På tirsdag 30. april 2019 kl. 15:43:16, skrev Tom Lane <t...@sss.pgh.pa.us <mailto:t...@sss.pgh.pa.us>>: Andreas Joseph Krogh <andr...@visena.com> writes: > With master (3dbb317d32f4f084ef7badaed8ef36f5c9b85fe6) I'm getting this: > visena=# CREATE INDEX origo_email_part_hdrvl_value_idx ON > public.origo_email_part_headervalue USING btree > (lower(substr((header_value)::text, 0, 1000)) varchar_pattern_ops); > psql: ERROR: failed to add item to the index page
Hm, your example works for me on HEAD. Usually, the first thing to suspect when you're tracking HEAD and get bizarre failures is that you have a messed-up build. Before spending any time diagnosing more carefully, do "make distclean", reconfigure, rebuild, reinstall, then see if problem is still there. (In theory, you can avoid this sort of failure with appropriate use of --enable-depend, but personally I don't trust that too much. I find that with ccache + autoconf cache + parallel build, rebuilding completely is fast enough that it's something I just do routinely after any git pull. I'd rather use up my remaining brain cells on other kinds of problems...) regards, tom lane I built with this: make distclean && ./configure --prefix=$HOME/programs/postgresql-master --with-openssl --with-llvm && make -j 8 install-world-contrib-recurse install-world-doc-recurse It's probably caused by the data: visena=# select count(*) from origo_email_part_headervalue; count ---------- 14609516 (1 row) I'll see if I can create a self contained example. -- Andreas Joseph Krogh
