SRI-CSL / yices2 / 25366656949 / 1
67%
master: 67%

LAST BUILD BRANCH: incr-cadical
DEFAULT BRANCH: master
Ran
Files 486
Run time 23s
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: 66.952% (+0.3%) from 66.687%
25366656949.1

Pull #606

github

disteph 
mcsat: extend zero-copy peek to all read-only single-leaf sites (L4)

Convert the remaining tuple_blast_get call sites whose only use of the
returned snapshot is a (size==1, data[0]) read to tuple_blast_peek:
NOT, SELECT, OR/XOR, POWER_PRODUCT, ARITH_POLY, ARITH_FF_POLY, BV64_POLY,
BV_POLY, and the generic-composite combine.

Each site already followed a uniform pattern -- tuple_blast_term, then
get into a one-shot ivector, then check size==1 and read data[0] -- so
nine independent malloc+memcpy+free triples per blasted DAG node go away.
The peek pointer is consumed within the same loop iteration before any
subsequent tuple_blast_term, preserving the lifetime invariant documented
on tuple_blast_peek.

The remaining tuple_blast_get callers (APP, UPDATE, top-level substitution
in tuple_blast_apply, model reconstruction in preprocessor_build_tuple_model
and the substitution-walk loop) all need a snapshot that survives further
tuple_blast_term calls and stay on the copying path. The peek doc-comment
is updated to spell out the split.

Verification: all 31 api tests pass in debug, release, and sanitize (the
unrelated type_macros UAF preexisting on master is unchanged); all 691
mcsat regression tests pass in all three modes.
Pull Request #606: MCSAT: add support for tuples (incl. nested with function types); blasts tuples in input constraints + reconstitutes them in models and interpolants

84411 of 126076 relevant lines covered (66.95%)

1635062.56 hits per line

Source Files on job 25366656949.1