Simplify source pages sidebar handling - #119016
Closed
GuillaumeGomez wants to merge 2 commits into
Closed
Conversation
rustbot
added
the
S-waiting-on-review
Collaborator
|
Some changes occurred in HTML/CSS/JS. cc @GuillaumeGomez, @jsha |
rustbot
added
the
T-rustdoc
GuillaumeGomez
commented
Dec 16, 2023
| #src-sidebar div.files > a.selected { | ||
| background-color: var(--src-sidebar-background-selected); | ||
| } | ||
| #src-sidebar-toggle > button { |
Member
Author
There was a problem hiding this comment.
I moved the block just under #src-sidebar-toggle rule as it made more sense like this.
This comment has been minimized.
This comment has been minimized.
GuillaumeGomez
force-pushed
the
simplify-src-sidebar-js
branch
from
December 16, 2023 16:41
c310228 to
ac97299
Compare
notriddle
reviewed
Dec 16, 2023
| content: ">"; | ||
| } | ||
| .src-sidebar-expanded #src-sidebar-toggle > button::before { | ||
| content: "<"; |
Contributor
There was a problem hiding this comment.
Didn’t we reduce using generated content for accessibility reasons?
Member
Author
There was a problem hiding this comment.
We absolutely did. The other possibility is to have two buttons. I'll go with this option then.
Contributor
There was a problem hiding this comment.
Could it be switched to <details>?
Member
Author
There was a problem hiding this comment.
Yes but I thought for accessibility reasons we didn't want to do that haha.
GuillaumeGomez
force-pushed
the
simplify-src-sidebar-js
branch
from
December 16, 2023 18:05
ac97299 to
e9faeea
Compare
Member
Author
|
Updated the JS by having two buttons instead of just one. |
Collaborator
|
☔ The latest upstream changes (presumably #119066) made this pull request unmergeable. Please resolve the merge conflicts. |
BlickandMorty
added a commit
to BlickandMorty/Epistemos
that referenced
this pull request
Apr 26, 2026
…with Apple ld64 trap
Per docs/audits/EXTENDED_PROGRAM_PLAN_2026_04_25.md Wave 6
Cross-ref dpp §6.1-6.3 Sprint 5 + 6
Lands the canonical scaffolding for Wave 6's two production deliverables
per the Wave 6 research findings:
1. PGO via cargo-pgo with the Apple ld64 lto="thin" workaround
2. bumpalo per-frame arenas with the serde_json::RawValue copy trick
The actual MIGRATION work (rewriting MCP dispatch + render hot paths
to allocate from `with_frame`, running the full PGO sweep against the
Swift host) is follow-up — both require differential testing against
the existing allocator paths and a multi-day instrumentation cycle.
NEW agent_core/src/arenas/ module:
- mod.rs — module doc + re-exports
- frame.rs — thread_local Bump per worker thread
(canonical research finding: "one arena per
worker, NOT one global; Bump is !Sync by
design — sharing requires a Mutex that
destroys the win"). Exposes `with_frame(|arena| ...)`
which resets the bump pointer (O(1)) before
invoking the closure.
- raw_value.rs — `raw_value_in(&RawValue, &Bump) -> BumpString`
helper for the `serde_json` heap-string gotcha.
serde_json::from_str ALWAYS allocates owned
Strings on the heap regardless of arena;
the canonical workaround per the research is
to read once into &RawValue, then copy the
slice into the arena via BumpString::push_str.
Used by the MCP dispatch path migration.
Initial chunk capacity: 4 KB per thread (sized for the median MCP tool
call: dispatcher header + small payload). The arena grows on demand if
a single call exceeds this; bumpalo's `reset()` keeps the LARGEST chunk
across resets so steady-state has zero chunk-allocation churn.
NEW agent_core/Cargo.toml dependencies:
- bumpalo 3.16 with `collections` feature (the canonical 2026 choice
per research: typed-arena is single-type only; slab is for stable
indices, wrong tool; no serious challenger has emerged)
- serde_json `raw_value` feature enabled
NEW [profile.release-pgo] in agent_core/Cargo.toml:
inherits = "release"
lto = "thin" ← critical Apple ld64 trap
Apple's ld64 LTO pass discards `__llvm_prf_*` PGO instrumentation
sections silently when lto = "fat". Verified against rust-lang/rust#119016.
Thin LTO retains them. The canonical [profile.release] keeps lto = "fat"
for non-PGO release artifacts; this separate profile is used only
during the cargo-pgo cycle.
NEW scripts/pgo-cycle.sh:
Drives one full cargo-pgo cycle for agent_core (the largest dylib at
14 MB — biggest PGO leverage). Three subcommands: `instrument`,
`optimize`, `all`. Embeds the Apple ld64 trap warning + reference to
rust-lang/rust#119016 inline so a future maintainer reading the
script understands why lto="thin" is non-negotiable here.
Designed for manual PGO sweeps (NOT CI — instrumentation cycle takes
~20 minutes wall-clock). Documents that the bench alone is not enough
coverage; the operator must drive the Swift host through real
workloads (idle scroll, agent turn, graph layout) for 5+ minutes per
the Wave 6 research finding ("3 distinct workloads totaling 5+ minutes;
fewer over-fits to the dominant path and regresses cold paths").
8 new tests in agent_core arenas module:
frame.rs (5):
- frame_arena_resets_between_calls (address-equality verifies
bump-pointer rewind across consecutive with_frame calls)
- frame_arena_strings_can_be_built_inside_closure
- frame_arena_handles_large_then_small_alloc_cycle (chunk reuse)
- nested_with_frame_panics (leaf-frame contract enforcement)
- separate_threads_have_independent_arenas (thread_local correctness)
raw_value.rs (3):
- raw_value_in_copies_bytes_into_arena
- raw_value_in_handles_unicode
- nested_raw_values_keep_their_serialised_form
6 new Swift source-guard tests in EpistemosTests/PGOAndArenasTests.swift:
1. arenas/ module shape (mod.rs + frame.rs + raw_value.rs all exist)
2. with_frame uses thread_local + arena.reset() per the research
3. raw_value_in helper exists with BumpString
4. agent_core/Cargo.toml declares bumpalo + serde_json raw_value
5. [profile.release-pgo] uses lto = "thin" (Apple ld64 trap guard)
6. pgo-cycle.sh exists, is executable, embeds the ld64/rust#119016 warning
Verification:
- cargo test --release on agent_core arenas:: → 8/8 pass
- cargo build --release --target aarch64-apple-darwin succeeds
- All 6 PGOAndArenasTests pass via xcodebuild
Wave 6 follow-ups (per dpp §6.2-6.3, NOT in this commit):
- Rewrite omega-mcp dispatcher to allocate scratch from with_frame
(uses raw_value_in for the JSON-RPC body parse)
- Bumpalo arena adoption in the Metal render driver path
- Full PGO sweep + ≥5% wall-clock measurement against the dpp budget
- xcrun xctrace report per dpp §8 acceptance criteria
Co-Authored-By: Claude Opus 4.7 (1M context) <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
5 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.
Since the text isn't really important in itself, instead of updating it manually in the JS, we can let the CSS do it instead.
r? @notriddle