Repository navigation
feat(release): publish Linux .deb and .rpm alongside the AppImage - #4071
chukfinley wants to merge 3 commits into
Conversation
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Repository UI Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
ApprovabilityVerdict: Needs human review This PR introduces new Linux distribution formats (.deb and .rpm) that will be published to end users, modifying the release workflow significantly. Changes to release infrastructure and new user-facing artifacts warrant human review. You can customize Macroscope's approvability policy. Learn more. |
|
Can this get merged? |
The desktop build script already accepts `--target deb`; the target string is
passed through to electron-builder untouched and the artifact copy step is
target-agnostic. A deb build fails only because app-builder-lib's FpmTarget
requires two metadata fields the staged package.json omits: `homepage`
(control "Homepage:") and `maintainer` (control "Maintainer:", otherwise
derived from `author`, which carries no email). AppImage needs neither, so
nothing surfaced this.
Rather than add a second Linux matrix entry -- whose upload artifact name
`desktop-linux-x64` would collide with the AppImage entry's -- let the existing
Linux entry emit both targets from one electron-builder run. `--target` now
takes a comma-separated list, and artifactName's ${ext} already keeps the output
files distinct, so there is one artifact name, one runner, and no collision.
The .deb also has to be listed in the collect step and in both
action-gh-release file lists, or it builds and never ships.
Packages declare Electron's shared-library dependencies per format, so a
minimal Debian/Ubuntu install pulls them in instead of the app failing to
start. The same wiring covers `rpm`, exposed as `dist:desktop:rpm` for local
builds; the release workflow stays AppImage+deb because the runners have no
`rpmbuild`.
Consolidates the overlapping .deb work from pingdotgg#4887, pingdotgg#4900 and pingdotgg#5139: the
dependency lists and the rpm target come from @bigpod98's pingdotgg#5139, the docs and
the download-page card from @benthecarman's pingdotgg#4900.
The .deb is not an updater payload -- in-app updates on Linux stay
AppImage-only, so .deb installs upgrade through dpkg/apt. Documented in
docs/operations/release.md.
b70739b to
ce3ca8d
Compare
Dismissing prior approval to re-evaluate ce3ca8d
|
@LouisDeconinck sorry for the slow reply. It had gone stale and conflicting against
There are four overlapping PRs for this (#4071, #4887, #4900, #5139) and a consolidation discussion is running in #5139 — that's the place to push, and I've said there that I'm fine with any of them being the base as long as it's one PR. Whatever lands, the release-workflow wiring in this PR (single Linux run emitting |
Matches the existing dist:desktop:dmg / dist:desktop:win families: the bare script builds for the host arch, with explicit :arm64 and :x64 variants. Keeps the Linux packaging scripts in line with what is already there rather than pinning x64.
The rpm target needs no code beyond what the deb target already added: same FpmTarget, same homepage/maintainer metadata, and the rpm dependency list is already in the build config. Only the release wiring was missing. The Linux job now builds AppImage,deb,rpm from the one electron-builder run, and *.rpm is added to the collect step and to both action-gh-release file lists, same as the deb. electron-builder shells out to rpmbuild for this. The Ubuntu runner image ships it in the `rpm` package (4.18.2 on 24.04), so the step is a no-op there; it installs the package only when the binary is missing, which keeps the target working if the image ever drops it.
