I looked into why buildfarm member hamerkop has been failing regression
tests for the last two months. The symptoms appear consistent with the
theory that the src/test/regress/data/tenk.data test file has been
converted to have DOS style line endings (\r\n). The test cases that
are failing involve reading data that should start 2030 bytes into that
file, but if you assume that the preceding \n line endings actually take
two bytes each, the data would be offset to match what we see in the
actual results. Also, the data load test that is the primary use of
that file wouldn't complain, because COPY is built to allow either \n or
\r\n line endings.
I assume this means that the git checkout was created with options that
allowed conversion of text files to \r\n line endings.
I'm not sure if we should just write this off as pilot error, or if we
should try to make the regression tests proof against such things. If
the latter, how exactly?
regards, tom lane
