Sunbelt Computer Software

PL/B Language Development and Support

ci: diff CodSpeed base vs. head on the same runner for PRs by moreal · Pull Request #8658 · RustPython/RustPython · GitHub

ci: diff CodSpeed base vs. head on the same runner for PRs - #8658

Draft

moreal wants to merge 13 commits into

RustPython:mainfrom

moreal:codspeed-local-base-head-diff

Commits on Sep 6, 2026

ci: diff CodSpeed base vs. head on the same runner for PRs

moreal
and
claude
committed
1bc6471
fix: look for built benches under target/codspeed/analysis

moreal
and
claude
committed
de345c8
fix: resolve --out to an absolute path before measuring

moreal
and
claude
committed
c6db639
fix: run codspeed-local-diff.py from a copy outside the checkout

moreal
and
claude
committed
f88d93d
fix: instrument from process start, not via instrument-hooks' toggles

moreal
and
claude
committed
29640a5
debug: print a callgrind file preview when Ir parses to 0

moreal
and
claude
committed
a307746
fix: set the env vars instrument-hooks needs to actually instrument

moreal
and
claude
committed
e0e8ebe
fix: drive measurement through the codspeed CLI instead of bare valgrind

moreal
and
claude
committed
b6835b4
fix: total instruction counts with callgrind_annotate, not summary: lines

moreal
and
claude
committed
c0a5402
debug: include callgrind_annotate's own output when it finds no totals

moreal
and
claude
committed
e8a5a6c
fix: match callgrind_annotate's actual "PROGRAM TOTALS (calculated)" label

moreal
and
claude
committed
21bc2eb
fix: tolerate a degenerate second PROGRAM TOTALS row

moreal
and
claude
committed
b135ff5
fix: treat an all-dots PROGRAM TOTALS row as 0, not a parse failure

moreal
and
claude
committed
a22f580