Take every dependency by version, leaving no git sources - #182
Merged
Conversation
added 3 commits
August 21, 2026 03:29
Five git dependencies, four of which no longer had to be one. tauri-runtime-blitz and blitz-control-protocol are published as 0.1.0, so both sites of each take a caret. The runtime could not be published before: it set publish = false, it carried a dead [patch] on usvg, and its macos-private-api feature forwarded nothing, so --all-features left the impl carrying a WindowBuilder method the trait did not declare. That last one is why the note above the dependency here changes: this manifest was enabling the feature on both edges by hand to work around it, and the runtime now forwards it itself. worktable moves from a git rev to =1.0.0-beta.10, which is the version that rev already resolved to. Exact, not a caret, and deliberately not beta.11: master has moved to an arctic-wt this workspace cannot select, and stepping the storage engine is its own change with its own testing. The [patch.crates-io] on usvg is gone. cargo reported it as "not used in the crate graph": the published ps-anyrender-svg depends on ps-usvg, so nothing asks for plain usvg any more. agent-experimental stays. It is a third-party GitLab repository behind an off-by-default feature, so there is nothing to publish and nothing a default build resolves. The one-rev script gains the same fix chuzz's needed: grep exits 1 when it matches nothing, and under pipefail that fails the check on a lockfile with no git sources at all. Only agent-experimental keeps that from firing here today.
Two comments said it was the dependency this workspace could not publish its way out of. It is going to be published under its own crates.io account, so the empty-lockfile case the one-rev script now handles is where this is heading rather than a hypothetical.
0.1.3 is published, so the last git dependency becomes a caret like the rest. Cargo.lock now resolves every crate from the registry. The comment above it is gone rather than reworded: it existed to say why this one was different and what would change when it was published, and neither is true now. The one-rev script's note changes for the same reason, from "only this dependency keeps the empty case from happening" to the empty case being what this workspace actually is.
pathscale
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Every git dependency in this workspace is now a published version.
Cargo.lockresolves no git sources at all.
The runtime
tauri-runtime-blitzandblitz-control-protocolare published as0.1.0, soall three sites take a caret. The runtime could not be published before, and one
reason was a real bug rather than a flag:
publish = falsein its workspace manifest.[patch]onusvgpointing at a DioxusLabs branch. crates.io refuses acrate with a git dependency.
macos-private-apiwas an empty feature that forwarded nothing.tauri-runtimedeclaresWindowBuilder::transparentbehind its own feature ofthat name and the runtime implements it behind the local one, so
--all-featuresturned the local flag on without turning ontauri-runtime's,leaving the impl carrying a method the trait did not declare. crates.io
verifies with every feature enabled, so it failed to package for the same
reason it failed to check.
That last one is why the note above the dependency here changes: this manifest
was enabling the feature on both edges by hand to work around it. The
runtime forwards it itself now.
Fixed in pathscale/tauri-runtime-blitz@794b5d4.
agent-experimental
0.1.3is published, so the last git dependency becomes a caret too. Itscomment is deleted rather than reworded — it existed to explain why this one was
different and what would change once it shipped, and neither holds now.
worktable
From a git rev to
=1.0.0-beta.10, the version that rev already resolved to.Exact rather than a caret, and deliberately not
beta.11: master has movedto an
arctic-wtthis workspace cannot select, and stepping the storage engineis its own change with its own testing.
The usvg patch
Gone.
cargoreported it as "not used in the crate graph": the publishedps-anyrender-svgdepends onps-usvgdirectly. A patch that matches nothingis not inert — it is a git source in the manifest.
The one-rev script
Same fix chuzz's needed (pathscale/chuzz#26).
grepexits 1 when it matchesnothing, and under
pipefailthat failed the check on a lockfile with no gitsources — which is now this workspace's normal state, so without this the check
would go red on the very thing it wants.
Verified
cargo test -p az-gui --features blitz-runtime: 271 pass, 0 failcargo test --workspace --exclude az-gui: all passcargo check -p az-gui --features blitz-inspector: clean (the glass path)cargo check -p az-gui --features blitz-runtime,experimental: cleancargo clippy ... -- -D warnings: clean ·cargo fmt --all -- --check: cleanscripts/check-one-rev-per-git-source.sh: "no git sources"grep -c 'source = "git+' Cargo.lock→0