Jeffrey Walton <noloader@gmail.com> writes:
> From below, I'm thinking --inputdir might not be quite correct. `mkdir
> ./testtablespace` creates testtablespace in regress/, not input/. I'd
> like to try --inputdir=./input/.
> The GNUmakefile in the top level directory and the regress/ directory
> do not include the string "--inputdir". Any ideas where I can tune it?
It really should not be necessary for you to hack the makefiles before
"make check" will pass. In any case, pg_regress seems to be reporting
that it searched the correct directory.
Given the upthread discussion about readdir having alignment issues in the
environment you're using, I'm suspecting that that is somehow causing
pg_regress to fail to find anything while it searches the directory.
Does that sound plausible at all? Look at pgfnames() in
src/common/pgfnames.c and convert_sourcefiles_in() in
src/test/regress/pg_regress.c.
regards, tom lane