Andrew Dunstan <andrew@dunslane.net> writes:
> Frankly this patch has significant infelicities. For example,
> what is the reason for removing the standard protection
> against double inclusion that header files should usually
> have from pg_config.h.win32?
I've got to admit, I don't recall that. It may be an oversight - I keep
mixing up pg_config.h.win32 and port.h and port/win32.h in my head ;-)
You will notice that the other two of those don't have it.
But it shouldn't be a problem - AFAIK it's only loaded from c.h, and
that one has protection.
> I assume that no test was done to see if this broke Cygwin,
> despite the fact that if you have a Windows box to test on,
> checking that you haven't broken Cygwin should not be too
> difficult..
Given the amount of damage I've seen it do, Cygwin is definitly *not*
making it onto my development machines. So no, I didn't build it on
cygwin. However, I didn't intend for it to break of course - I tried
manual checking. Which I obviously didn't do good enough.
I'll try to set it up in a VM for testing.
> And why do win32 and cygwin now not include at all pg_config_os.h?
It's a way to avoid the step to copy win32\port.h in msvc. configure
copies it to pg_config_os.h. Since for win32 platforms (unfortunatly, at
this point it considers cygwin win32..)that will always be port/win32.h,
it explicitly includes that one instead.
Tom writes:
> Magnus, this was your patch, can you see about fixing the
> collateral damage to the Cygwin build?
Will do. May be a ocuple of days before I can set up a VM with cygwin,
but I'll try to get it done as soon as I can.
//Magnus
