The standard and fuzz matrix jobs share the same github.job value
(windows-native-dll), so both try to save the vcpkg tools cache with the
same key.
Since the tools are identical across build types, let them share a
single cache entry by restricting the save to the standard job only.
CMAKE_COMPILE_WARNING_AS_ERROR is not a cache variable by default in
CMake, so it has no value in the configure summary when not set, and
even when set cannot be toggled in ccmake. Define it as an option() to
make it a cache BOOL with a default of OFF.
The vcpkg tools cache was using the combined actions/cache action,
which saves on every run regardless of branch. Split it into the
restore/save pattern used by the other caches, so that saves only
happen on default branch pushes.
The current SHA1 LLVM signing key is considered not secure since
2026-02-01T00:00:00Z which makes this run fail when downloading
packages.
See: https://github.com/llvm/llvm-project/issues/153385
Apply the fix from the issue to temporarily to treat this error as a
warning, until the upstream key can be updated.
This PR should be reverted once the upstream key is updated.
With 15 runners we get 150GB of cache space to use, and we currently
have 18 jobs using ccache.
Although each run only generates ~ 200-300 MB of cache data on each run,
the small the small size may be contributing to poor hitrate.
Bump ccache to 2GB per job ~ --> 36GB of the total 150GB cache space to
try and reduce any thrashaing and generally increase hitrate.
Add a ci/lint.py script to run the linter both locally or inside the CI
(replacing .github/ci-lint-exec.py) which supports running from a
worktree.
Determines whether we are in a worktree, and mounts the real `.git`
directory as a read-only volume if we are.
The options used were wrong in two ways: firstly they were not enforced
as a "choice" (i.e. invalid input valudes could be provided without
error) and one of the options was listed as `gh` when we passed it as
`gha` from ci.yml.
"Fix" this by removing the choice altogether but sanity-testing the
input value against an expected list using a GHA "warning" to notify of
unknown inputs.
Fixes: 33735
Correct runner type selection for the lint job.
This was erroneously left-out during refactor of the runner selection
mechanism in #33302 causing the lint job to run on GH hosts (and
therefore not be able to acces local cirrus caches).
When using `docker buildx build` in conjunction with the `gha` backend
cache type, it's important to specify the URL and TOKEN needed to
authenticate.
On Cirrus runners this is working with only `ACTIONS_CACHE_URL` and
`ACTIONS_RUNTIME_TOKEN`, but this is not enough for the GitHub backend.
Fix this by exporting all `ACTIONS_*` variables.
This fixes cache restore/save on forks or where GH-hosted runners are
being used.
Co-authored-by: Max Edwards <youwontforgetthis@gmail.com>
Add an optional matrix field allowing opt-out of configuring cirrus
GHA cache when not using cirrus runners.
This is not needed for the cirruslabs/[save|restore]-cache actions, as
they automatically fallback based on runner type.
Handle multi-part netlink responses to prevent truncated results from
large routing tables.
Previously, we only made a single recv call, which led to incomplete
results when the kernel split the message into multiple responses (which
happens frequently with NLM_F_DUMP).
Also guard against a potential hanging issue where the code would
indefinitely wait for NLMSG_DONE for non-multi-part responses by
detecting the NLM_F_MULTI flag and only continue waiting when necessary.
ci/lint_run.sh: Only used in .cirrus.yml. Refer to test/lint/README.md on how to run locally.
ci/lint_run_all.sh: Only used in .cirrus.yml for stale re-runs of old pull request tasks.
Docker currently warns that we are missing a default value.
Set this to scratch which will error if an appropriate image tag is not
passed in to silence the warning.
Previously jobs were running on a large multi-core server where 10 jobs
as default made sense (or may even have been on the low side).
Using hosted runners with fixed (and lower) numbers of vCPUs we should
adapt compilation to match the number of cpus we have dynamically.
This is cross-platform compatible with macos and linux only.
When using hosted runners in combination with cached docker images,
there is the possibility that the host runner image is updated,
rendering the linux-headers package (stored in the cached docker image)
incompatible.
Fix this by doing a re-install of the headers package in
03_test_script.sh.
If the underlying runner kernel has not changed thie has no effect, but
prevents the job from failing if it has.
To remove multiple occurances of the respository name, against which we
compare `${{ github.repository }}` to check if we should use Cirrus
Runners, introduce a helper job which can check a single environment
variable and output this as an input to subsequent jobs.
Forks can maintain a trivial patch of their repo name against the
`REPO_USE_CIRRUS_RUNNERS` variable in ci.yml if they have Cirrus Runners
of their own, which will then enable cache actions and docker build
cache to use Cirrus Cache.
