Filip Hajny <filip%joyent.com@localhost> writes: >> So therefore I probably should be patching postgresqlXX-client to drop >> the -u -g from pgxs.mk, or perhaps to make postgis not use pgxs.mk. >> > > This doesn’t normally become a problem because it’s assumed both the > client and the postgis2 package are built as the same user. E.g. our > postgresql94-client has "bsdinstall -c -o pbulk -g pbulk”. Did you > build the client package yourself on the same setup, or using a binary > package from elsewhere? > > Otherwise omitting the ownership bits sounds fine to me. Yes, I had rebuilt postgresql93-client as a different user. I don't think we document a rule that all builds must be with the same user, so I think this is a bug. But with using the same user, I don't see it, and thus I no longer think it's new with the new postgis version. Given Joerg's comment, I agree that somehow the user/group names should be passed in or just omitted. This is messy because pgsql is unusual in that it installs makefile fragments with information determined from its build environment.
Attachment:
pgp5NY6SDM1iH.pgp
Description: PGP signature