some wrong with 'contains_nonconstant_walker' about gpdb HOT 2 CLOSED

On 20/01/16 12:35, liruto wrote:

Supposed the first argument is given a VAR node,
'contains_nonconstant_walker' would return FALSE It's obviously wrong
that VAR doesn't contain anything non-constant,isn't it?

Hmm, I think you're right. The saving grace is that the function is used
only in one place, to test if an INSERT .. VALUES() contains anything
but constants. I believe there can't be any Vars in the list in that
context. It would still be prudent to check, IMO.

Also, why does the walker call contain_volatile_functions() on every
node it traverses? That's a waste of time, as
contain_volatile_functions() recurses; it'd be enough to call it once at
the root node.

Poking around that case, I bumped up into this:

create table pt (i int, j int, k int) distributed by (i) partition by
range(k)
(start(1) end(10) every(2));
alter table pt_1_prt_2 set distributed randomly;
insert into pt values (1, 1, random());

psql:/tmp/pt-const-bug.sql:4: FATAL: Unexpected internal error
(cdbmutate.c:564)
DETAIL: FailedAssertion("!(((((Node*)(tle->expr))->type) == T_Const))",
File: "cdbmutate.c", Line: 564)
HINT: Process 20240 will wait for gp_debug_linger=120 seconds before
termination.
Note that its locks and other resources will not be released until then.
server closed the connection unexpectedly
This probably means the server terminated abnormally
before or while processing the request.

I'm not sure what, but there's something wrong there..

• Heikki

from gpdb.

Fixed, by PR #2972. Thanks for the report!

from gpdb.