coverallsapp / coverage-reporter / 32906553634
95%

DEFAULT BRANCH: master
Ran
Jobs 1
Files 44
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: 95.438%. Remained the same
32906553634

push

github

web-flow 
docs: correct the cause behind the libevent pin and its regression gate (#215)

We attributed the v0.6.19-v0.6.21 startup crash to sandboxed container runtimes
(gVisor, Cloud Run, GKE Autopilot). That was a guess, and it was wrong.

The reporting customer's environment turned out to be ordinary Docker and
Kubernetes on RHEL/CentOS 7 hosts -- kernel 3.10, a genuine Red Hat build, no
sandbox and no emulation. CLOCK_BOOTTIME timerfds require Linux >= 3.15
(June 2014), so 3.10 returns EINVAL. Same missing capability, different reason.

That distinction matters for anyone reading these comments later. "gVisor"
invites the reader to conclude it does not apply to them; a kernel floor is
something they can actually check. It also means the exposed population was
larger than we assumed -- RHEL 7 ELS runs to 2028 and enterprise CI fleets are
full of it -- not the niche we described.

Comments only. No build flags, no seccomp rule, no behavior change, so no
release: the `build_cmd` line is byte-identical and the profile's `syscalls`
rule is untouched (only its inert `_comment` key changed). Re-verified the gate
still discriminates against the real published binaries: v0.6.21 exits 1,
v0.6.22 exits 0.

The regression gate itself was always correct -- it tests for the missing
capability, which is cause-agnostic, and would have caught this regardless of
which explanation we believed.

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

1046 of 1096 relevant lines covered (95.44%)

2.22 hits per line

Jobs
ID Job ID Ran Files Coverage
1 32906553634.1 44
95.44
 GitHub Action Run
Source Files on build 32906553634