test(desktop): keep update tests independent of the workspace version

[leoisadev1]

Summary

The 0.2.1 version PR (#76) fails CI because six desktop update tests served 0.2.0 as the newer build. The app compares manifest builds against its own crate version, so once Changesets bumped the workspace to 0.2.1, 0.2.0 became a downgrade and VerifiedManifest::select filtered it out. Newer test builds now use 9.2.0 and 9.3.0. Only test code changes.

 running build = crate VERSION (0.2.0 on main, 0.2.1 on the version PR), dated 2026-10-01
 manifest builds:
   ("0.1.0", "2026-10-01")   # older than any real version; unchanged
-  ("0.2.0", "2026-10-03")   # newer only while the app is <= 0.2.0
-  ("0.3.0", "2026-10-04")
+  ("9.2.0", "2026-10-03")   # newer than the app at any realistic version
+  ("9.3.0", "2026-10-04")

Evidence

• Before: CI run 37678677744 on chore(release): version packages #76 (workspace at 0.2.1), Rust (Linux, desktop) and Rust (Windows, desktop):

  ---- ui::tests::a_covered_update_shows_and_opens_the_download_page stdout ----
  assertion `left == right` failed
    left: NotCovered { version: "0.3.0", date: "2026-10-04", purchase_url: "https://convt.test/pricing" }
   right: Available { version: "0.2.0", date: "2026-10-03", uncovered: Some("0.3.0") }
  test result: FAILED. 99 passed; 6 failed
  

• After: CI on this PR runs cargo test -p convt-app with the workspace at 0.2.0. Once this merges, Version Packages regenerates chore(release): version packages #76 and runs the same tests at 0.2.1.

Not run locally: the box doesn't have the pinned 1.95 toolchain or the gpui system libraries, so CI is the proof for both versions.

Merge Danger

Door: two-way. Test-only change, so a revert undoes it completely.

Blast radius: convt-app update tests only. The app's update selection and release workflows don't change.

Fixes CNV-58

Created with Grok Bot in Grok Bot.

---

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.

[greptile-apps]

Confidence Score: 4/5

[Low risk] Test file updates version numbers in update checks.

The PR appears safe to merge, with a non-blocking suggestion to preserve coverage for same-version rebuilds.

Findings

1.  Same-version updates lose coverage ▶

Summary

This PR changes desktop update test builds to 9.2.0 and 9.3.0, updates their expected text, and explains why the fixtures use higher versions.

• The covered-update and renewal expectations remain consistent.
• One non-blocking test gap remains: same-version builds with later dates lose coverage.
• Tests were not run during this review.

_{Reviews (1) · Last reviewed commit: "test(desktop): keep update tests indepen..." · Reviewed by Greptile}

[greptile-apps]

Same-version updates lose coverage

Changing the covered build to 9.2.0 removes coverage for a build with the same version as the running app but a later date. On the current workspace version, the old 0.2.0 fixture exercised that case. The existing convt-update tests do not cover it separately, so a change that rejects these builds could pass unnoticed.

Keep these workspace-independent fixtures, but add a dedicated VerifiedManifest::select test with a fixed running version and a same-version, later-date build. This is non-blocking. Confidence: 5/5.