The ci "test-each-commit" job fetches the PR branch being tested with a depth of (# of commits in PR + 2), and then tries to run tests on commits after the most recent merge commit.
When a PR is opened against a bitcoin core branch, a merge commit is always guaranteed to exist within the fetch depth, because bitcoin core branches always point at merge commits.
However, in fork repositories, pull requests can be opened based on other branches that don't contain recent merge commits, and this will currently cause git rev-list to fail with fatal: bad revision '^^@'.
Work around this problem by not requiring a recent merge commit, and just testing on all fetched commits if a merge commit can't be found.
Co-authored-by: Ryan Ofsky <ryan@ofsky.org>
DANGER_CI_ON_HOST_CACHE_FOLDERS if set will mount caches in directories on the host rather than in docker volumes. Supports saving and restoring caches on Github Actions.
9eea51d905 ci: move Asan / LSan / USDT job to Github Actions (Max Edwards)
4b527fa93b ci: add IPV6 network to ci container (Max Edwards)
Pull request description:
PR for moving the ASAN + LSAN + USDT + friends job to github actions from Cirrus.
The motivation for this PR is that this task needs a full VM (or bare metal) to function, because of the tracepoints. It can not run in a container on an arbitrary Linux, because the outside machine must exactly match the specification of the distro used in the CI task config. This requires more maintenance for the persistent worker, and I think moving to GHA will reduce the maintenance burden, or at least make it possible for anyone to work on.
Also, it makes it easier to run the task on forks (bitcoin-inquisition, bitcoin-knots, devel forks, ...) without having to set-up a real machine.
ACKs for top commit:
maflcko:
review ACK 9eea51d905
achow101:
ACK 9eea51d905
hebasto:
ACK 9eea51d905.
Tree-SHA512: 1111c1c9e3a11e725dff1344643fff3c91fb9b4d7c1cc9a7d507a8f146f5223316a00272030b41ae37ecb59d044f2e90e1cd907450049b25f094f0b60643d4c7
GHA Windows images previously had multiple VC Build Tools installed,
which required specifying the `VCPKG_PLATFORM_TOOLSET_VERSION`
explicitly to avoid linker errors. This issue has been resolved as per
https://github.com/actions/runner-images/issues/9701.
115c283516 ci: add print of powershell version to win64 job (Max Edwards)
Pull request description:
Extraction of just printing powershell version from closed PR: https://github.com/bitcoin/bitcoin/pull/29581
See https://github.com/bitcoin/bitcoin/pull/29581#issuecomment-1984212990 for the cause of a CI failure which was a powershell update.
This PR will make it easier to notice in the future that PS has changed.
ACKs for top commit:
hebasto:
ACK 115c283516. We still use PowerShell in some steps of the "Win64 native" CI job.
Tree-SHA512: 4c7ba9df4f0a98491120326f05e877a995f43a387fe9bbd193549b32f5a4488f85f83e472c9277db457110a7deda04f08832fe6e8129aff4b0b7278be23d4e35
Homebrew attempts to check for outdated dependents or those with broken
linkage. Such behavior might lead to failures when Homebrew updates them
on old macOS images.
This change prevents such behavior.
The GHA VS installation includes its own vcpkg package manager, which is
available since VS 17.6. This change avoids any ambiguity about which
copy of vcpkg we run.
88c8e3a0e4 github actions: Fix test-one-commit when parent of head is merge commit (Ryan Ofsky)
Pull request description:
Instead of figuring out the commit *after* the last merge and rebasing on that with a ~1 suffix, just figure out the last merge commit directly and rebase on it. This way, if HEAD happens to be a merge commit, the rebase just succeeds immediately without blank variables or errors.
Explanation of the problem from https://github.com/bitcoin/bitcoin/pull/28497#issuecomment-1743430631:
> The problem is that the PR only contains a one commit after the last merge, so the job _should_ be skipped, but the `pull_request.commits != 1` check is not smart enough to skip it because the PR is based on another PR and has merge ancestor commits. So specifically what happens is that after HEAD~ is checked out, the new HEAD is a merge commit, so the range `$(git log --merges -1 --format=%H)..HEAD` is equivalent to HEAD..HEAD, which is empty, so the `COMMIT_AFTER_LAST_MERGE` variable is empty and the rebase command fails.
Note: In the current version of this PR, the "test each commit" job is skipped, because this PR only contains a single commit. But I manually verified the code works in earlier versions of the PR that included dummy commits.
ACKs for top commit:
maflcko:
lgtmrecr ACK 88c8e3a0e4
RandyMcMillan:
utACK 88c8e3a
Tree-SHA512: a6865b5c8b96eb0b622b3255971a3cf050dd0f5a356cdfcf7f0cbb659e4a363612e8e62b3ae4fd6b5d9a40bc29176891bc4690659b026c5ef8feea25c8e263cc
Instead of figuring out the commit *after* the last merge and rebasing on that
with a ~1 suffix, just figure out the last merge commit directly and rebase on
it. This way, if HEAD happens to be a merge commit, the rebase just succeeds
immediately without blank variables or errors.
From https://github.com/bitcoin/bitcoin/pull/28497#issuecomment-1743430631:
The problem is that the PR only contains a one commit after the last merge,
so the job _should_ be skipped, but the `pull_request.commits != 1` check
is not smart enough to skip it because the PR is based on another PR and
has merge ancestor commits. So specifically what happens is that after
HEAD~ is checked out, the new HEAD is a merge commit, so the range `$(git
log --merges -1 --format=%H)..HEAD` is equivalent to HEAD..HEAD, which is
empty, so the `COMMIT_AFTER_LAST_MERGE` variable is empty and the rebase
command fails.
Hennadii Stepanov