feat(cache): store colored task logs, strip at display when needed#378
Merged
Conversation
Spawn cached tasks with FORCE_COLOR=1 so cached output is always colored. Color-related env vars (NO_COLOR, COLORTERM, TERM, TERM_PROGRAM) are no longer passed through by default; FORCE_COLOR is pre-inserted with value "1" before pattern filtering. The reporter wraps its output writers with `anstream::StripStream` when the user's terminal doesn't support colors, so colored cache replays display correctly on plain terminals. Color support is now decided once at the CLI binary level via `supports_color::on(Stream::Stdout)` and threaded into the reporter builders as a `bool` parameter; the reporter no longer reads NO_COLOR itself. New `vtt print-color` helper and `color_env_handling` e2e fixture exercise the colour pipeline end-to-end, with a new opt-in `formatted-snapshot` mode on the e2e harness (via vt100's `Screen::contents_formatted`) for asserting ANSI bytes in cached replays. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
Replace the hand-rolled escape table in `render_formatted_screen` with a straight pass through `std::ascii::escape_default`, keeping only the newline as a special case so snapshots remain multi-line in markdown. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
branchseer
changed the title
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 92de5b89db
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
`Screen::contents_formatted` emits the full terminal state including cursor positioning (`\x1b[5;1H`), trailing `\r` and cursor-visibility sequences. The exact bytes depended on platform-specific PTY/runner behavior (Linux had `\x1b[5;1H`, macOS had `\x1b[4;1H`, Windows collapsed differently), making the new color e2e snapshot flaky. Switch `screen_contents_formatted` to `Screen::rows_formatted`, which yields per-row content with SGR attribute escapes only — no cursor moves, no screen erase. Trailing empty rows are dropped and the remaining rows are joined with `\n`. The e2e renderer then just feeds the bytes through `std::ascii::escape_default`; the hand-rolled SGR filter is gone. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
Windows ConPTY rewrites the byte stream that vt100 sees, eliding bare `\x1b[m` resets between styled spans and at the end of styled runs. Unix PTYs keep them, so the same screen state produced different formatted bytes per platform. Strip `\x1b[m` from each `rows_formatted` row in `screen_contents_formatted`. The non-reset SGR transitions (`\x1b[34m`, `\x1b[32m`, etc.) — which are the bytes the test actually asserts on — survive intact on every platform. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
Reporters previously took a single `color_support: bool` derived from
stdout. When stdout is redirected (non-TTY) but stderr is still a TTY,
this incorrectly stripped colour from stderr too.
Replace the bool with a `ColorSupport { stdout, stderr }` struct and
plumb it through every reporter builder. The reporter's own writes
(command lines, summaries) continue to use the stdout flag, and each
pipe writer in [`InterleavedLeafReporter`]/[`LabeledLeafReporter`]/
[`PlainReporter`] now picks the flag matching its stream. The grouped
reporter funnels everything into a single buffer flushed via the main
writer, so it keeps using the stdout flag.
[`Session`] detects each stream with its own
`supports_color::on(Stream::*)` cache. Also bumps the Windows-only plan
snapshot that stored `NO_COLOR: 1` (now `FORCE_COLOR: 1` after the
default untracked-env list dropped the colour vars).
https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
`ColorizeExt::style` used to short-circuit to a `Style::new()` when `NO_COLOR` was set; with stripping now handled at the writer layer the trait reduced to a pure passthrough to `OwoColorize::style`. Use `OwoColorize::style` directly at the call sites. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
Member
Author
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8ae0cec795
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
`show_last_run_details` formats the saved summary via `format_full_summary` and used to rely on the `ColorizeExt` wrapper to silently drop styles when `NO_COLOR` was set. After the wrapper was removed, the formatted buffer always carried ANSI escapes, so users with `NO_COLOR=1` or a redirected stdout would see escape codes in the saved-summary output. Route the buffer through `maybe_strip_writer(stdout_supports_color())` so the strip layer runs in the same way as the live reporter path. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
Restrict `maybe_strip_writer` to the `StdioConfig.writers` returned from `LeafExecutionReporter::start` — those receive bytes the runner does not control (child stdout/stderr, cache replay) so stripping at the writer level is the only correct place. For the reporter's own writes (command banners, error lines, summary blocks, --last-details rendering) introduce a thin `ColorizeExt::style` extension that delegates to `OwoColorize::if_supports_color(Stream::Stdout, …)`. Call sites keep the existing `.style(Style::new().bold())` syntax; the result is now a no-op on terminals that lack colour support (honouring `NO_COLOR`, `FORCE_COLOR`, and TTY detection via the `supports-color` crate). For the grouped reporter, child output is buffered before display, so its pipe writers wrap `GroupedWriter` with `maybe_strip_writer` and the main writer is stored unwrapped. Plain/Interleaved/Labeled/Summary builders likewise no longer wrap their main writer. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