|
Note 🤖 GPT-5.6 Sol responding on behalf of Theo Closing this PR after an automated pass over open pull requests. Theo already rejected adding more Linux package formats to release CI. External package mirrors exist. |
|
if anyone wants the packages my packaged versions are avaialble at https://git.tuxbase.com/t3code-stable/-/packages and https://git.tuxbase.com/t3code-daily/-/packages i have setup repos for both deb and rpm |
The Linux release build now also makes a .deb in the same electron-builder run as the AppImage, and the release publishes it. electron-builder lists it in latest-linux.yml and writes resources/package-type into it, so the app updates a .deb install through electron-updater, which installs the new .deb with dpkg. The desktop updater now allows that path. The .deb uses xz in threaded mode (XZ_DEFAULTS=-T0): 113 MB in place of 141 MB with gzip. The Linux arm64 release job moves to a 16-vCPU Blacksmith arm64 runner. The download page, README, install guide, and release docs list the .deb. Based on community work in #4071, #5139, #4900, and #4887. Co-authored-by: chukfinley <[email protected]> Co-authored-by: Primož Ajdišek <[email protected]> Co-authored-by: benthecarman <[email protected]> Co-authored-by: NaveDanan <[email protected]> Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
* test(relay): remove constant-restating database mode test (pingdotgg#13932) * fix(server): let agents use simctl and adb alongside device tools (pingdotgg#13908) * feat(observability): honor OTEL_*_EXPORTER=none per signal (pingdotgg#13736) Signed-off-by: Yordis Prieto <[email protected]> * fix(clients): hide duplicate Cursor Keychain prompts (pingdotgg#13870) Co-authored-by: Claude Opus 5.5 <[email protected]> * fix(build): validate Linux node-pty prebuilds in Windows artifacts (pingdotgg#13867) * test(usage): assert contract mismatch details (pingdotgg#13861) * fix(usage): identify client version mismatches (pingdotgg#8208) * fix(usage): omit Cursor warning when no login is saved (pingdotgg#13820) * fix(usage): tolerate newer provider variants (pingdotgg#10076) * fix(server): background git status fetches no longer fill the disk with failed repacks (pingdotgg#13812) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(server): cut idle wakeups from the Connect relay and session reaper (pingdotgg#13774) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(clients): saving the thread list cache no longer freezes the UI (pingdotgg#13767) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(server): background sweeps only read threads that can still settle (pingdotgg#13765) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(observability): the renderer trace proxy stops tracing itself (pingdotgg#13761) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(clients): sort projects and settled threads without re-parsing dates per comparison (pingdotgg#13759) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(server): opening Diagnostics no longer loads the whole trace ring into memory (pingdotgg#13763) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(observability): stop writing empty spans on spawns, projected events, and idle polls (pingdotgg#13756) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(server): let Antigravity inspect unsupported files by path (pingdotgg#13339) Co-authored-by: Julius Marminge <[email protected]> Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * Show a focus ring on sidebar thread and draft rows (pingdotgg#13344) * fix(server): bump node-pty to 1.2.0-beta.15 for linux-arm64 prebuild (pingdotgg#13748) * fix(server): prune expired replay-protection files from the secrets directory (pingdotgg#13695) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(server): installed editors no longer vanish when discovery is slow (pingdotgg#13669) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(editors): stop treating the agy CLI as the Antigravity IDE (pingdotgg#7079) * fix(opencode): accept v2 serve ready line when spawning server (pingdotgg#13651) * fix(usage): hide the Cursor keychain prompt when Cursor isn't set up (pingdotgg#13714) * perf(server): stop re-running git for every project each minute (pingdotgg#13689) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * feat(observability): record event loop stalls in the server trace (pingdotgg#13697) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(server): per-thread settlement and PR checks no longer rebuild the whole thread list (pingdotgg#13691) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(client): slow servers finish loading the thread list instead of loading it twice (pingdotgg#13683) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(server): build the thread list snapshot without decoding it twice (pingdotgg#13693) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * feat(observability): write a server heap snapshot on SIGUSR2 (pingdotgg#13694) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(server): pull request sync reads only threads with linked pull requests (pingdotgg#13704) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * feat(cli): summarize the server trace file from the command line (pingdotgg#13698) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(server): the SQLite WAL file shrinks back after large writes (pingdotgg#13684) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(cli): t3 triage points agents at log files that exist (pingdotgg#13685) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * feat(observability): name