- Add the ArtifactNode port over radicle-artifact-client
- Seed, unseed, download and fetch artifacts per release
- Unseed artifacts on redaction and release deletion
- Report artifact node status and installed binaries
- Probe the artifact node alongside the radicle node
- Add node commands and mirror them in test API
Co-authored-by: Brandon Oxendine <brandon.oxendine@gmail.com>
Co-authored-by: Rūdolfs Ošiņš <rudolfs@osins.org>
- Add release types and the Releases and ReleasesMut ports
- Create, delete and redact releases
- Register, attest and locate artifacts
- Filter releases with radicle-artifact
- Add release commands and mirror them in test API
Co-authored-by: Brandon Oxendine <brandon.oxendine@gmail.com>
Co-authored-by: Rūdolfs Ošiņš <rudolfs@osins.org>
Relative images in markdown and AsciiDoc files were broken, nothing
resolved them against the repository. They're read through a new
`repo_blob_bytes` command, which skips the text encoding and last
commit lookup of `repo_blob`. Images now also keep their alt text.
Co-authored-by: Rūdolfs Ošiņš <rudolfs@osins.org>
Uses @asciidoctor/core, lazily imported so the main bundle is
unchanged, with its own DOMPurify instance because the markdown
allowlist is pinned globally via setConfig.
Include directives resolve through a registered HttpCache that maps
Asciidoctor's URI reads onto repo_blob, so tag, line-range and
nesting handling stay upstream. A per-conversion token in base_dir
scopes each read to one repository and commit.
Embeds went over IPC as JSON number arrays, so a 9MB image turned into
~36MB of JSON. On upload Tauri parses that on the main thread, which
beachballed the app while dropping a large screenshot.
Pasting a single image went through save_embed_by_clipboard, which
stored the clipboard's raw RGBA pixels instead of an encoded image.
The result could not be decoded, so the embed showed as a broken image.
Always store the pasted file's own bytes instead. Nothing else uses
save_embed_by_clipboard, so it's removed.
Co-Authored-By: Rūdolfs Ošiņš <rudolfs@osins.org>
Seeding a repo only fetched from seeds we were already connected to, so
one nobody had announced to us yet stayed in the fetching state. Also
try the preferred seeds and connect to known ones, like `rad seed`.
When looking at merged patches you want to see what landed recently, not
what was opened recently. We take the earliest merge, which is exact for
repos that need a single delegate to merge.
References are picked or pasted in the composer, stored as `rad:` URIs
with a readable label, and shown as chips or links that open in-app when
the entity is here.
Co-Authored-By: Rūdolfs Ošiņš <rudolfs@osins.org>
Mentioning someone starts from their alias, but aliases are
self-declared and unrelated nodes often share one. Results say who is
followed or is you, so the composer can tell them apart.
Co-Authored-By: Rūdolfs Ošiņš <rudolfs@osins.org>
Seeding only set the policy, so a repo stayed unreplicated until a
seed happened to announce it again. Fetch from connected seeds in the
background, like `rad seed` does.
Rebuild the issue page as a single column matching the patch page:
the state dropdown sits next to the title, with author, issue ID,
labels and assignees in a row beneath it.
Let issue authors change the state, not only delegates. Add a
delete_issue command so authors can delete their own issues.
Share the delete confirmation and ID chip with the patch page.
Adds a read-only page for a repository's identity document. It shows the
current document together with its full revision history. The data and
wording follow heartwood and the rad id / rad inspect CLI.
Co-Authored-By: Rūdolfs Ošiņš <rudolfs@osins.org>
A repository can sit in storage without being seeded, which `rad ls`
hides and `rad ls --all` shows as "local". A new Settings preference
picks the `Show` passed to `list_repos_summary`, defaulting to seeded.
Unseeding only deletes the policy row, so a node defaulting to `allow`
keeps seeding: `unseed` now reports `DefaultPolicySeeds` instead of
claiming success, and both callers refresh in a `finally` so a failed
`clean` no longer leaves a stale row behind.