It's not possible to use `${{ env.USE_CIRRUS_RUNNERS }}` in the
`runs-on:` directive as the context is not supported by GitHub.
If it was, this job would no longer be necessary.
Whilst the action cirruslabs/actions/cache will automatically set this
host, the docker `gha` build cache backend will not be aware of it.
Set the value here, which will later be used in the docker build args to
enable docker build cache on the cirrus cache.
This sets the build dir at build time so that Apple SDK gets installed
in the correct/expected location for the runtime to find it.
Co-authored-by: Max Edwards <youwontforgetthis@gmail.com>
Reverts: e87429a2d0
This was added in PR #31545 with the intention that self-hosted runners
might use it to save build cache.
As we are not using hosted runners with a registry build cache, the bulk
of this commit can be reverted, simply using the value of
$DOCKER_BUILD_CACHE_ARG in the script.
link: https://github.com/bitcoin/bitcoin/pull/31545
Using buildx is required to properly load the correct driver, for use
with registry caching. Neither build, nor BUILDKIT=1 currently do this
properly.
Use of `docker buildx build` is compatible with podman.
Another action to reduce boilerplate in the main ci.yml file.
This action will set up a docker builder compatible with caching build
layers to a container registry using the `gha` build driver.
It will then configure the docker build cache args.
If set, Cirrus runners will be used on pushes to, and pull requests
against, this repository.
Forks can set this if they have their own cirrus runners.
Add "Restore" and "Save" caching actions.
These actions reduce boilerplate in the main ci.yml configuration file.
These actions are implemented so that caches will be saved on `push`
only.
When a pull request is opened it will cache hit on the caches from the
lastest push, or in the case of depends will hit on any matching depends
hash, falling back to partial matches.
Depends caches are hashed using
`$(git ls-tree HEAD depends "ci/test/$FILE_ENV" | sha256sum | cut -d' ' -f1)`
and this hash is passed in as an input to the actions. This means we
direct cache hit in cases where depends would not be re-built, otherwise
falling back to a partial match.
Previous releases cache is hashed similarly to depends, but using the
test/get_previous_releases.py file.
The cirruslabs cache action will fallback transparently to GitHub's
cache in the case that the job is not being run on a Cirrus Runner,
making these compatible with running on forks (on free GH hardware).
exFAT is known to cause corruption on macOS. See #28552.
Therefore we should warn when using this fs format for either the blocks
or data directories on macOS.
Co-authored-by: l0rinc <pap.lorinc@gmail.com>
Currently there is a warning for this in guix-build, but we also need
one in guix-codesign, otherwise the codesigned hashes are not
reproducible.
Move common functionality into prelude and call the function in both
guix actions.
Current recommendation is to add a new remote fetching all PRs, but this
is resource-intensive.
Document a better way to fetch a single PR, and to update a PR which has
been force-pushed.
Current behaviour will by-default use SOURCE_DATE_EPOCH from the
environment without warning. This breaks the default reproducibility
from a guix build.
Warn when and exit when this variable is set, and
FORCE_SOURCE_DATE_EPOCH is unset.
Closes: #11800
Note: In certain rare edge cases, monotonically increasing estimates may
not be guaranteed. Specifically, given two targets N and M, where M > N,
if a sub-estimate for target N fails to return a valid fee rate, while
target M has valid fee rate for that sub-estimate, target M may result
in a higher fee rate estimate than target N.
See: https://github.com/bitcoin/bitcoin/issues/11800#issuecomment-349697807
Filter netlink responses to only consider default routes by checking the
destination prefix length (rtm_dst_len == 0).
Previously, we selected the first route with an RTA_GATEWAY attribute,
which for IPv6 often resulted in choosing a non-default route instead of
the actual default.
This caused occasional PCP port mapping failures because a gateway for a
non-default route was selected.
...interface_usdt_utxocache.py
We have encountered an instance where the tracepoints were not collected
in the same order they were fired (#31951).
Tracepoint ordering is not guaranteed in userspace for a number of
reasons.
As this test does not require a strict collection/processing order
collect `expected` and `actual` events into dicts and compare them.
This will gracefully handle both the number of events, and out-of-order
events should they reoccur in the future.
This should be bumped every major release to avoid falling too far
behind, therefore making this feature as useful as it can be.
Document this in release-process.md to avoid forgetting to add a new
hardcoded height during release.
Fixes: #31044
This MLC update includes a change which will ignore files being ignored
by git, and help avoid false-positives when linting in this repo.
