externs DELETED + #include hir_c_api.h ADDED in
    each §1b TU. New 3-arg canonical sig visible at compile time.
    2-arg call sites now fail compile with clear errors.
  Drift caught BEFORE link; no silent UB possible.

NOW LEGITIMATELY: Step B closes §1b drift surface — empirically validated
per spec §5 acceptance criterion. Per pythia python#83 python#2 framing-precision:
mechanical work + dual-typedef collision resolution landed at push 75
ab07641; formal §5.3 closure landed HERE.

AUTHORSHIP: testkeeper executed §5.3 procedure (chat L2327); generalist
commits artifact per role boundary (push 67 precedent).

Authorization chain:
  - W25 spec §5: theologian L2017 (procedure design)
  - §5.3 procedure with corrections: 2ed74a9 doc landed in push 67
  - testkeeper EXECUTED + verbatim output: chat L2327
  - Supervisor task assignment: chat L2328