Testing

Everything here assumes a build — see Building from source.

Functional tests run through corTest (installed by the corLibs umbrella into ~/git/corLibs/bin/corTest):

make test                    # whole suite (mongoc; use corTest -db corDB for in-memory)

Tests live under test/funcTests/.

Coverage

make coverage                # corDB   → coverage-corDB/index.html
make coverage DB=mongoc      # mongoc  → coverage-mongoc/index.html
make coverage-etsi           # ETSI TP suite → coverage-etsi/index.html

Measured 2026-08-27 on 227042c:

Run	Tests	Lines	Functions	Branches
DB=mongoc	620 / 620 pass	81.8%	95.2%	60.8%
DB=corDB	571 / 571 pass	76.7%	89.5%	56.3%

Branch coverage is the honest number of the three, and the one to move. Of the uncovered lines, some is a failure path no test can reach without fault injection — a database that fails on demand, a socket that dies mid-write — and 12.3% sits in 35 functions the suite never enters at all. Nearly half of those, 158 lines, are the high-availability cache-sync paths, which need a MongoDB replica set rather than a test nobody has written. coverage.md has the full breakdown, the method behind it, and why the two runs are separate measurements rather than two views of one.

The ETSI target instruments the broker and the NGSI-LD libs (whole-archived, so they flush through the broker's gcov runtime) and the mongoc/timescale plugin .sos.