Release 0.2.0 #44
Release 0.2.0 #44
Conversation
ueno
force-pushed
the
wip/release-0.2.0
branch
from
b5678e7 to
5240815
Compare
Signed-off-by: Daiki Ueno <[email protected]>
To make all crates use the same version of dependencies, this switches to using the workspace.dependencies feature. This also updates the following dependencies to make packaging easier: - page_size to 0.6 - probe version to 0.5 - toml to 0.7 - serde_with to 3 Signed-off-by: Daiki Ueno <[email protected]>
Signed-off-by: Daiki Ueno <[email protected]>
ueno
force-pushed
the
wip/release-0.2.0
branch
from
5240815 to
64ab6d9
Compare
|
@t184256 could you check this as well? |
| serde.workspace = true | ||
| serde_cbor.workspace = true | ||
| time = { workspace = true, features = ["formatting", "local-offset", "macros"] } | ||
| tokio = { workspace = true, features = ["fs", "io-util", "signal"] } | ||
| tokio-uring = { version = "0.4", optional = true } |
There was a problem hiding this comment.
any reason to single it out?
There was a problem hiding this comment.
According to the documentation, optional dependencies cannot be declared in workspace.dependencies.
| "crypto-auditing", | ||
| "event-broker", | ||
| "log-parser" | ||
| ] | ||
| resolver = "2" |
There was a problem hiding this comment.
(I know it's not changed in here, but I'd still like to ask)
Features for target-specific dependencies are not enabled if the target is not currently being built.
is either a rather confusing feature description or a rather confusing feature. My testing seems to support the idea that the enabled feature set becomes the superset of the packages' features of the invocation. That suggests that 1. cargo build result would differ from make, 2. library featureset depends on whichever application target was built first, etc.
a. Should we strive for more isolation, e.g., by building the library first in a separate invocation?
b. Given that it still doesn't guarantee we've specified the features correctly, should we, maybe, give up and define just the all-encompassing superset at the top level and at least have the consistency?
There was a problem hiding this comment.
I'm not sure if I get it right, but I can for example turn off io_uring in the agent while keeping libsystemd in the event-broker with make RELEASE=1 CARGO_ARGS="--no-default-features --features=libsystemd". Isn't that sufficient?
There was a problem hiding this comment.
(clarified off-PR. We don't have builds that go "one app + library" and library inheriting a featureset of whatever app is built first, as I've mistakenly thought after seeing four app targets in Makefile. We'll either build the entire workspace at once with the union-of-all featureset or build subprojects one-by-one with their individual featuresets. As long as we test all of the approaches we use, this is fine.)
| crypto-auditing.workspace = true | ||
| futures.workspace = true | ||
| inotify.workspace = true | ||
| libsystemd = { version = "0.7", optional = true } |
There was a problem hiding this comment.
Same, optional dependency.
agent/src/bpf/vmlinux.h
Outdated
There was a problem hiding this comment.
- can it be trimmed down?
- can it be pulled in from a separate crate, with its own repo, release cycle, versioning..?
There was a problem hiding this comment.
I'd say (1) is not practical, because functions in bpf-helpers are also defined there and have tight dependencies on other part of the header. (2) might be an interesting idea, though I would defer it to "cargo build" time, rather than "cargo package".
There was a problem hiding this comment.
In the end I managed to add a logic to generate the header file on the fly.
ueno
force-pushed
the
wip/release-0.2.0
branch
from
64ab6d9 to
b6fa8f7
Compare
This adds a logic to generate "vmlinux.h" in build.rs, needed for compiling BPF programs. If vmlinux.h is supplied by the makefile, it will be copied to the build directory; otherwise bpftool will be used to extract it from the running kernel. Signed-off-by: Daiki Ueno <[email protected]>
Signed-off-by: Daiki Ueno <[email protected]>
ueno
force-pushed
the
wip/release-0.2.0
branch
from
b6fa8f7 to
0177e4a
Compare
|
Thanks for the review! |
This prepares all changes needed to publish 0.2.0. Note that #43 needs to be merged first.