Copilot on #297: `pr-reviews` only calls `date +%s`, so the REAL_DATE
pass-through never ran. It is deleted rather than tested. The fake now refuses
any other arguments with exit 2, so a new `date` call shows up as a failure.
Rival: with pr-reviews calling `date -u +%s`, "a review arriving on the third
poll is seen" and "--wait keeps polling until its deadline" fail (23/25).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
This follows the user's rulings: CI should not care which bash or which
standard utilities it runs, and jobs take what the ci dev shell provides.
- checkers.sh: remove find_modern_bash, the 5.1 floor (improvise-fd0j) and
exit 3. Suites run under $BASH with PATH unchanged.
- checkers-test.sh: remove sections 2, 5 and 5b (the fake bashes and utilbin).
They existed only to pin the floor.
- scripts/kb: drop the BSD `stat -f` fallback. kb needs the dev shell for tpl,
and that shell's stat is GNU.
- ci.yml: remove every `shell: bash` override, so every step runs in the dev
shell. Remove the web-toolchain version probe and the Playwright
version/browser probe. The shell's PLAYWRIGHT_BROWSERS_PATH pins the
browsers. npm ci stays because @playwright/test is not nix-provided.
- flake.nix: update the comments that promised those CI checks.
Reasoning: context/code/tooling.md section 5. Evidence:
context/test-evidence/tooling.md.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
On the NixOS runners /bin holds only sh. With no /bin/sleep, the watchdog in
bounded_reviews exits at once and wins the `wait -n`, so every --wait case
reports 124. Both tools are now resolved with `type -P` before any case stubs
PATH.
Red: with both paths pointed at /nonexistent/, the suite fails 25 of 25.
Rival: REAL_SLEEP=/nonexistent/sleep fails 25 of 25 again. The date
pass-through has no discriminating rival, because pr-reviews only calls
`date +%s`. Evidence: context/test-evidence/tooling.md.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
Section 5b (the bash 5.1 floor, #237) was written on a branch cut before the
NixOS-runner change (af44653f). That change moved section 5 off /usr/bin:/bin,
because on NixOS that path holds only env and sh. The two merged with no
conflict. On NixOS the runner then dies on `dirname` with exit 1, which turned
the KB job red on every PR at 666de199.
5b now uses utilbin too, and utilbin gains head and env for the 5.1 control.
Evidence and rivals: context/test-evidence/tooling.md.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
This head already carries #272 (p6v6), merged at d9c65181.
Conflicts:
- resolve.rs advanced: main's `reopened_at` (decline an aggregate or a
formula's result) now takes this branch's `writes` and answers the
decline arm first, then `awaiting_seed` or `opened_on(state, ..)`, as
the #282 review prescribed for this combination.
- context/test-evidence/{clients,command}.md: both sides' entries kept.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
nixpkgs' WebKit (WPEWebProcess) aborts on its first page inside the
self-hosted runner VMs, though not on the host with the same sandbox;
its OPFS specs skip on a runtime probe anyway. Re-add --project=webkit
once the abort is understood.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Every CI build is now a derivation in nix/ci.nix, so the self-hosted
runners substitute it from titan's harmonia cache instead of rebuilding:
- checks.clippy, checks.test (with the --check-evidence gate, which
needs the compiled tests), checks.wasm-check (per crate, no feature
unification), checks.harness-lint (opfs-contract fmt + clippy)
- packages.web: harness/build.sh unchanged -> the three wasm bundles
and harness/serve; packages.site re-assembles with the commit as the
cache-bust token
- packages.ci-deps: the crane dependency builds, for titan's cache
timer to keep rooted
Dependencies are separate buildDepsOnly derivations keyed on the lock,
so a PR that leaves Cargo.lock alone rebuilds only the workspace. The
script-only checks (node, kb, checkers, comment/CI-graph gates, fmt)
stay dev-shell steps: they compile nothing, and their enforced
`#!/usr/bin/env bash` cannot run in the build sandbox.
ci.yml's build, test, wasm and harness-lint jobs now run `nix build` on
these; rust-cache and the PATH workaround for it are gone. Job names
and the deploy graph are unchanged.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
GIT_CONFIG_GLOBAL=/dev/null and GIT_CONFIG_NOSYSTEM=1 in run_git were
witnessed by a config excludesFile, which vwjj's -c
core.excludesFile=/dev/null outranks. So dropping either line passed
every test from vwjj onwards.
The hostile child's config, read as both GIT_CONFIG_GLOBAL and
GIT_CONFIG_SYSTEM, now also sets core.sharedRepository = nonsense.
Measured on git 2.55: git init and git add die on it, git ls-files
does not read it (so the walk, which does not null the user's config,
is unaffected), and run_git passes no -c for it. init.defaultObjectFormat
and init.defaultRefFormat were tried first; they only warn.
Dropping either env line is now red. Every other rival in the combined
test is unchanged, and there is still one child per test.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
The two generator strategies re-parsed the grammar inside their
prop_map, once per generated case: the same defect improvise-8shu fixed
in improvise-io. A OnceLock takes the lib test binary from ~0.9s to
~0.06s on the same machine and load.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
tests/shipped_documents_load.rs re-ran itself in 18 child processes, one
per hostile setting, and each child spawns several gits. So the binary
took 2.1-3.3s against design-principles 7's 2s-per-binary tripwire.