A desktop launch inherits a `PATH` that holds none of the user's own
install locations, so `Command::new("git")` finds nothing usable, or on
macOS finds Apple's shim and only prompts to install the developer
tools.
Resolve git to an absolute path instead, searching `PATH` first and the
usual install locations after, and cache the result for the process.
`rad` is looked for the same way, which also fixes the CLI check for a
Homebrew or Nix install.
`rad unseed` only drops the seeding policy and leaves the repository in
storage, so while `list_repos_summary` listed everything on disk an
unseeded repository never left the sidebar and unseeding looked like it
had done nothing. That call now filters by seeding policy.
The confirmation dialog also offers to delete the files, running the
equivalent of `rad clean` after the unseed.
Co-authored-by: Rūdolfs Ošiņš <rudolfs@osins.org>
`list_commits` walked back from the revision head and stopped at the
first commit equal to `base`. A merge commit reaches `base` through one
of its parents, so the walk truncated there and dropped everything the
other parent contributed: a revision whose head merges the target branch
in looked like a single-commit revision.
Serve `targetBranch` so the UI can tell where a patch is headed. Editing
a title no longer sends its own target, which reset branch-targeted
patches to the default branch.
The tree-sitter highlighter goes with it: the hunk types name its
`Paint` spans and it implements `ToPretty` for them, so neither side
can be removed on its own.
`revision_by_patch_and_id` and `review_by_patch_and_revision_and_id`
were registered as Tauri commands but never invoked: the patch view
resolves a revision or review from the revisions it already holds. Drop
the commands, their registrations and the `revision_by_id` / `review_by_id`
port methods, which had no other callers.
`fileDiffName` in `diffText` has had no callers since it was added.
radicle-surf 0.29 switched from radicle-git-ext to radicle-oid for
its Oid type, making radicle_surf::Oid and radicle::git::Oid the
same type. The crate::oid::{from_surf, into_surf} helpers were
therefore no-op round-trips through git2::Oid.
explorerUrl() hardcoded both the seed ("iris.radicle.network") and
the explorer base ("https://radicle.network"), so every "Copy Link"
and "Open in …" action ignored the user's config and could point at
a host that does not replicate the repo.
Expose publicExplorer and preferredSeeds on the Config type (both
already carried by radicle::profile::Config) and thread config
through to every explorerUrl() call site. URLs are now built from
the publicExplorer template, filling $host from the first preferred
seed (falling back to rosa only when none are configured), and link
labels show the configured explorer's host.
Co-authored-by: Rūdolfs Ošiņš <rudolfs@osins.org>
PatchTeaser requests diff_stats per row, and each call re-opened the
whole repo through libgit2-surf, so a list of patches serialized into
tens of seconds on a large repo. Shell out to `git diff --numstat`,
which opens and diffs in one fast pass, falling back to the surf diff
when git is unavailable.
repo_blob and repo_readme resolved the blob through radicle-surf, which
also walks history to find the last commit that touched the path.
libgit2 ignores the commit-graph and diffs every commit, so opening a
file last changed long ago took a long time and varied by how far back
the file changed, not by its size.
Resolve the content with a direct tree lookup and the last commit via
`git rev-list -1`, falling back to the libgit2 walk when git is
unavailable.
libgit2's revwalk ignores the pack bitmap and commit-graph, so counting
walks every commit. Shell out to `git rev-list --count --use-bitmap-index`,
which uses those indexes, and fall back to the revwalk if git is
missing, too old, or errors, so the result is unchanged, just faster on
maintained storage.
Navigating away from a list and back via history (the sidebar
back/forward buttons) returns to the same scroll position with the
previously loaded pages intact, instead of refetching only the first
page at the top.
list_issues previously read through heartwood's id-ordered issue cache,
forcing it to load and sort every issue in memory on each page. Mirror
the patch path instead: a domain/issue Service backed by an IssueStorage
adapter that queries the issues table directly.
The issues list shipped each issue's full root comment (body) over IPC,
but the teaser never renders it, only the detail view does. Make
Issue.body optional and add an Issue::summary constructor that skips it;
list_issues now builds summaries while issue_by_id keeps the full body.
Daniel Norman