fix: discard-dist-on-failure keeps the step's status when its message cannot be written (closes #342)

With stderr closed, the wrapper's message write failed and set -e ended it
with status 2; with stderr a pipe whose reader had gone, the write killed it
with 141. The message is now written with SIGPIPE ignored and its failure
ignored, after the step has run. An interrupt while a step runs now exits with
130 once the step has ended, removing nothing: under bash, a step that caught
the interrupt and exited with a status, as check-censored does, used to get
dist/ removed. A failed check-censored --require-dist still removes dist/. The
header states both. Also the README bullet's missing period.

Model: opus-5-5

README.md

@@ -277,7 +277,7 @@ provide:
   step's exit status, saying on stderr that it did and why. Every step of
   `make build` runs through it; `make build-debug` runs none of them through it.
   A step that succeeds removes nothing, and a removal that cannot be completed
-  is reported as loudly as one that was
+  is reported as loudly as one that was.
 - `script/test-verify-build` — exercise every failure mode of
   `script/verify-build` against a fixture tree in a temp dir, asserting the exit
   status and the message of each, assert the state of `dist/` on disk after a
@@ -1838,6 +1838,9 @@ view would leave a wallet one click from deletion.
       try again." on the error line, nothing deleted
     - "I have lost my password" → **DeleteWalletLostPassword**
     - "Back" → previous screen (Settings)
+    - Settings gear → **Settings**, whose "Back" never lands back on this
+      screen: leaving drops the wallet the screen was showing, so it also takes
+      the screen off the Back stack
 
 #### DeleteWalletLostPassword (`delete-wallet-lost-password`)
 
@@ -1876,6 +1879,9 @@ view would leave a wallet one click from deletion.
       selection comes back with it. The two delete screens are siblings rather
       than parent and child: nothing is pushed on the way here, so both have
       Settings as their Back target.
+    - Settings gear → **Settings**, whose "Back" never lands back on this
+      screen: leaving drops the wallet the screen was showing, so it also takes
+      the screen off the Back stack
 - **Deliberately not password-gated.** A password in front of _discarding_ a
   secret protects nobody: an attacker at the popup who wants the wallet gone can
   uninstall the extension, so the only person such a gate stops is the owner who

TODO.md

@@ -45,6 +45,29 @@ but the review is broader than any of them.
 
 # Completed Steps
 
+- 2026-10-07: `script/discard-dist-on-failure` returns the failed step's own
+  exit status even when it cannot write its message, to a closed stderr or to a
+  pipe nobody reads any more
+  ([#342](https://git.eeqj.de/sneak/AutistMask/issues/342)); it used to return 2
+  or 141 instead. An interrupt while a step runs now removes nothing and says
+  nothing whichever shell `/bin/sh` is, even when the step catches it and exits
+  with a status of its own, as `script/check-censored` does: the wrapper exits
+  with 130 once the step has ended. Under bash such a step used to have `dist/`
+  removed. A failed `check-censored --require-dist` still removes `dist/` like
+  any other step, because a `dist/` not cleared of the name `RULES.md` bars must
+  not ship. The header states both. `script/test-verify-build` runs the wrapper
+  with stderr closed and with stderr a pipe nobody reads, and interrupts it
+  under dash and under bash.
+
+- 2026-10-06: Back from Settings no longer lands on the delete wallet or
+  lost-password screen after either was left by the settings gear
+  ([#480](https://git.eeqj.de/sneak/AutistMask/issues/480)), the defect
+  [#461](https://git.eeqj.de/sneak/AutistMask/issues/461) fixed for the two
+  secret screens. Leaving drops the screen's wallet selection, so its leave
+  handler now also takes it off the Back stack, as a reopened popup already
+  does. `tests/deleteWalletLostPassword.test.js` drives the gear and then Back
+  on both screens.
+
 - 2026-10-06: `script/bootstrap` no longer reports success while node cannot
   find a package listed in `dependencies` or `devDependencies` of `package.json`
   ([#263](https://git.eeqj.de/sneak/AutistMask/issues/263)). After the install

