[wan9chi]

Motivation

Before starting a long run, there's no way to see which tasks would come from the cache and which would run, or why one would miss, without running them (#791, voidzero-dev/vite-plus#2848). The dry-run RFC specifies vp run --dry-run for this. This is the first PR of the stack that implements it: the text output. The PRs above it add every change of a miss (#961), -v (#963), --json (#966, #967), unknown (#968) and -vv (#969).

This supersedes #792 by @lifeiscontent, whose write-free lookup and test scenario it builds on, so it credits them as co-author.

Closes #791.

Changes

vp run --dry-run plans the run and checks the cache as a real run does, the remote cache included, then prints each command's line and the summary. It runs nothing and records nothing. The cache is opened as in a real run, but no entry is written, a hit's outputs aren't downloaded or restored, and last-summary.json isn't written, so --last-details still shows the last real run.

$ vp run --dry-run test
$ vtt print-file src/a.txt ◉ cache hit
$ vtt print tested

---
vp run: 1/2 cache hit (50%), nothing was run.

• A dry run goes through the real run's lookup and spawn path, which skips each step that would write or run: recording a local hit, downloading a remote hit, replaying a hit, and spawning the command.

• Each line is a real run's, without , executing or , replaying, and with no blank line between commands. --log is ignored, since it only shapes task output: --dry-run and --log move from RunFlags, which selects tasks, to RunCommand, and resolve to a RunMode of Real { log } or Dry, so a dry run can't carry a log mode.

• The summary follows a real run's rules, with "nothing was run" in place of the successful count and the time saved, and without the --last-details hint. The detailed summary (-v) has no ✓ or exit codes.

  Real run, unchanged:

  vp run: 1/2 cache hit (50%), 2/2 successful. (Run `vp run --last-details` for full details)
          │ │            │     │ │
          │ │            │     │ └─ tasks
          │ │            │     └─ tasks that succeeded
          │ │            └─ cache hit rate
          │ └─ tasks
          └─ cache hits
  

  Dry run:

  vp run: 1/2 cache hit (50%), nothing was run.
          │ │            │
          │ │            └─ cache hit rate
          │ └─ tasks
          └─ cache hits
  

• A cache lookup error is printed under its command and stops the dry run, as it stops a real run, and both exit 1. Real runs stopped but exited 0 before, since the summary only counted tasks that started; a task that fails before it starts now fails the run. Otherwise a dry run exits 0 once planning succeeds.

• A nested vp run --dry-run isn't inlined. It's started as a process that prints its own dry run, so an enclosing dry run doesn't start it.

• Without a task, --dry-run is an error, like vp run -v. After the task name, it's passed to the task, like any other flag.

• RFC: the summary and cache-error rules are written down, and the sentence about the first changed file being the first in path order is removed. Without -v, the file check stops at whichever changed file it finds first.

🤖 Generated with Claude Code

---

Note

Medium Risk
Touches cache lookup, remote fetch, spawn, and exit-code paths across the runner; mistakes could skip real execution side effects or mis-report failures, though behavior is gated on RunKind::Dry.

Overview
Adds vp run --dry-run, which plans the task graph and performs full cache lookups (including remote) but does not spawn processes, replay outputs, download remote archives, write cache entries, or persist last-summary.json.

The CLI introduces RunMode/RunKind: --dry-run and --log move onto RunCommand (dry runs always use interleaved output). Execution and cache layers thread RunKind through lookup, spawn, and reporters so dry runs short-circuit after reporting cache status (WouldSpawn / DryRun not-updated reasons).

Output and summary match a real run’s command lines and cache indicators, but omit , executing / , replaying, skip extra blank lines, use “nothing was run” instead of success/saved time, and omit the --last-details hint. -v detailed summaries drop exit icons and saved-time wording for hits.

Bugfix: runs now exit 1 when a task’s cache cannot be checked (e.g. DB unreadable), including failures before start(); nested vp run --dry-run is not inlined. Changelog and dry-run RFC/e2e snapshots document the behavior.

^{Reviewed by Cursor Bugbot for commit 3010605. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

fspy benchmark

linux

dynamic/launch             change  -1.22%  [ -7.00% ..  +6.55%]  overhead   +65.22%
dynamic/access             change  +0.03%  [ -5.52% ..  +6.82%]  overhead   +19.67%
dynamic/access-relative    change  +0.52%  [ -4.23% ..  +5.60%]  overhead   +74.12%
dynamic/access-contended   change  +4.71%  [ -2.41% .. +14.14%]  overhead   +23.72%
static/launch              change  +0.41%  [ -5.20% ..  +7.94%]  overhead  +189.32%
static/access              change  +0.15%  [ -3.99% ..  +5.44%]  overhead +1431.01%
static/access-relative     change  +0.07%  [-11.09% ..  +3.49%]  overhead +1818.83%
static/access-contended    change  -1.09%  [-17.04% ..  +4.09%]  overhead +1189.53%

macos

dynamic/launch             change  -0.53%  [ -3.93% ..  +3.99%]  overhead  +237.41%
dynamic/access             change  -0.34%  [-27.82% .. +75.26%]  overhead    +4.59%
dynamic/access-relative    change  -1.15%  [-55.68% .. +15.72%]  overhead  +237.74%
dynamic/access-contended   change  +1.24%  [-17.45% .. +24.99%]  overhead    +5.28%

windows

dynamic/launch             change  +0.01%  [ -1.80% ..  +1.66%]  overhead   +25.84%
dynamic/access             change  +0.19%  [ -1.12% ..  +1.14%]  overhead    +1.33%
dynamic/access-relative    change  -0.19%  [ -0.94% ..  +0.56%]  overhead    +0.97%
dynamic/access-contended   change  +0.00%  [ -0.55% ..  +1.32%]  overhead    +1.30%

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 9f8e9ee. Configure here.}

[wan9chi]

No need for step to display cache folder.

[wan9chi]

Removed both cache-folder listings from this case: the repeated miss and --last-details already show nothing was recorded. The listing in remote_cache::dry_run_remote_hit stays, since it's what shows the hit's archive wasn't downloaded.

[wan9chi]

Correction: the listing in remote_cache::dry_run_remote_hit is gone too. The backend's log already shows the archive wasn't downloaded: a real remote hit logs GET /blob/1, and the dry run has no GET /blob request.