the command on subprocess spans (pingdotgg#13701) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * perf(server): stop remapping every thread on each thread event (pingdotgg#13720) Co-authored-by: Claude <[email protected]> * refactor(observability): name each service after its application (pingdotgg#13699) Signed-off-by: Yordis Prieto <[email protected]> * fix(usage): price Cursor cache savings by base model (pingdotgg#13731) * perf(server): avoid rereading unchanged files in review previews (pingdotgg#13395) * fix(server): load Cursor keyring with createRequire (pingdotgg#13678) * fix(terminal): settling a thread closes its idle shells (pingdotgg#13673) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * feat(observability): honor the standard OTLP endpoint, headers, and protocol variables (pingdotgg#13492) Signed-off-by: Yordis Prieto <[email protected]> * fix(sqlite): retry failed statement preparations (pingdotgg#10584) Co-authored-by: shivam <[email protected]> * feat(usage): read cursor, opencode, and antigravity history (pingdotgg#10409) Co-authored-by: Yash Singh <[email protected]> * fix(usage): price Claude fast-mode requests at the fast rate (pingdotgg#13599) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(dev): one t3.json setup action that works on every OS (pingdotgg#13589) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * feat(release): ship a Linux .deb that updates itself (pingdotgg#13575) The Linux release build now also makes a .deb in the same electron-builder run as the AppImage, and the release publishes it. electron-builder lists it in latest-linux.yml and writes resources/package-type into it, so the app updates a .deb install through electron-updater, which installs the new .deb with dpkg. The desktop updater now allows that path. The .deb uses xz in threaded mode (XZ_DEFAULTS=-T0): 113 MB in place of 141 MB with gzip. The Linux arm64 release job moves to a 16-vCPU Blacksmith arm64 runner. The download page, README, install guide, and release docs list the .deb. Based on community work in pingdotgg#4071, pingdotgg#5139, pingdotgg#4900, and pingdotgg#4887. Co-authored-by: chukfinley <[email protected]> Co-authored-by: Primož Ajdišek <[email protected]> Co-authored-by: benthecarman <[email protected]> Co-authored-by: NaveDanan <[email protected]> Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]> * fix(clients): a preview app no longer knocks the desktop's own server offline (pingdotgg#13577) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(mcp): preview errors tell agents what to do instead (pingdotgg#13559) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(mcp): preview snapshots fit in the agent's tool output again (pingdotgg#13558) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(server): newer Codex models get T3 Code's instructions again (pingdotgg#13547) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(clients): sync status no longer flickers when opening running threads (pingdotgg#13551) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * ci(relay): add a forced manual relay deploy (pingdotgg#13550) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(server): report the Grok account email so usage limits merge across environments (pingdotgg#12588) Co-authored-by: Claude Opus 5 (1M context) <[email protected]> Co-authored-by: macroscopeapp[bot] <170038800+macroscopeapp[bot]@users.noreply.github.com> * fix(server): Grok accounts with no usage yet no longer vanish from Limits (pingdotgg#12799) Co-authored-by: Claude Opus 5 (1M context) <[email protected]> * fix(relay): export tunnel cleanup counters to Axiom (pingdotgg#13528) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * feat(grok): offer one-click updates through `grok update` (pingdotgg#13523) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(connect): remove tunnels after hosts go offline (pingdotgg#9386) Co-authored-by: Claude Fable 5.1 <[email protected]> Co-authored-by: Julius Marminge <[email protected]> Co-authored-by: Julius Marminge <[email protected]> * feat(codex): require Codex 0.156 and regenerate its protocol (pingdotgg#13481) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(codex): the protocol generator runs again on Effect rc.115 (pingdotgg#13480) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(antigravity): keep Windows runtime unpacking under MAX_PATH (pingdotgg#13389) Co-authored-by: javiergusart <[email protected]> Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(acp): keep one answer when a running tool reports progress (pingdotgg#13386) Co-authored-by: adeebahmad01 <[email protected]> Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(marketing): use the official OpenCode and Antigravity logos (pingdotgg#13365) Co-authored-by: Claude Opus 5.5 <[email protected]> * fix(observability): a malformed OTEL_RESOURCE_ATTRIBUTES no longer stops startup (pingdotgg#13469) Signed-off-by: Yordis Prieto <[email protected]> * feat(server): show and redeem Claude banked resets (pingdotgg#13118) Co-authored-by: Julius Marminge <[email protected]> Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> Co-authored-by: Julius Marminge <[email protected]> * chore(ci): use GPT 6 Sol Max for check agents (pingdotgg#13473) * fix(server): keep Codex's reset answer when the re-probe fails (pingdotgg#13363) Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> * fix(server): preserve racy edits in review diff previews (pingdotgg#12613) Co-authored-by: shivam <[email protected]> * feat(observability): honor the OpenTelemetry kill switch (pingdotgg#13355) Co-authored-by: Yordis Prieto <[email protected]> Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]> Signed-off-by: Yordis Prieto <[email protected]> * fix(shared): preserve final quoted empty CSV records (pingdotgg#11425) Co-authored-by: Simone <[email protected]> * fix(server): stop replaying old agent alerts on restart (pingdotgg#13340) * fix(providers): restore compatibility ranges for every harness (pingdotgg#13328) * fix(ci): shard release tests like pull request CI (pingdotgg#13321) * fix(sync): resolve wave1 test/build fallout from upstream rebase After rebase of 78 upstream commits (78 picks, 70 drops) onto c37294d, the following merge-conflict artefacts remained because the rebase strategy kept fork-exclusive V2/Pi source while picking upstream implementations of shared files: - packages/shared/src/usageMerge.test.ts: removed orphan nested it() declaration from commit 9, removed dangling expect+}) from rebase residue, and renamed staleEnvironments -> contractMismatches to match the upstream field that commit 9/11/13 introduced (the fork exclusively uses the contractMismatches field name). - apps/server/src/usage/usagePricing.ts: collapsed the doubly-nested conflict markers left by commits 61 and 75 in cacheSavingsUsd into the upstream clause that resolves to rateModel ?? record.model. - apps/server/src/provider/Drivers/CursorDriver.ts: replaced with the upstream main implementation so the file matches the upstream auth/modelDiscovery shape; the fork V2 Cursor path was already covered by PiAdapterV2.test.ts (61/61 green) and is not touched. - apps/server/src/provider/Layers/CursorProvider.test.ts: restored the fork-main version (252 lines) because the upstream version exercises a CursorProvider path that this fork does not implement (fork ships only the V2 Pi adapter and the legacy Cursor provider layer tests); 7/7 green. Wave1 critical-point verification (vp test): packages/shared/src/usageMerge.test.ts 30/30 apps/server/src/orchestration-v2/Adapters/PiAdapterV2.test.ts 61/61 apps/server/src/usage/UsageService.test.ts 14/14 apps/server/src/provider/Layers/CursorProvider.test.ts 7/7 apps/server/src/provider/cursorCredentialStore.test.ts 1/1 packages/shared/src/usageFormat.test.ts 9/9 packages/client-runtime/src/state/usage.test.ts 9/9 apps/server/src/serverLogger.test.ts 6/6 137/137 --------- Signed-off-by: Yordis Prieto <[email protected]> Co-authored-by: t3-code[bot] <269035359+t3-code[bot]@users.noreply.github.com> Co-authored-by: Julius Marminge <[email protected]> Co-authored-by: Yordis Prieto <[email protected]> Co-authored-by: Yash Singh <[email protected]> Co-authored-by: Claude Opus 5.5 <[email protected]> Co-authored-by: Tristan Knight <[email protected]> Co-authored-by: Theo Browne <[email protected]> Co-authored-by: Bilal Bakr <[email protected]> Co-authored-by: Ryan Ilano <[email protected]> Co-authored-by: Ephraim <[email protected]> Co-authored-by: Bob Fowler <[email protected]> Co-authored-by: Ishaan Kothari <[email protected]> Co-authored-by: Shirish Pothi <[email protected]> Co-authored-by: Guillermo Casanova <[email protected]> Co-authored-by: shivam <[email protected]> Co-authored-by: Aditya Garud <[email protected]> Co-authored-by: maria <[email protected]> Co-authored-by: chukfinley <[email protected]> Co-authored-by: Primož Ajdišek <[email protected]> Co-authored-by: benthecarman <[email protected]> Co-authored-by: NaveDanan <[email protected]> Co-authored-by: Jake Leventhal <[email protected]> Co-authored-by: macroscopeapp[bot] <170038800+macroscopeapp[bot]@users.noreply.github.com> Co-authored-by: Julius Marminge <[email protected]> Co-authored-by: javiergusart <[email protected]> Co-authored-by: adeebahmad01 <[email protected]> Co-authored-by: oliver <[email protected]> Co-authored-by: Simone <[email protected]> Co-authored-by: Simone <[email protected]>
Closes #298.
One open question
Maintainer:is a mandatory Debian control field, so the deb target needs an address. I used[email protected]as a placeholder — I have no way to know which mailbox you want on it, and I did not want to quietly invent one that bounces. Tell me the right address and I'll amend; it is a single string.Why this is smaller than you'd expect
--target debalready works.createBuildConfigpasses the target straight through to electron-builder, and the artifact copy step is target-agnostic. Nothing in the pipeline needs restructuring.A deb build fails today for exactly one reason:
app-builder-lib'sFpmTargetrequires two metadata fields the stagedpackage.jsonomits. AppImage needs neither, so nothing ever surfaced it:Add
homepage, addmaintainer. That is the feature.The artifact-name collision
#1655 was closed partly because "the artifact-name collision remains unresolved". That collision is a consequence of adding a second Linux matrix entry: the upload step names artifacts
desktop-${{ matrix.platform }}-${{ matrix.arch }}, so alinux/deb/x64entry producesdesktop-linux-x64— the same name the AppImage entry already uses.upload-artifactthen fails.This PR adds no matrix entry. The existing Linux entry builds both targets in one electron-builder run (
target: AppImage,deb), so there is one artifact name, one runner, and no collision by construction. The output files don't collide either —artifactName's${ext}already separates.AppImagefrom.deb.Side effect: no extra runner and no second full desktop build, so release wall-clock is unchanged.
