tarantool / go-tarantool / 34971349072
76%
master: 76%

LAST BUILD BRANCH: bigbes/gh-no-request-accounting
DEFAULT BRANCH: master
Ran
Jobs 1
Files 19
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 76.499% (-0.1%) from 76.595%
34971349072

Pull #604

github

bigbes 
tests: stop the reconnect race workers on every exit path

TestFutureReleaseRaceOnReconnect closed the stop channel only at the end
of the test body, while the restart loop above it uses require. A failed
require calls t.FailNow(), which unwinds with runtime.Goexit() and skips
that close, so the 64 worker goroutines kept looping on an instantly
failing Do() over the already closed connection for the rest of the test
binary. Move the shutdown into a defer registered right after the workers
are spawned, so it runs before the deferred conn.Close() and no worker
touches the connection afterwards.

Bound the join with a timeout as well: a regression that orphans a future
now fails in seconds with a readable message instead of after the ten
minute `go test` timeout and its dump of every stuck goroutine.

Align the instance start and reconnect timings with the other reconnect
tests: RetryTimeout 50ms gave Tarantool 600ms to boot on each of six
starts against the 5.1s every sibling allows, and Reconnect 50ms halved
the per-attempt dial timeout for no benefit - the race window comes from
the in-flight requests, not from the reconnect period.

While at it, name the cleanup the test races against - futureList.clear -
in the comments instead of "future clean" and "futures clear".
Pull Request #604: tests: stop the reconnect race workers on every exit path

3190 of 4170 relevant lines covered (76.5%)

80242.88 hits per line

Coverage Regressions

Lines Coverage ∆ File
4
83.4
 -0.38% connection.go
Jobs
ID Job ID Ran Files Coverage
1 34971349072.1 19
76.5
 GitHub Action Run
Source Files on build 34971349072