Acceptance evidence recorded in each detail file before the content-identical
archive moves:
- G-112: fresh-rebuild bake products BYTE-IDENTICAL to pre-rename vfs products
(punk91 zip-wrap + punk86 kit-wrap; vacuous punkcheck-skip pass caught and
redone), vfslibs phase runs inside bake, aliases + bakehouse dirty-refusal
verified live, docs/layout copies swept, punkexe suites 54/0 green.
- G-113: zero-ESC piped on tclsh90 + punk kit + 8.6, force overrides, real
Windows-console and WSL-pty mode=tty evidence, maketclcolour.test green;
manual-visual note recorded.
- G-104: list/info/build + self-description contract verified incl. a copied
tree; punk-kit hosting; docs in place.
Sweep: G-005/G-018/G-035/G-047/G-099 pending-tense references rewritten with
archive pointers; forward-notes pushed to G-108/G-061/G-105; G-099<->G-108
re-bridged with a direct Related line (their bridges were all archived).
Also: tclsh9.0.5-punk working runtime re-materialized to r2 (stale local
materialization exposed by the window's bake - G-118 piperepl pins green again).
Assisted-by: harness=claude; primary-model=claude-fable-5; api-location=anthropic.com
Acceptance: the registration holder is named (module, proc, mechanism) with reproduction evidence; the disposition is recorded here (deliberate-and-documented, or fixed); a script-mode run of the recipe `chan close stderr; open NUL w` on the current 9.1 kit yields a channel auto-named stderr with working puts stderr, and the console + repl module suites (src/tests/modules/punk/console/, src/tests/modules/punk/repl/) pass.
### G-113 [achieved 2026-07-25] make.tcl colour output is terminal-aware: SGR only on interactive tty, off when piped → detail: goals/archive/G-113-maketcl-tty-aware-colour.md
Scope: src/make.tcl (startup colour policy, define_global_ansi); src/AGENTS.md (invocation guidance); src/tests/shell/testsuites/punkexe/ (piped-output characterization if a suite proves feasible); layout make.tcl copies via established sync channels
Acceptance: With NO_COLOR unset, representative informational and build subcommands (at minimum check, help, workflow, and one build command such as modules) run with stdout redirected to a file produce zero ESC (0x1B) bytes on stdout and stderr, verified on tclsh 9.x and a repo punk kit. The same commands on a real interactive terminal retain colour (manual verification, Windows console plus one unix tty). NO_COLOR=1 still suppresses all colour; a documented force-colour override re-enables colour on piped output. The Tcl 8.6 probe-fallback policy is documented and behaves as decided. The interim src/AGENTS.md NO_COLOR bullet is revised to describe the implemented behaviour.
### G-112 [achieved 2026-07-25] make.tcl subcommand rename: stage-true vocabulary around the promotion gates → detail: goals/archive/G-112-maketcl-subcommand-rename.md
Acceptance: (1) `make.tcl bake` on a clean checkout produces kits byte-comparable to what `make.tcl vfs` produced at the same tree (allowing only known-nondeterministic zip metadata), and additionally runs the vfslibs phase so a bake cannot ship stale binary libs; (2) `make.tcl project` and `make.tcl vfs` still work as deprecated aliases that print a one-line deprecation notice naming their replacement; (3) `make.tcl bakehouse` defaults `-dirty-abort` on and its abort message names the granular developer flow; (4) `make.tcl help`/`workflow` output and every doc in Scope describe the two-persona flow (consumer: `bakehouse`; developer: packages -> test -> bootsupport/vfscommonupdate -> commit -> bake) with no remaining instruction to run bare `make.tcl project` as the recommended path; (5) the punkexe testsuites pass against a kit built with the new names.
