[estebanzimanyi]

test/sql/aggregates_catalog.test runs each of the 262 aggregate signatures MobilityDuck registers
over two rows of sample values of its first argument, the other arguments fixed, and states the
answer MobilityDB gives over the same rows in the time zone UTC, at the MobilityDB commit
vcpkg_ports/meos/portfile.cmake pins (7146de0fba). Each aggregate is called by the name the MEOS
catalog states for it, its altSqlName where it has one. A value of a type whose text carries a
coordinate or a float, the 26 types both engines give an asText with a number of decimal digits,
is written with six decimals in both engines, asText(x, 6): the last digits of a coordinate an
aggregate computes, the extent and the centroid of network points among them, differ between
builds. The test also states the set of aggregate
signatures registered under the catalog's names, which is the catalog's but for the entries over
raster and geography, types MobilityDuck does not have. The point cloud values name pcid 1, whose
schema the test registers as test/sql/tpcpatch.test does.

Measured: MobilityDB answers all 262 queries and MobilityDuck answers each as MobilityDB does,
529 assertions, but wAvgAgg(tbigint), which MobilityDB answers with the bits of its values read as
a double (5e-324 for 1), a MEOS defect the test states in place of the answer.

Why: a test per aggregate stating MobilityDB's answer is what shows an aggregate MobilityDuck
registers computing what MobilityDB computes, and the registered set stated whole is what shows a
change of the catalog reaching MobilityDuck.