[Cced to hackers list] > I'm seeing a regression test failure with the latest CVS code, in the > 'conversion' test. I've attached the 'regression.diff' file -- the > failure occurs consistently on my machine. > > I'm mailing you because I believe the test in question is for code you > wrote). Let me know if you need any more information.
Do you still have the failure after doing "make install"? If so, it's news to me. If not, it's a known problem I want to fix before the beta freeze. The basic problem here is I'm using following statement while doing initdb: CREATE OR REPLACE FUNCTION utf8_to_iso8859_1 (INTEGER, INTEGER, OPAQUE, OPAQUE, INTEGER) RETURNS INTEGER AS '$libdir/utf8_and_iso8859_1', 'utf8_to_iso8859_1' LANGUAGE 'c'; The $libdir variable is defined at the compile time and it points to $prefix/lib. Apparently it points to different place while doing regression tests. One idea is replacing $lindir with the absolute path to $prefix/lib. However I wonder this would break some installations, for example RPM. Any idea? -- Tatsuo Ishii ---------------------------(end of broadcast)--------------------------- TIP 4: Don't 'kill -9' the postmaster
