Closed amanda11 closed 4 years ago
@armab - Because our spec files has a post-install then it only does the stripping by the brp-strip. And the build root in the cryptography binding .so files Isn’t getting stripped out by that. There is still the question of why we didn’t get this issue before, so I want to do a little more digging before I merge (unless this is needed urgently)
@amanda11 That makes sense, thanks for the deeper investigation on this issue!
@armab @punkrokk I've revised the code so that its less invasive. Now only performs strip on: 1) EL8 (as that's where we build crypography only for) 2) Only strip on .so files where the build root is actually found I've still not located the root cause, could you re-check you are happy with code files.
Also is there a way of getting a E2E run using the packages from this branch? If not, then I'll fire up an EL8 StackStorm and use the st2 package manually, but if we have it setup so that it can use the packages from this branch that would be even better.
I'd say merge the PR and see what happens. We can revert it if e2e tests go wrong or any other regressions spot.
Packages are initially deployed to staging-unstable
package repository. Nightly builds check new packages periodically and if e2e tests succeeded, - promote them to unstable
packagecloud repository. So that's quite a safety net for any e2e regressions.
For some reason based on https://app.circleci.com/pipelines/github/StackStorm/st2/374/workflows/6bb47baf-c2a5-4b11-be22-5f86e318d9f9/jobs/11646 a few
.so
files forEL8
rpm packages started to containBUILDROOT
in them which resulted in failed builds:Call strip for any
.so
libraries, as per logic used in https://github.com/kevinconway/rpmvenv/issues/48 or https://stackoverflow.com/questions/52766091/removing-paths-from-so-files-so-that-rpm-check-buildroot-succeeds.