- p2p_node_network_limited -v*transport
- feature_assume_utxo
On CI runners these tests are taking longer than their current bucket
suggests, often being among the last to finish.
Re-bucket them to improve CI efficiency.
When using `sendrawtransaction` the ALREADY_IN_CHAIN error help string
may be confusing.
Rename TransactionError::ALREADY_IN_CHAIN to
TransactionError::ALREADY_IN_UTXO_SET and update the rpc help string.
Remove backwards compatibility alias as no longer required.
Updating to MLC v0.18.0 includes a new feature which will ignore all
files ignored by git: `--gitignore`.
This helps avoid false-positives flagged by this linter in non-project
files, such as a developer might expect to have in their directory (e.g.
guix-builds, python venvs, etc.)
This test checks that we can successfully process PSBTs and opt out of
finalization.
Previously trying to call `walletprocesspsbt` would attempt to
auto-finalize (as a convenience), and would not permit opt-out of
finalization, instead aborting via `CHECK_NONFATAL`.
Fix cases of calls to `FillPSBT` returning `complete=true` when it's not
the case.
This can happen when some inputs have been signed but the transaction is
subsequently modified, e.g. in the context of PayJoins.
Also fixes a related bug where a finalized hex string is attempted to be
added during `walletprocesspsbt` but a CHECK_NONFATAL causes an abort.
Reported in #30077.
PermsToSymbolicString will convert from fs::perms to string type
'rwxrwxrwx'.
InterpretPermString will convert from a user-supplied "perm string" such
as 'owner', 'group' or 'all, into appropriate fs::perms.
This adds a markdown hyperlink check task to the lint test_runner. It
relies on having the [`mlc`](https://crates.io/crates/mlc) binary found
on $PATH, but will fail with `success` if the binary is not found.
`mlc` is also added to the ci/04_install.sh script run by the
containerfile.
Note that broken markdown hyperlinks will be detected in untracked
markdown files found in a dirty working directory (including e.g.
.venv).
Without explicitly declaring the move, these UniValues get copied,
causing increased memory usage. Fix this by explicitly moving the
UniValue objects.
Used by `rest_block` and `getblock` RPC.
As described in #10542 (and numerous other places), message signing in
Bitcoin Core only supports message signing using P2PKH addresses, at
least until a new message-signing standard is agreed upon.
Therefore update the possibly-misleading error message presented to the
user in the GUI to detail more specifically the reason their message
cannot be signed, in the case that a non P2PKH address is entered.
Fixes: #809
Previously it was possible through the GUI to enter an IP address:port
into the "Proxy IP" configuration box. After the node was restarted the
errant setting would prevent the node starting back up until manually
removed from settings.json.
Previously, starting a second bitcoind using the same datadir would
correctly fail to init and shutdown. However during shutdown the PID
file belonging to the first instance would be erroneously removed by
the second process shutting down.
Fix this to only delete the PID file if we created it.
Recently usage of undeclared functions became an error rather than a
warning, in C2x. https://reviews.llvm.org/D122983?id=420290
This change has migrated into the build tools of Ubuntu 23.10 which now
causes the USDT tests to fail to compile, see
https://github.com/bitcoin/bitcoin/issues/28600
Fix this by setting `-Wno-error=implicit-function-declaration` for the
tracing programs.
Closes#21070
v21.0 introduced a behaviour changed noted in #21070 where using a config value
`rpcallowip=::0` no longer also permitted ipv4 ip addresses.
The rpc_bind.py functional test covers this new behaviour already by checking
that the list of bind addresses exactly matches what is expected so this
commit only updates the documentation.
Disable boost multi index safe mode by default when configuring with
--enable-debug.
This option can cause transactions to take a long time to be accepted
into the mempool under certain conditions; iterator destruction takes
O(n) time vs O(1) as they are stored in a singly linked list. See
27586 for more information.
Re-enable it on the CI builds which previously had it enabled.
Re-enable it on the msan fuzz target so that we have fuzz tasks testing
with it enabeld and disabled in this repo.
fixes#22638
If we find a duplicate key and error, clear `values` before returning so that
WriteSettings will write an empty file, therefore clearing it.
This aligns with GUI behaviour added in 1ee6d0b.
fixes#20246
Document both JSON-RPC endpoints, when they are active and which types
of requests they are able to service.
Adds two example curl requests, one for each endpoint.
.. only in bitcoind and bitcoin-qt
This changes behaviour of GetConfigFilePath which now always returns the
absolute path of the provided -conf argument.
Changes to the default mempool allocation size now documented.
Provides users with guidance on the mempool implications of -blocksonly
mode, along with instructions on how to re-enable old behaviour.
will