rpm
.rpmships from the same job. It needs no code beyond what the deb target already required — sameFpmTarget, samehomepage/maintainer, and the rpm dependency list sits next to the deb one — so the Linux job buildsAppImage,deb,rpmin one electron-builder run and*.rpmjoins*.debin the collect step and both upload lists.electron-builder shells out to
rpmbuildfor this. The Ubuntu 24.04 runner image ships it in therpmpackage (4.18.2), so the added step is a no-op there; it installs the package only if the binary is missing, so the target survives an image change.Changes
scripts/build-desktop-artifact.tshomepage+maintainer; splittargeton commas so one Linux run can emit several targets; declare Electron's shared-librarydependsper package format (deb and rpm).scripts/build-desktop-artifact.test.tsdeb/rpmsections only appear for the formats actually built..github/workflows/release.ymlAppImage,deb,rpm; add*.deband*.rpmto the collect step and to both release upload lists; ensurerpmbuildis present.package.jsondist:desktop:debanddist:desktop:rpm, each with:arm64/:x64variants, matching the existingdist:desktop:dmg/:winfamilies (bare script uses host arch).docs/internals/scripts.md,docs/operations/release.md.debis not an updater payload.apps/marketing/src/pages/download.astro.debcard on the download page (amd64.debasset suffix).Note the two
softprops/action-gh-releaseblocks (first-release and subsequent-release) each carry their own file list — both needed*.deb, otherwise the package builds and never ships.Verification
Built on Ubuntu 24.04, x64, from this branch:
dpkg -cshows the executable at/opt/T3 Code*/t3codeand the.desktopentry at/usr/share/applications/t3code.desktop. Installs and launches viasudo apt install ./T3-Code-*-amd64.deb.Consolidation with the other .deb PRs
There were four overlapping PRs for this (#4071, #4887, #4900, #5139). Per the discussion in #5139 this one now folds in the parts that were missing here, so there is a single PR to review:
dependslists for deb and rpm, and therpmtarget, from @bigpod98's feat(packaging): add deb and rpm desktop targets #5139Credit to them for those pieces; happy to consolidate the other way round instead if a maintainer prefers one of the other PRs as the base — the point is one PR, not this PR.
Not included, deliberately:
/usr/bin/t3launcher. feat(packaging): add deb and rpm desktop targets #5139 has one; it is an entrypoint change rather than packaging metadata — it changes what lands on users'PATH— so it deserves its own review. @bigpod98 agreed to split it out.dist:desktop:deb:arm64,dist:desktop:rpm:arm64build locally), but shipping it needs a second matrix entry on an arm64 runner — a cost decision only a maintainer can make. Follow-up PR, happy to write it.Verification of this revision
main, no conflictsvitest run scripts/build-desktop-artifact.test.ts— 32 passedvp linton the touched scripts — clean;astro checkonapps/marketing— 0 errorsNote
Medium Risk
Changes the release artifact matrix and GitHub Release file matching (
fail_on_unmatched_files), so a failed rpm/deb build blocks shipping; packaging metadata uses a placeholder maintainer email pending confirmation.Overview
Linux desktop releases now emit AppImage,
.deb, and.rpmfrom the existing single Linux matrix job by passing a comma-separatedAppImage,deb,rpmtarget—avoiding a second runner and duplicate artifact names.The desktop packager gains
resolveLinuxTargets, stagedhomepage/ maintainer metadata required by electron-builder’s deb/rpm targets, and per-formatdependslists; CI installsrpmbuildwhen needed and attaches*.deb/*.rpmto GitHub Releases. Localdist:desktop:deb/rpmscripts, docs, tests, and a marketing.debdownload card follow the same packaging story.Linux in-app auto-update stays AppImage-only; deb/rpm installs are expected to upgrade via the system package manager.
Reviewed by Cursor Bugbot for commit 6815958. Bugbot is set up for automated code reviews on this repo. Configure here.
Note
Publish Linux .deb and .rpm packages alongside AppImage in releases
AppImage,deb,rpmin a single Linux x64 run, installsrpmbuildon the runner when needed, and uploads.deb/.rpmfiles as release assets alongside the AppImage.scripts/build-desktop-artifact.tsgainsresolveLinuxTargetsto parse comma-separated target strings, and conditionally injectsdeb.depends/rpm.depends, maintainer, homepage, and other metadata into the Tauri build config.dist:desktop:debanddist:desktop:rpmnpm scripts (includingarm64/x64variants) allow local builds of each format..debdownload card for Debian/Ubuntu (x86_64)..deband.rpmare shipped in the release but are not updater payloads.Macroscope summarized 6815958.