Bruce Momjian wrote:
> Andrew Dunstan wrote:
>
>>> Not sure why ware are not. Should we enabled that code on Win32 and see
>>> how it works? Can you test it? Was it some MinGW limitation? I do see
>>> isatty() being used on lots of platforms.
>>>
>>> This is kind of odd. Ah, I bet it came from libpq's PQprint(), which I
>>> think we had working on Win32 long before we had psql working and
>>> perhaps I copied it from there. I don't see the Win32 checks around
>>> isatty() anywhere else.
>>>
>>>
>>>
>>>> In fact, it looks to me like it would be much more sensible to #include
>>>> "settings.h" and then simply test pset.notty for all platforms.
>>>>
>>>>
>>> Yes, we could do that but does the isatty() value ever change while psql
>>> is running? When you do '\g filename' does stdout then have isatty as
>>> false?
>>>
>>>
>> Good point. I think the best thing would just be to remove the #ifndef
>> WIN32 / #endif lines
>>
>
> OK, patch applied to remove the Win32 test in both places.
>
>
>
This broke the buildfarm and finally explains the following kluge which
has been puzzling me for four years:
/*
* for some reason MinGW (and MSVC) outputs an extra newline, so
this
* suppresses it
*/
#ifndef WIN32
fputc('\n', fout);
#endif
I have removed the kluge (and yes, I tested it).
cheers
andrew