Doc-link `[\`Self::start\`]` resolves against the builder type (which has no `start` method) and `[\`ColorizeExt\`]` is out of scope in `summary_reporter.rs`, so `cargo doc -D warnings` rejected the new comments. Demote both to plain code-fenced names — they still read clearly without the rustdoc cross-link. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
`[\`Self::finish\`]` on `GroupedReporterBuilder` resolves against the builder struct (which has no `finish` method), so `cargo doc -D warnings` rejected it. Replaced with prose. Re-ran `RUSTDOCFLAGS=-D warnings cargo doc --workspace --no-deps` and the workspace docs build cleanly. https://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3
branchseer
changed the title
branchseer
changed the title
branchseer
added a commit
to voidzero-dev/vite-plus
that referenced
this pull request
May 13, 2026
Bumps the vite-task git dependency from `88bacaa` to `c63db22`. ## Notable upstream changes - feat(cache): add `output` globs for cache restoration ([vite-task#375](voidzero-dev/vite-task#375)) - feat(cache): store colored task logs, strip at display when needed ([vite-task#378](voidzero-dev/vite-task#378)) - fix(plan): move FORCE_COLOR fallback after pattern filtering ([vite-task#379](voidzero-dev/vite-task#379)) - fix: preserve `PATHEXT` for Windows cached tasks ([vite-task#366](voidzero-dev/vite-task#366)) ## Vite+ side changes - Added the new `output: None` field to all `EnabledCacheConfig` initializers in `packages/cli/binding/src/cli/{handler,resolver}.rs` to match the new field added in vite-task#375. - Bumped `rusqlite` to `0.39.0` to match vite-task's new requirement and resolve a `libsqlite3-sys` links conflict. - Regenerated `packages/cli/src/run-config.ts` types snapshot (vite-task task map type tightened). - Removed the `pass-no-color-env` snap test, which is no longer relevant now that `NO_COLOR` is not in vite-task's default env passthrough. - Updated `docs/config/run.md`: - Added an `output` cache config section. - Refreshed the default-passthrough terminal env vars list (only `FORCE_COLOR` is auto-passed; other color vars need opt-in). Compare: voidzero-dev/vite-task@88bacaa...c63db22#diff-06572a96a58dc510037d5efa622f9bec8519bc1beab13c9f251e97e657a9d4ed --------- Co-authored-by: Claude <noreply@anthropic.com>
otnc
pushed a commit
to otnc/viteplus-ja
that referenced
this pull request
May 13, 2026
Bumps the vite-task git dependency from `88bacaa` to `c63db22`. ## Notable upstream changes - feat(cache): add `output` globs for cache restoration ([vite-task#375](voidzero-dev/vite-task#375)) - feat(cache): store colored task logs, strip at display when needed ([vite-task#378](voidzero-dev/vite-task#378)) - fix(plan): move FORCE_COLOR fallback after pattern filtering ([vite-task#379](voidzero-dev/vite-task#379)) - fix: preserve `PATHEXT` for Windows cached tasks ([vite-task#366](voidzero-dev/vite-task#366)) ## Vite+ side changes - Added the new `output: None` field to all `EnabledCacheConfig` initializers in `packages/cli/binding/src/cli/{handler,resolver}.rs` to match the new field added in vite-task#375. - Bumped `rusqlite` to `0.39.0` to match vite-task's new requirement and resolve a `libsqlite3-sys` links conflict. - Regenerated `packages/cli/src/run-config.ts` types snapshot (vite-task task map type tightened). - Removed the `pass-no-color-env` snap test, which is no longer relevant now that `NO_COLOR` is not in vite-task's default env passthrough. - Updated `docs/config/run.md`: - Added an `output` cache config section. - Refreshed the default-passthrough terminal env vars list (only `FORCE_COLOR` is auto-passed; other color vars need opt-in). Compare: voidzero-dev/vite-task@88bacaa...c63db22#diff-06572a96a58dc510037d5efa622f9bec8519bc1beab13c9f251e97e657a9d4ed --------- Co-authored-by: Claude <noreply@anthropic.com>
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
2 participants
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.
Fixes #358.
Summary
FORCE_COLOR=1, so the cache always stores task logs with ANSI colors intact. On a cache hit the original bytes are replayed verbatim — cached logs keep their colors regardless of the parent'sFORCE_COLORvalue.supports-colorcrate.NO_COLOR,COLORTERM,TERM, andTERM_PROGRAMare no longer inDEFAULT_UNTRACKED_ENV. Tasks that need them can opt in viaenv/untrackedEnv.Implementation notes
ColorSupport { stdout, stderr }struct threads per-stream support through the reporter builders.maybe_strip_writeris applied only to theStdioConfig.writersreturned fromLeafExecutionReporter::start(child stdout/stderr and the grouped buffer); banners and summaries use a thinColorizeExt::styleextension that delegates toOwoColorize::if_supports_color.formatted-snapshot = true(per-step) and switchedscreen_contents_formattedtovt100::Screen::rows_formatted. Raw\x1b[mresets are stripped from the rendered rows so snapshots match across Linux/macOS/Windows ConPTY.Test plan
cargo test -p vite_task -p vite_task_plan -p vite_task_graphcargo test -p vite_task_bin --test e2e_snapshotscargo clippy --testson Linuxcargo doc --workspace -D warningshttps://claude.ai/code/session_01EAhpw6TzUWExMeF7hdwuF3