daniel1noble / BACE / 25296309311
91%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 15
Run time 1min
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
coverage: 42.923%. Remained the same
25296309311

push

github

daniel1noble 
Switch globtherm + leptraits to n_cores=1 (sequential)

n_cores=2 still hit "runner has received a shutdown signal"
mid-final-imputation in run 25294935456. The pattern is consistent
across globtherm (twice) and leptraits (twice): both have
taxonomy-built trees (as.phylo + multi2di + Grafen branch lengths)
that produce a per-iteration MCMCglmm that's noticeably slower
than the real published phylogenies used by avonet/bien/pantheria.

Diagnosis: parallel R workers (parallel/parallel package) do not
stream stdout to the parent process, so the GHA runner sees no log
output during long worker fits and kills the runner on its
no-heartbeat watchdog. Datasets with faster per-iter fits (real
phylogenies) finish each iteration quickly enough that the watchdog
never fires, even at n_cores=4.

Fix: sequential execution (n_cores=1). Every per-iteration message
goes through main R's stdout, the runner sees a steady stream of
logs, the watchdog stays calm. Cost is ~2x final-imputation time
but well within the 5h45m budget given the reduced trait sets
(globtherm 5 traits, leptraits 4 traits).

This also has a side benefit: easier debugging if anything goes
wrong during final imputation -- per-iter logs are visible instead
of being captured silently inside parallel worker processes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

1574 of 3667 relevant lines covered (42.92%)

39.49 hits per line

Jobs
ID Job ID Ran Files Coverage
1 25296309311.1 15
42.92
 GitHub Action Run
Source Files on build 25296309311