Expected performance for 'Splicing Cargo workspace"

[rbtcollins]

Fetching repository @crate_index; Splicing Cargo workspace. 400s

is not uncommon for a fairly modest Bazel monorepo, which has 2 Rust projects in it (and many more other things). Is this usual, and if not, how do I get the splicer to provide a profile / opentelemetry trace / debug logging?

[UebelAndre]

In the next release, setting CARGO_BAZEL_DEBUG should enable some tracing to show you what's happening in splicing. In general though this time is spent having cargo fetch a copy of the registry and then all crates you depend on. Setting CARGO_BAZEL_ISOLATED=false should allow crate universe to use the standard cargo location on your machine to speed things up.

[rbtcollins]

Does that isolated setting cache the registry in some manner that permits incremental updates? This could explain it somewhat.

[rbtcollins]

Tested on a faster machine.
Splicing with the current config ~250s, stays about that when repeated
with CARGO_BAZEL_ISOLATED=false ~ 100s
repeating the isolated run a second time - ~20s

[erenon]

Try reducing the number of supported platforms, by specifying supported_platform_triples:

crate.from_specs(
  supported_platform_triples = [
    "x86_64-unknown-linux-gnu",
    "aarch64-apple-darwin",
  ],
)