script/discard-dist-on-failure

@@ -3,22 +3,21 @@
 # step fails, remove dist/ before returning its exit status. Our own extension
 # to scripts-to-rule-them-all, wrapped around every step of make build.
 #
-# Why: with AUTISTMASK_DEBUG=1 exported in the calling shell, make build
+# Why: with AUTISTMASK_DEBUG=1 exported, make build compiles a debug bundle and
-# compiles a debug bundle and then fails on it in script/verify-build — but the
+# then fails on it in script/verify-build, after the bundle is written. It is
-# bundle is already written. It is loadable, and every wallet it creates gets
+# loadable, and every wallet it creates gets the publicly committed test
-# the publicly committed test recovery phrase from src/shared/constants.js. A
+# recovery phrase from src/shared/constants.js, so a failed build must not leave
-# failed release build that leaves that behind is a smaller version of the trap
+# it behind. The removal is never silent: it says on stderr that dist/ is gone
-# the verifier exists to close, and "the failure was loud" only works on an
+# and why.
-# operator who does not load dist/chrome/ anyway. Removing the artifact does not
-# depend on that.
 #
-# Two things this deliberately does not do. It does not wrap make build-debug: a
+# A failed check-censored --require-dist removes dist/ like any other step: a
-# debug build that failed is not a mistakable artifact, and its output is the
+# dist/ not cleared of the name RULES.md bars must not ship either. A step that
-# evidence of what went wrong. And it never removes anything on a step that
+# succeeds removes nothing. An interrupt (Ctrl-C) while a step runs removes
-# SUCCEEDS, including the final check-censored --require-dist pass.
+# nothing and says nothing, even when the step catches it and exits with a
-#
+# status of its own: the wrapper exits with 130 once the step has ended. An
-# The removal is never silent: it says dist/ is gone and why, on stderr, above
+# interrupt is not a build failure, and whoever interrupted the build knows it
-# the build's own failure.
+# did not finish. make build-debug is not wrapped: a debug build that failed is
+# not a mistakable artifact, and its output is the evidence of what went wrong.
 set -eu
 
 ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
@@ -66,12 +65,20 @@ main() {
         exit 1
     }
 
+    # An interrupt is not a build failure: once the step has ended, exit
+    # without removing anything, even if the step caught the interrupt and
+    # exited with a status of its own. bash as /bin/sh would otherwise carry on.
+    trap 'exit 130' INT
+
     _status=0
     "$@" || _status=$?
 
     [ "$_status" -ne 0 ] || return 0
 
-    discard_dist
+    # A message that cannot be written, to a closed stderr or to a pipe nobody
+    # reads any more, must not replace the step's status.
+    trap '' PIPE
+    discard_dist || true
     exit "$_status"
 }
 

script/test-verify-build

@@ -915,6 +915,76 @@ run_cases() {
     discard_case "the wrapper given no command removes nothing" \
         c_control 1 kept "no command given" ""
 
+    # With stderr closed the wrapper cannot write its message, and must still
+    # remove dist/ and return the step's own status.
+    build_fixture
+    _status=0
+    (cd "$FIXTURE" &&
+        "$FIXTURE/script/discard-dist-on-failure" sh -c 'exit 6' 2>&-) ||
+        _status=$?
+    if [ "$_status" -eq 6 ] && [ ! -e "$FIXTURE/dist" ]; then
+        PASSED=$((PASSED + 1))
+        echo "  ok: a failed step's status survives a closed stderr"
+    else
+        FAILED=$((FAILED + 1))
+        echo "  FAIL: a failed step's status survives a closed stderr"
+        echo "    exit status $_status, wanted 6, and dist/ must be gone"
+    fi
+
+    # The same with stderr a pipe nobody reads any more, where the write would
+    # kill the wrapper with SIGPIPE. The FIFO's only reader opens it, exits and
+    # is waited for before the wrapper runs, so the pipe never has a reader.
+    build_fixture
+    mkfifo "$WORK/stderr-fifo"
+    _status=0
+    (
+        : <"$WORK/stderr-fifo" &
+        exec 3>"$WORK/stderr-fifo"
+        wait "$!"
+        cd "$FIXTURE" &&
+            "$FIXTURE/script/discard-dist-on-failure" sh -c 'exit 6' 2>&3
+    ) || _status=$?
+    if [ "$_status" -eq 6 ] && [ ! -e "$FIXTURE/dist" ]; then
+        PASSED=$((PASSED + 1))
+        echo "  ok: a failed step's status survives a pipe nobody reads"
+    else
+        FAILED=$((FAILED + 1))
+        echo "  FAIL: a failed step's status survives a pipe nobody reads"
+        echo "    exit status $_status, wanted 6, and dist/ must be gone"
+    fi
+
+    # An interrupt while a step runs removes nothing and says nothing, even
+    # when the step catches it and exits with a status of its own, as
+    # script/check-censored does. The step interrupts the wrapper and then
+    # itself, as Ctrl-C interrupts every process of the build at once. Run
+    # under dash and under bash, which /bin/sh may each be: bash carries on
+    # after such a step unless the wrapper stops it.
+    for _shell in dash bash; do
+        _name="an interrupt under $_shell removes nothing"
+        if ! command -v "$_shell" >/dev/null 2>&1; then
+            SKIPPED=$((SKIPPED + 1))
+            SKIPPED_NAMES="$SKIPPED_NAMES##   - $_name ($_shell not found)$NEWLINE"
+            echo "  SKIP ($_shell not found): $_name"
+            continue
+        fi
+        build_fixture
+        _status=0
+        _out="$(cd "$FIXTURE" && "$_shell" \
+            "$FIXTURE/script/discard-dist-on-failure" \
+            sh -c 'trap "exit 4" INT; kill -INT "$PPID" $$; exit 5' 2>&1)" ||
+            _status=$?
+        if [ "$_status" -eq 130 ] && [ -z "$_out" ] &&
+            [ -f "$FIXTURE/dist/chrome/src/popup/index.js" ]; then
+            PASSED=$((PASSED + 1))
+            echo "  ok: $_name"
+        else
+            FAILED=$((FAILED + 1))
+            echo "  FAIL: $_name"
+            echo "    exit status $_status, wanted 130; dist/ must be intact" \
+                "and nothing said. Output: $_out"
+        fi
+    done
+
     check_makefile_wiring
 }
 

