DrudgeCAS / gristmill / 31896274029
92%

DEFAULT BRANCH: master
Ran
Jobs 1
Files 4
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: 91.249% (+0.05%) from 91.198%
31896274029

push

github

web-flow 
deps: Bump fbitset and libparenth, and cover the crash they fixed (#36)

libparenth applied its unnatural-partition skip to every problem, though
it is only valid for classical contractions where each summation is
involved by exactly two factors.  Outside that it could skip every
candidate, leaving a subproblem with no evaluation, and then read the
first element of an empty vector.  From Python that is a segfault of the
interpreter, with no exception and no traceback.

    targets = [dr.define_einst(s, x[a] * y[a] * z[a] * w[a])]
    optimize(targets, contr_strat=ContrStrat.OPT)      # segmentation fault
    optimize(targets, contr_strat=ContrStrat.GREEDY)   # segmentation fault

It also returned suboptimal parenthesizations on non-classical problems,
for 68 of 500 random problems, where ContrStrat.OPT is documented to find
the global minimum.

fbitset separately had its iterator read one limb past the end of the
storage, on every complete walk over a bit set, and could not be included
from two translation units.

Add a test for the four-factor summation, since the crash was only ever
reachable from Python and nothing here would have caught it coming back.
On the previous submodules it takes the whole pytest process down; on
these it passes, with every strategy agreeing on the cost.  It is stable
across hash seeds.

Test suite: 31 passed, 2 xfailed.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

1783 of 1954 relevant lines covered (91.25%)

0.91 hits per line

Jobs
ID Job ID Ran Files Coverage
1 31896274029.1 4
91.25
 GitHub Action Run
Source Files on build 31896274029