On Mon, Jul 6, 2015 at 12:52 PM, Alvaro Herrera
<alvherre@2ndquadrant.com> wrote:
> I suggest CC'ing Peter as a first measure. I already suggested this (or
> something similar) to him months ago.
This would be a worthwhile effort. tuplesort.c only has 46% coverage.
There is no coverage for functions that I know are used all the time
in production, like dumptuples(), or ExecHashIncreaseNumBatches().
We should make increasing test coverage an explicit goal. Ideally,
there'd be a lower quality set of tests that fill in certain gaps in
code coverage, but are used less frequently, and may take much longer
to run. Simply having the code executed will allow tools like Valgrind
to catch bugs.
--
Peter Geoghegan