• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

openvax / vaxrank / 25856501687
96%
master: 88%

Build:
Build:
LAST BUILD BRANCH: v3.1.8
DEFAULT BRANCH: master
Ran 14 May 2026 11:07AM UTC
Jobs 0
Files 0
Run time –
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

pending completion
25856501687

push

github

web-flow
test.sh: strip to 13 lines; coverage moves to CI (#296)

* test.sh: default to serial pytest; xdist is opt-in (TEST_SH_XDIST=1)

Measurement on the current suite shows xdist is consistently slower
than serial regardless of worker count:

  serial, no cov     49s   ← default
  serial + --cov     78s
  xdist -n 1         60s   (worker fork + IPC overhead)
  xdist -n 2         95s   (slower than -n 1; disk contention)
  xdist -n 4 + cov   often 20+ min (memory pressure → swap)

Two reasons xdist loses here:

1. ``test_reference_peptide_logic`` runs ~20s on its own — first-time
   mouse kmer-set pickle load. xdist's wall-time floor IS that test;
   parallelism over the other 831 fast tests can't beat the single
   long pole.
2. Every xdist worker is a fresh Python process. Each independently
   loads pyensembl + mhctools + topiary (~3 GB resident) AND
   decompresses its own copy of the mouse kmer-set pickle. Concurrent
   loads contend on disk I/O and (with --cov on) thrash memory into
   swap.

Set TEST_SH_XDIST=1 to opt back in if a future suite has many
genuinely-parallel slow tests. The memory-aware worker math is kept
intact for that path.

* test.sh: strip to 13 lines

xdist is consistently slower than serial for this suite (measurements
in the prior commit message stand: serial 49s vs xdist -n 2 95s vs
xdist -n 4 + cov 20+ min wedge). The previous revision kept the
memory-aware xdist path as an opt-in escape hatch; this revision
deletes it. Future suite shape changes can re-introduce parallelism
later if it actually pays off.

* test.sh: drop --cov; move coverage into CI

Addresses review feedback:
- ./test.sh now runs plain pytest (~69s on this suite). Coverage adds
  ~60% wall, which is friction for the inner-loop dev path.
- CI runs pytest --cov inline (was already the path that consumed the
  .coverage file via Coveralls). DYLD guard isn't needed on Linux CI,
  so the inlined invocation skips it.
- deploy.sh is unchanged — calls ./test.sh, no ... (continued)
Source Files on build 25856501687
Detailed source file information is not available for this build.
  • Back to Repo
  • Github Actions Build #25856501687
  • d4fc97de on github
  • Prev Build on main (#25831138353)
  • Next Build on main (#25925849444)
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc