PFC <lists@boutiquenumerique.com> writes:
>> You might try comparing the results of
>> select ev_action from pg_rewrite where ev_class = 'pg_user'::regclass;
>> from the two builds.
> I'll spare your eyes and tell you right away that the results are
> identical... they're at the bottom of the email.
Hmm. Your data matches mine except for several spots where words are run
together:
> :colnames ("usename""usesysid" "usecreatedb" "usesuper" "usecatupd"
^^^^^^^^^^^^^^^^^^^
> :resorigcol 5 :resjunk false} :expr {VAR :varno 3 :varattno 5:vartype 16
^^^^^^^^^
> :resdom{RESDOM :resno 7 :restype 702 :restypmod -1 :resname valuntil
^^^^^^^^^^^^^^
> :ressortgroupref 0 :resorigtbl 1260 :resorigcol 7 :resjunkfalse} :expr
^^^^^^^^^^^^^^
> :varnoold 3 :varoattno 7}} {TARGETENTRY:resdom {RESDOM :resno 8 :restype
^^^^^^^^^^^^^^^^^^^
I suppose that this is just a cut-and-paste issue but you should double
check it.
> peufeu=> SELECT pdb.datname FROM pg_database pdb, pg_user pu;
> ERREUR: unexpected right parenthesis
> Which is interesting, as there is no parenthesis in this query ;)
Yes, the complaint is about something wrong with the stored form of the
view rule for pg_user.
> I'm totally perplexed. Have you got any idea ?
I'm starting to wonder about compiler bugs breaking the code in read.c
and readfuncs.c. If so, it's entirely likely that initdb wouldn't fix
it (and more than likely would fail halfway through :-()
If you want, you can try building with --enable-debug and getting a gdb
traceback from the call to errfinish(). That would at least give us
some clue where in the rule text it's getting confused.
regards, tom lane
