Thread: geo_ops.c is broken

The recently checked-in backend/utils/adt/geo_ops.c does not compile
under gcc 2.7.2.2, because it has

#define LINEDEBUG

followed by several instances of

#if LINEDEBUG

... in other words #if with an empty expression, which I think gcc
has every right to object to.  Either the #if's ought to be #ifdef's
or LINEDEBUG has to be #define'd as 1 rather than empty.

I changed the #if's to #ifdef's and found that it *still* doesn't
compile, because the code within the if's is seriously broken ---
it's got lines like
        result->C = pt1->x;
where result is a char * and pt1 isn't even declared.

Backing out the #define LINEDEBUG makes it compile, but who checked
this stuff in anyway?

            grumble, tom lane

> The recently checked-in backend/utils/adt/geo_ops.c does not compile
> under gcc 2.7.2.2,
> Backing out the #define LINEDEBUG makes it compile, but who checked
> this stuff in anyway?

Probably t'was moi. Will look at it, though I'm suprised that a #define
LINEDEBUG was in the code. I _always_ do that from the Makefile. I had
gotten some patches from Gautam Thaker which addressed a boundary
problem in one of the routines, and something funny must have crept in.
I'm not sure what went wrong considering that I took a (fairly) fresh
source and passed the regression tests with it...

Sorry to slow you up on testing int8 and type conversion stuff :)

                    - Tom

Looks like they are gone now.


> The recently checked-in backend/utils/adt/geo_ops.c does not compile
> under gcc 2.7.2.2, because it has
>
> #define LINEDEBUG
>
> followed by several instances of
>
> #if LINEDEBUG
>
> ... in other words #if with an empty expression, which I think gcc
> has every right to object to.  Either the #if's ought to be #ifdef's
> or LINEDEBUG has to be #define'd as 1 rather than empty.
>
> I changed the #if's to #ifdef's and found that it *still* doesn't
> compile, because the code within the if's is seriously broken ---
> it's got lines like
>         result->C = pt1->x;
> where result is a char * and pt1 isn't even declared.
>
> Backing out the #define LINEDEBUG makes it compile, but who checked
> this stuff in anyway?
>
>             grumble, tom lane
>
>


--
Bruce Momjian                          |  830 Blythe Avenue
maillist@candle.pha.pa.us              |  Drexel Hill, Pennsylvania 19026
  +  If your life is a hard drive,     |  (610) 353-9879(w)
  +  Christ can be your backup.        |  (610) 853-3000(h)