I have completed upgrading the cube codebase to use the V1 call
protocol. However, before I submit a patch I would really like to work
out why I have never been able to get cube to pass installcheck.
This is what I get when I run installcheck on a clean version of 8.1.4
(I get similar results
What version of bison do you have?
cheers
andrew
Joshua Reich wrote:
I have completed upgrading the cube codebase to use the V1 call
protocol. However, before I submit a patch I would really like to work
out why I have never been able to get cube to pass installcheck.
This is what I get
I upgraded to the latest version (2.3) as per the warning after running
configure.
Josh
Andrew Dunstan wrote:
What version of bison do you have?
cheers
andrew
Joshua Reich wrote:
I have completed upgrading the cube codebase to use the V1 call
protocol. However, before I submit a patch
Then maybe you need to remove cubeparse.[ch] and try again. Bison
changed the error message from parse error to syntax error in
version 1.875 and it has been the same ever since, AFAIK.
cheers
andrew
Joshua Reich wrote:
I upgraded to the latest version (2.3) as per the warning after
Joshua Reich [EMAIL PROTECTED] writes:
Andrew Dunstan wrote:
What version of bison do you have?
I upgraded to the latest version (2.3) as per the warning after running
configure.
Better look again, because you definitely seem to be using an old bison
to build cubeparse.c.
I just downgraded from 2.3 to 1.875 and rebuilt everything, and now it
works fine.
As per Andrew's previous email, in 2.3 the error messages were changed
from parse error to syntax error, causing the diff to fail.
Josh
Tom Lane wrote:
Joshua Reich [EMAIL PROTECTED] writes:
Andrew Dunstan
Joshua Reich [EMAIL PROTECTED] writes:
if (PG_ARGISNULL(0) || PG_ARGISNULL(1))
{
ereport(ERROR,
(errcode(ERRCODE_ARRAY_ELEMENT_ERROR),
errmsg(Cannot work with NULL arrays)));
}
This is useless code if the function is declared STRICT, as C functions
most often are. What you *do* need to be