### G-104 [achieved 2026-07-25] make.tcl buildsuite surface: list / info / build → detail: goals/archive/G-104-maketcl-buildsuite-surface.md
Scope: src/make.tcl (buildsuite subcommand group); src/buildsuites/*/ (suite self-description contract: an 'info'/describe affordance per suite - suite.tcl action or manifest record); documentation (make.tcl help text, src/buildsuites README/AGENTS)
Acceptance: list shows suite_tcl90 (and any copied tree) with descriptions sourced from the suites themselves; info renders the sources.config-derived table plus zig pin and products for a named suite; build forwards arguments and streams driver output, exiting with the driver's status; the surface works when make.tcl runs under a plain tclsh and under punk kit executables; help text documents the group; the self-description contract is documented where suite authors will find it (src/buildsuites docs) and holds for a tree created by copying suite_tcl90.
Scope: src/buildsuites/suite_tcl90/ (target parameterization of recipe + driver; per-target out prefixes); src/buildsuites/ (layout/naming decision: target is an INVOCATION DIMENSION, suite folders stay version-named - no per-target tree copies); bin/runtime platform-dir naming alignment (punkshell's win32-x86_64-style names vs zig triples); punkbin per-target artifact layout (with G-103 metadata carrying the target); local WSL as the first cross-verification environment
Scope: modpod canonical source (c:/repo/jn/tclmodules/modpod fossil checkout, src/modules/modpod-999999.0a1.0.tm) with re-vendor into shellspy (src/vendormodules/modpod-0.1.5.tm + bootsupport copies via established channels); punkshell-side tarjar artifacts ASSESSED ONLY (cli #tarjar-* module-type branch, libunknown skip-globs, vendored tarjar-2.4.3.tm, src/modules/_build/#tarjar-tarjar-2.3 - dispositions individually user-confirmed, preservation-aware per the c:/tclmodules/main manual-migration intent); new modpod testsuite (home settled in the work; shellspy src/tests/modules/ harness is the natural fit)
Detail: goals/G-111-modpod-tidy-tests.md
### G-112 [active] make.tcl subcommand rename: stage-true vocabulary around the promotion gates
### G-113 [active] make.tcl colour output is terminal-aware: SGR only on interactive tty, off when piped
Scope: src/make.tcl (startup colour policy, define_global_ansi); src/AGENTS.md (invocation guidance); src/tests/shell/testsuites/punkexe/ (piped-output characterization if a suite proves feasible); layout make.tcl copies via established sync channels
# G-104 make.tcl buildsuite surface: list / info / build
Status: active
Status: achieved 2026-07-25
Scope: src/make.tcl (buildsuite subcommand group); src/buildsuites/*/ (suite self-description contract: an 'info'/describe affordance per suite - suite.tcl action or manifest record); documentation (make.tcl help text, src/buildsuites README/AGENTS)
Goal: an ergonomic make.tcl surface for defined buildsuite builds: 'make.tcl buildsuite list' discovers the suites under src/buildsuites (excluding _build and non-suite dirs) with one-line descriptions; 'make.tcl buildsuite info <name>' surfaces the configured detail for review - source records from sources.config (urls/refs/pins), the pinned zig, the product set, doc pointers; 'make.tcl buildsuite build <name> ?args?' runs the suite's driver forwarding suite options (e.g -tclbranch, -refresh, -target when G-105 lands). Suites stay SELF-describing so copy-and-tweak trees (G-096 workflow) appear in list/info automatically without make.tcl edits.
Acceptance: list shows suite_tcl90 (and any copied tree) with descriptions sourced from the suites themselves; info renders the sources.config-derived table plus zig pin and products for a named suite; build forwards arguments and streams driver output, exiting with the driver's status; the surface works when make.tcl runs under a plain tclsh and under punk kit executables; help text documents the group; the self-description contract is documented where suite authors will find it (src/buildsuites docs) and holds for a tree created by copying suite_tcl90.
@ -56,9 +56,15 @@ products appear in info output), G-105 (target dimension forwarded when it exist
suite_tcl90 with an edited description appeared in list/info automatically
and was removed after the check. suite.tcl direct CLI re-verified
(usage-error path through the new config-load position).
- Remains for acceptance: a real driver run through the surface (satisfied
naturally when the G-099 suite_tcl86 arc drives builds via 'make.tcl
buildsuite build'); window doc sweep.
