You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 30b2bc8
Browse filesBrowse the repository at this point in the historyBrowse files
lane-u5a-2359
committed
test: give back the leaf lock when a coexist row fails before its release
The three rows in the coexistence block acquire a lock in this process and release
it inline, so an assertion that lands between the two leaves the lock held with its
renewal timer keeping the artefact fresh. That is what the finding on row 4555925401
says about the row that reaps its child in a finally and never releases the leaf,
and the same hole sits in the row beside it whose two releases follow its last
assertion and in the posix row whose two releases follow its own.
Guard the failure path for the block, in the shape the merge block guard took in the
previous commit: record what each row is holding, clear the record when the row lets
go, and in the afterEach release only what is still recorded, clearing before awaiting
so a release that lands late is never attempted twice and the row that failed still
reports its own assertion. The success paths did not move, and the diff is 81 added
lines with no removed line.
Pin it with two rows that need no alias and no symlink, so they are measured
wherever the block runs rather than only where short names are generated. The first
forces the failure the finding names, an assertion that lands while the leaf lock is
held and before any release line in the block, and records that the artefact is still
there at that instant, which is the hole and not the fix. The second runs after the
guard and asserts what the guard left, namely no artefact under this process and the
same path acquirable again by an immediate call. Measured by taking the guard away:
without it the file printed 1 failed, 32 passed and 8 skipped of 41 with the message
that an artefact list containing the leaf lock directory did not equal an empty list;
with it the file printed 33 passed and 8 skipped of 41 and the misc lane printed 2127
passed and 21 skipped of 2148 over 119 files where the commit before this one printed
2125 passed and 21 skipped of 2144. The prettier check, the type check and the lint
pass came back clean on both sides. No production file changed, so the priced unit
delta against the unit base is unchanged, and the mutation leg for these commits is
unmeasured locally.
0 commit comments