Re: Setup in debian 13 (trixie): bugs + fixes

Am Sun, Aug 17, 2025 at 07:02:18PM -0300 schrieb Mar=EDa Scappini:

> */DATABASE SETUP/*
>=20
> *1- Syntax (?) error:*
>=20
> See file attached for the tail of the log after the error (bootstrap_err=
or_1.txt). Ran some
> tests, did some tinkering.
>=20
> Had to modify /var/lib/gnumed/server/bootstrap/bootstrap_gm_db_system.py=
 to fix.
>=20
> Changing line 420 to 425...
>=20
> =A0=A0=A0 curs =3D self.conn.cursor()
> =A0=A0=A0 curs.execute(u"select setting from pg_settings where name =3D =
'lc_ctype'")
> =A0=A0=A0 data =3D curs.fetchall()
> =A0=A0=A0 lc_ctype =3D data[0][0]
> =A0=A0=A0 _log.info(u'template database LC_CTYPE is [%s]', lc_ctype)
> =A0=A0=A0 lc_ctype =3D lc_ctype.lower()
>=20
> ... for this...
>=20
> =A0=A0=A0 curs =3D self.conn.cursor()
> =A0=A0=A0 # PG17+: lc_ctype/lc_collate are per-database attrs, not GUCs
> =A0=A0=A0 curs.execute(
> =A0=A0=A0=A0=A0=A0=A0 u"SELECT datcollate, datctype "
> =A0=A0=A0=A0=A0=A0=A0 u"FROM pg_database "
> =A0=A0=A0=A0=A0=A0=A0 u"WHERE datname =3D current_database()"
> =A0=A0=A0 )
> =A0=A0=A0 row =3D curs.fetchone()
> =A0=A0=A0 if not row:
> =A0=A0=A0=A0=A0=A0=A0 _log.error(u"Could not read datcollate/datctype fo=
r template DB")
> =A0=A0=A0=A0=A0=A0=A0 return None
> =A0=A0=A0 lc_collate, lc_ctype =3D row
> =A0=A0=A0 _log.info(u'template database LC_COLLATE is [%s]', lc_collate)
> =A0=A0=A0 _log.info(u'template database LC_CTYPE=A0=A0 is [%s]', lc_ctyp=
e)
> =A0=A0=A0 lc_ctype =3D lc_ctype.lower()
>=20
> ... fixed the bug, and the bootstrapper was able to continue (until the =
next error detailed below).
>=20
> Explanation (textually from the LLM, I'm still a newbie to sql databases=
): In PostgreSQL 17,
> lc_collate and lc_ctype were removed from pg_settings because they are n=
o longer GUCs. They
> are per-database attributes stored in pg_database, not cluster-wide runt=
ime parameters. So
> the query SELECT setting FROM pg_settings WHERE name=3D'lc_ctype' return=
s zero rows, data
> becomes an empty list [], and we get an IndexError which aborts the boot=
strap.

That is basically true. I have updated both the bootrapper
and gmPG2 to work on both v17- and v17+.

Best,
Karsten
=2D-=20
GPG  40BE 5B0E C98E 1713 AFA6  5BC0 3BEA AC80 7D4F C89B