It doesn't sound like this is blocking the release of 0.4.2. Should we move it to 0.4.3 with the current backlog of 0.4.2 tickets still being in place?
I've added a comment about the way that we're invoking configure here, and one about CFLAGS is handled.
Additionally, I'd like to request additional review from Teor if possible, on the question of CI speed. Teor has done good work recently on CI performance, and I'd like their opinion on:
whether this will slow down CI much,
whether there is any sensible workaround for that,
and whether we should try to apply the workaround before or after merge.
Trac: Status: needs_review to needs_revision Reviewer: nickm to teor
I've added a comment about the way that we're invoking configure here, and one about CFLAGS is handled.
Additionally, I'd like to request additional review from Teor if possible, on the question of CI speed. Teor has done good work recently on CI performance, and I'd like their opinion on:
whether this will slow down CI much,
whether there is any sensible workaround for that,
and whether we should try to apply the workaround before or after merge.
Thanks Nick!
I'd like to avoid adding any extra jobs, if possible. Instead, we could:
try to build all existing jobs with ALL_BUGS_ARE_FATAL, unless there's some reason we want to turn hardening options off (coverage, distcheck)
make ALL_BUGS_ARE_FATAL apply to Appveyor, as well as Travis
If that causes too many CI failures, we could:
add a few jobs on Travis. I recommended a list of jobs we could add, without adding too much extra time. (Due to Travis parallelism.)
Oh, and if we're skipping some unit tests when ALL_BUGS_ARE_FATAL, let's try to make sure that we still run those tests on some configurations, particularly rare configurations (disabled modules, NSS, macOS).