src/popup/views/deleteWallet.js

@@ -49,9 +49,9 @@ function confirmKey(name) {
 }
 
 // Drop the password from the DOM and the wallet selection from the
-// closure. Registered as the view-leave handler as well as run on entry,
+// closure. Run by the view-leave handler as well as on entry, so the typed
-// so the typed password does not sit in the hidden view after the user
+// password does not sit in the hidden view after the user navigates away
-// navigates away by any route, including the Settings gear.
+// by any route, including the Settings gear.
 function clear() {
     deleteWalletIndex = null;
     $("delete-wallet-password").value = "";
@@ -147,8 +147,25 @@ async function finishDelete(walletIdx) {
 function init(_ctx) {
     ctx = _ctx;
 
-    onViewLeave("delete-wallet-confirm", clear);
+    // Leaving drops the wallet selection, so each screen also comes off the
-    onViewLeave("delete-wallet-lost-password", clearLostPassword);
+    // Back stack, where the settings gear has just put it: Back from
+    // Settings must not land on a screen whose button can only answer "No
+    // wallet selected for deletion." A reopened popup drops them from the
+    // stack the same way (https://git.eeqj.de/sneak/AutistMask/issues/480).
+    onViewLeave("delete-wallet-confirm", () => {
+        clear();
+        const stack = state.viewStack;
+        if (stack[stack.length - 1] === "delete-wallet-confirm") {
+            stack.pop();
+        }
+    });
+    onViewLeave("delete-wallet-lost-password", () => {
+        clearLostPassword();
+        const stack = state.viewStack;
+        if (stack[stack.length - 1] === "delete-wallet-lost-password") {
+            stack.pop();
+        }
+    });
 
     // No wipe here: goBack() routes through showView(), which runs the
     // leave hook.

tests/deleteWalletLostPassword.test.js

@@ -615,3 +615,44 @@ describe("the password route's confirm button", () => {
         ]);
     });
 });
+
+// https://git.eeqj.de/sneak/AutistMask/issues/480: leaving either delete
+// screen drops its wallet selection, so Back onto one showed a screen whose
+// button could only answer "No wallet selected for deletion."
+describe("Back from Settings after leaving by the settings gear", () => {
+    test("does not land on the delete screen", () => {
+        const { helpers, deleteWallet, state } = load();
+        deleteWallet.show(1);
+        // The settings gear: push the current view, then show Settings.
+        helpers.pushCurrentView();
+        helpers.showView("settings");
+
+        expect(state.viewStack).toEqual(["main", "settings"]);
+        helpers.goBack();
+        expect(state.currentView).not.toBe("delete-wallet-confirm");
+    });
+
+    test("does not land on the lost-password screen", async () => {
+        const { helpers, deleteWallet, state } = load();
+        await openLostPassword(deleteWallet, 1);
+        // The settings gear: push the current view, then show Settings.
+        helpers.pushCurrentView();
+        helpers.showView("settings");
+
+        expect(state.viewStack).toEqual(["main", "settings"]);
+        helpers.goBack();
+        expect(state.currentView).not.toBe(VIEW);
+    });
+
+    // The lost-password screen's own Back is "Back returns to the delete
+    // screen with its wallet still chosen", above.
+    test("the delete screen's own Back leaves the rest of the stack alone", async () => {
+        const { deleteWallet, state } = load();
+        deleteWallet.show(1);
+
+        await click("btn-delete-wallet-back");
+
+        expect(state.currentView).toBe("settings");
+        expect(state.viewStack).toEqual(["main"]);
+    });
+});