Repository navigation
perf(ci): DEBUG=0 for the shard build (P1a, Refs #5486) - #5745
Conversation
…1a, Refs #5486) Operator 12:35 Madrid P1a: build once, link less. Debug info dominates link time and target-dir size; nightly keeps line tables. Seven env pairs in workspace-test-shard. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
Seat: aprender-w5382 (claude-sonnet-5-5, non-author) at fec85e1. Diff read in full: ci/sections.yml only, 7 hunks (14 lines) changing CARGO_PROFILE_TEST_DEBUG and CARGO_PROFILE_DEV_DEBUG from line-tables-only to 0, in the build/test/check steps. Nothing else changed. Grepped scripts, ci, crates, .github and Makefile: no guard or test pins line-tables-only; the one remaining mention in sections.yml is a comment. Panic locations do not depend on debuginfo; only symbolized backtrace line numbers are lost. APPROVE the diff. CI not complete when read (5 pending, 1 success), so diff-only.
noahgift
left a comment
noahgift
left a comment
There was a problem hiding this comment.
seat 1, aprender-pace @fec85e142bc7e9c1d1f63ae5ee24b8c1dc8e182b
Non-author (0 aprender-pace trailers). Read the full delta: ci/sections.yml only, 7 docker run sites, CARGO_PROFILE_TEST_DEBUG and CARGO_PROFILE_DEV_DEBUG line-tables-only -> 0, symmetric, no site missed or half-changed. Compile-time lever only; no gate or test is added or removed.
Trade-off to name: debug=0 drops line tables, so a CI panic backtrace loses file:line frames (panic Location is unaffected). Fine if nobody reads CI backtraces.
NOT verified: CI (5 still in progress), the claimed build-time saving, that no test asserts on backtrace text.
|
§13.11 rung 1 — quorum shadow verdict Shadow mode: this records a verdict and merges nothing. A refusal |
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
1 participant
Operator 12:35 Madrid P1a (build once, link less): CARGO_PROFILE_DEV_DEBUG=0 and CARGO_PROFILE_TEST_DEBUG=0 in the seven env blocks of workspace-test-shard in ci/sections.yml. Nightly workflows keep line tables (untouched).
Measure (to post on merge): shard build 'Finished' time and target dir size per run vs MG 39227292002 (shard builds 6m16s/7m38s/7m31s; shard 1 7m33s). First run is cold for sccache (key change), so judge the second MG.
Gate: target dir <=25 GB/run; MG p90 <=35 min by 10-21.
Revert rule: MG p90 not lower after 2 MGs, revert.
🤖 Generated with Claude Code