The six per-setting tests (53zz x2, vwjj, pr3i, a0i3 and 2iws/vav0) are
now one test, a_hostile_git_environment_changes_nothing. It has two
children: the gzip shape test and every_shipped_document_loads. Each
carries every setting at once:
- the template dir;
- the global and system config (with trace2 targets);
- XDG_CONFIG_HOME;
- GIT_DIR and GIT_INDEX_FILE;
- the five pathspec and attribute settings;
- the six trace variables.
Afterwards the test asserts the outer index is still empty.
Each setting breaks the walk on its own, so a rival that drops one
defence leaves that setting live, and the child fails. Every rival that
was red against the per-setting tests is red against the combined one.
The rivals that were green before are green now:
- ICASE;
- the trace2 names under the 0 override;
- the GIT_CONFIG_GLOBAL and NOSYSTEM nulls, whose excludes witness vwjj's
-c override masks. That last pair is filed as improvise-3j70.
Measured on an Apple M4 Max, three workspace runs each: this binary went
from 2.10-3.31s (load 37-39) to 0.57-0.91s (load 30-33). It is now
below the next-largest binary in every run.
The evidence sections keep their reasoning, under prose headings.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
The review of #276 found that an exact list of the three names the test
set passed it. The child now sets all six variables that write to their
target on git 2.55: GIT_TRACE, _SETUP, _PERFORMANCE, 2, 2_EVENT and
2_PERF. An exact three-name list leaves GIT_TRACE_SETUP behind and
fails.
vav0: a trace2.*Target in user config is not an environment variable,
so removal cannot reach it, and the walk's git does not null the user's
config. untraced_git now also sets GIT_TRACE2, GIT_TRACE2_EVENT and
GIT_TRACE2_PERF to 0, which outranks the config. The child adds a
GIT_CONFIG_GLOBAL whose [trace2] section sends all three targets to
/dev/stdout.
Rivals, each of which fails:
- no strip;
- the exact three names;
- the prefix minus GIT_TRACE, minus _SETUP, or minus _PERFORMANCE;
- no 0 override.
The evidence entry moves from the web_demo_model section to the
shipped_documents_load one.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
GIT_TRACE, GIT_TRACE2_EVENT and GIT_TRACE_PERFORMANCE are process-wide,
so --local-env-vars lists none of them. Pointed at /dev/stdout, git
wrote trace lines into the ls-files -z output, and the walk read them
as paths. git() now builds on untraced_git(), which removes every
variable starting with GIT_TRACE: a prefix, because the family is
open-ended. The rev-parse --local-env-vars call goes through it too,
since its stdout is parsed as well.
the_walk_ignores_a_git_trace_sent_to_stdout reruns the gzip shape test
and every_shipped_document_loads in children with all three set to
/dev/stdout. It failed at #269's head. Rivals, each of which fails it:
- no strip;
- the exact name only;
- keep GIT_TRACE_PERFORMANCE;
- keep GIT_TRACE2_EVENT.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
The advance no longer reaches the commit door on a fold or a formula's
cell, so the two older advance tests stopped seeing it (measured: the
cell_key rival passes them). Two tests seat the editor there directly.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
EnterEditAtCursor asks core::uneditable, the commit door's own question,
now shared with committable_cell_key. The browser's advance asks the
projected id at the landing cell; lockstep pins both ends on a fold and
on a formula's cell.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
mdfc put the "any unanswered write" check inside `advanced` alone, so an
editor opened with i/a (`edit_or_drill`) still seeded from the cache the
write had made stale: `i 9 Tab Esc i 5 Enter`, typed faster than the
answer, stored 15 in the browser where the TUI stores 195.
The in-flight half now lives in `opened_on`, so no opener can forget it.
`advanced` keeps the "this frame writes" half, since the commit being
resolved is not in flight yet; dropping it fails the 3tt4/3uzg tests.
Stacked on #227 (fix/fy9o-reopened-editor-reseeded, de4a7d78).
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
GIT_LITERAL/GLOB/NOGLOB/ICASE_PATHSPECS and GIT_ATTR_SOURCE are
process-wide, so --local-env-vars names none of them. Remove them in
git() beside the local-vars loop.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
improvise-uln5. The browser painted every help-overlay row plainly,
including six whose keys go upstream to a host that answers nothing:
/, n, H, z, Enter, and search mode's Enter/Esc. Every ex-command row was
painted plainly too, although `:` opens a command line the browser
cannot run. A dead key is worse than a gap: the keys typed after it land
on live bindings (improvise-koh6).
- resolve::NEEDS_THE_HOST is now the one list of names this client has
no arm for. KeyRow keeps the command it was built for.
- The browser's RenderHelpRow draws a listed row, and every ex row while
the command line is stranded, as `dim` with "(not in the browser yet)".
It marks rather than omits: both front ends keep drawing the same rows,
the feature stays findable, and web/ needs no change.
- the_command_names_that_still_need_the_host_are_these now drives that
list rather than a table of its own.
- every_help_row_the_browser_cannot_run_is_one_it_marks holds it from
the other side: silent help rows == listed rows, and no hint line
names one.
- every_advertised_command_answers_in_its_own_mode no longer restates
the black holes.
Red: a_help_row_the_browser_cannot_run_is_drawn_marked at 202e7eac.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VdfEGXA7aJ6ftMsCc9PJze
Ed Langley