- Acceptance close 2026-07-25 (achieved flip): every acceptance clause is
verified - list (suite_tcl90 + a copied tree, descriptions sourced from the
Goal: make.tcl's subcommand vocabulary names pipeline stages truthfully: `vfs` becomes `bake` (kit assembly only), `project` is retired in favour of `bakehouse` - a consumer umbrella that chains packages+bake and refuses to run on dirty src - and no command silently spans the promotion gates (`bootsupport`, `vfscommonupdate`). `packages`, `modules`, `libs`, `bootsupport`, `vfscommonupdate`, `bin` keep their names.
Acceptance: (1) `make.tcl bake` on a clean checkout produces kits byte-comparable to what `make.tcl vfs` produced at the same tree (allowing only known-nondeterministic zip metadata), and additionally runs the vfslibs phase so a bake cannot ship stale binary libs; (2) `make.tcl project` and `make.tcl vfs` still work as deprecated aliases that print a one-line deprecation notice naming their replacement; (3) `make.tcl bakehouse` defaults `-dirty-abort` on and its abort message names the granular developer flow; (4) `make.tcl help`/`workflow` output and every doc in Scope describe the two-persona flow (consumer: `bakehouse`; developer: packages -> test -> bootsupport/vfscommonupdate -> commit -> bake) with no remaining instruction to run bare `make.tcl project` as the recommended path; (5) the punkexe testsuites pass against a kit built with the new names.
@ -91,9 +91,32 @@ RESOLVED at activation 2026-07-25 (user-directed dispositions):
standalone narrowing - it serves quick iteration on the
src/runtime/vendorlib_vfs.toml declarations without a full bake, matching the
modules/libs narrowings of packages; its helptext names it a phase of bake.
- Remaining for acceptance: doc + layout-copy sweep (Scope docs, three layout
make.tcl copies via sync channels), clean-tree bake-vs-vfs byte comparison,
punkexe suites against a kit built via the new names.
# G-113 make.tcl colour output is terminal-aware: SGR only on interactive tty, off when piped
Status: active
Status: achieved 2026-07-25
Scope: src/make.tcl (startup colour policy, define_global_ansi); src/AGENTS.md (invocation guidance); src/tests/shell/testsuites/punkexe/ (piped-output characterization if a suite proves feasible); layout make.tcl copies via established sync channels
Goal: make.tcl emits ANSI SGR colour only when its output goes to an interactive terminal. Piped/redirected runs (agent harnesses, CI, log capture) produce SGR-free output with no caller action required; interactive terminal runs keep full colour by default; explicit env overrides work in both directions.
Acceptance: With NO_COLOR unset, representative informational and build subcommands (at minimum check, help, workflow, and one build command such as modules) run with stdout redirected to a file produce zero ESC (0x1B) bytes on stdout and stderr, verified on tclsh 9.x and a repo punk kit. The same commands on a real interactive terminal retain colour (manual verification, Windows console plus one unix tty). NO_COLOR=1 still suppresses all colour; a documented force-colour override re-enables colour on piped output. The Tcl 8.6 probe-fallback policy is documented and behaves as decided. The interim src/AGENTS.md NO_COLOR bullet is revised to describe the implemented behaviour.
@ -59,9 +59,15 @@ Agents drive `tclsh src/make.tcl ...` with captured/piped output on every build-
guidance. Remaining manual item: human visual confirmation of colour
appearance on live terminals (the mechanical decision + rendering chain is
fully evidenced above).
- Remains for acceptance: none beyond the recorded manual-visual note; flip is
deferred to the make.tcl-window acceptance pass (with G-112/G-104) so the
layout make.tcl copies sync once for the whole window.
- Window-acceptance close 2026-07-25 (achieved flip): the layout make.tcl
copies synced through the normal thin-layout channels carrying the colour
policy (basic + project-0.1 + templates modpod payload); the
maketclcolour.test suite rode the full punkexe acceptance run green
(54 total, 0 failed). Remaining manual item stands as recorded: human visual
confirmation of colour appearance on live terminals - every mechanical link
(probe decision on real Windows console and WSL pty, colour-on rendering via
the force path, zero-ESC piped on tclsh 9.x + punk kit + 8.6) is evidenced
