Compile error in current cvs (~1230 CDT July 4)

Bruno Wolff III wrote:

I get the following compilation error when making postgres from current CVS:
timestamp.c: In function `tm2timestamp':
timestamp.c:69: warning: implicit declaration of function `elog'
timestamp.c:69: `ERROR' undeclared (first use in this function)
timestamp.c:69: (Each undeclared identifier is reported only once
timestamp.c:69: for each function it appears in.)
make[4]: *** [timestamp.o] Error 1

This is on a machine with RH 6.1.

The following configure command was used:

./configure --prefix=/usr/local/pgsql --enable-integer-datetimes --with-pgport=5433

Yeah, I've been getting that since Wednesday morning (west coast USA
time), and reported it Wednesday evening, but no one else has replied to
that post, so I thought maybe it was somehow related to the othee ecpg
issues being discussed.

I sync'd up after Tom committed the lost commits from Wednesday, and I'm
still seeing the issue. Adding

+ #include "utils/elog.h"

to timestamp.c lets me compile, but I'm left with two warnings:

timestamp.c: In function `PGTYPEStimestamp_from_asc':
timestamp.c:315: warning: overflow in implicit constant conversion
timestamp.c:319: warning: overflow in implicit constant conversion

Joe

Joe Conway <mail@joeconway.com> writes:

Yeah, I've been getting that since Wednesday morning (west coast USA
time), and reported it Wednesday evening, but no one else has replied to
that post, so I thought maybe it was somehow related to the othee ecpg
issues being discussed.

I've committed fixes for the problems noted by gcc. I wouldn't care to
bet that the code actually works though. The HAVE_INT64_TIMESTAMP paths
in ecpg seem to be totally untested :-(

regards, tom lane

Tom Lane wrote:

Joe Conway <mail@joeconway.com> writes:

Yeah, I've been getting that since Wednesday morning (west coast USA
time), and reported it Wednesday evening, but no one else has replied to
that post, so I thought maybe it was somehow related to the othee ecpg
issues being discussed.

I've committed fixes for the problems noted by gcc. I wouldn't care to
bet that the code actually works though. The HAVE_INT64_TIMESTAMP paths
in ecpg seem to be totally untested :-(

Thanks, Tom. It does at least compile cleanly now. I don't use ecpg, so
I can't say whether the changes actually work.

Joe

On Fri, Jul 04, 2003 at 11:23:57 -0700,
Joe Conway <mail@joeconway.com> wrote:

Thanks, Tom. It does at least compile cleanly now. I don't use ecpg, so
I can't say whether the changes actually work.

I am still seeing the problem in anoncvs, but I seem to remember there being
a lag between the real cvs and the anoncvs servers, so I might just need
to wait a bit.

I don't use ecpg either, so I won't be testing out the code there.