Open m-span opened 1 month ago
hmm CI tests are failing in an interesting way:
thread 'main' panicked at prqlc/prqlc/src/lib.rs:124:26:
Invalid prqlc version number: c91cacc: Error("unexpected character 'c' while parsing major version number")
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
I don't know how git describe --tags
is only outputting c91cacc
, especially since that hash commit doesn't seem to exist.
Maybe there's something about how the CI is run that I don't understand? Is git
available via command line within the CI?
Nice!
Interesting re the git issue. I agree the commit doesn't seem to exist. But OTOH it does state here that that does seem to be the commit... https://github.com/PRQL/prql/actions/runs/9395042476/job/25873760775?pr=4546#step:2:56
I think it's probably creating a merge commit to simulate this branch being merged into main. And it's possibly doing a shallow clone so the tags aren't there.
Do you know if we can change the clone to be less shallow?
Either way we should make this robust to not having tags there — e.g. fall back to an empty string, I think.
Yes seems like we need something like: https://github.com/actions/checkout/issues/217#issuecomment-900923602
No obligation to use vergen, but:
cargo run -p prqlc -- --version
Finished dev [unoptimized + debuginfo] target(s) in 0.07s
Running `target/debug/prqlc --version`
prqlc 0.11.5
rerun-if-changed
approach is correct? I think vergen tries to handle that correctly (but don't know detail)I'm not sure how this works with downstream packages. What happens if I install from crates.io?
I'm not sure how this works with downstream packages. What happens if I install from crates.io?
It would need to fall back to just the version number
Adds feat. requested in #3393
example output:
On my M1 pro, it appears to add ~0.5s to the re-build process, bringing the total to somewhere between 3.6 and 5.7 seconds