Thread: Compile Failue on win32 - pipe.c using ereport

For the past few days I have experienced compile failure with CVS HEAD 
on win32.
The culprit is src/port/pipe.c which uses ereport() when it only 
#includes "c.h".

I cured the error by #including "postgres.h" - but after a bit of 
thought suspect that "utils/elog.h" is all that is necessary!

Is anyone else able to confirm this? I am using mingw 3.1.0-1, msys 
1.0.10 and gcc 3.3.1 on win 2003 server.

regards

Mark

> For the past few days I have experienced compile failure with CVS HEAD
> on win32.
> The culprit is src/port/pipe.c which uses ereport() when it only
> #includes "c.h".

Try replacing "c.h" with "postgres.h".  I fixed that on the win32
binaries and forgot to mention it.

p.s. this probably belongs on win32 hackers list.

Merlin

"Merlin Moncure" <merlin.moncure@rcsonline.com> writes:
>> For the past few days I have experienced compile failure with CVS HEAD
>> on win32.
>> The culprit is src/port/pipe.c which uses ereport() when it only
>> #includes "c.h".

> Try replacing "c.h" with "postgres.h".  I fixed that on the win32
> binaries and forgot to mention it.

There was a recently submitted patch to change postgres.h to c.h in a
bunch of src/port modules, which I applied without looking closely.
Apparently it was a tad overenthusiastic.

The big question here is whether pipe.c is interesting for use in
frontend programs.  If so, it can't use ereport/elog at all, and we'll
need to change the code.  If not, we can just revert the inclusion
mistake.

Are there failures in any other src/port modules now?

            regards, tom lane

You are right! (goes away to sign up....)

Merlin Moncure wrote:

>
>p.s. this probably belongs on win32 hackers list.
>
>Merlin
>
>