Dave Page dpage
Thu Jul 7 16:04:45 PDT 2005
 

> -----Original Message-----
> From: Darcy Buskermolen [mailto:darcy at wavefire.com] 
> Sent: 07 July 2005 15:55
> To: slony1-general at gborg.postgresql.org
> Cc: Dave Page
> Subject: Re: [Slony1-general] [PATCH] Improved Win32 libpq detection
> 
> On Thursday 07 July 2005 03:44, Dave Page wrote:
> > The attached patch prevents the use of pg_config when 
> figuring out the
> > PostgreSQL installation paths on Win32 (because gcc doesn't 
> like the DOS
> > style paths returned). Instead, default locations for $PG_LIBDIR,
> > $PG_INCLUDEDIR etc. are assumed relative to the location of the
> > pg_config binary.
> 
> Rather than do it that way, would it make more sence to 
> convert the path 
> from / to \ where needed inside of the m4 ?  Assumptions 
> while thought to be 
> sound can often lead to looking like an ass later on.
> 
> Just my 2 ct's worth

Unfortunately it's quite that simple due to the way that Msys mounts
directories to emulate a *nix style filesystem within the windows
filesystem. For example, PostgreSQL is installed in /usr/local/pgsql in
Msys, however:

$ /usr/local/pgsql/bin/pg_config --bindir
C:/msys/1.0/local/pgsql/bin

The real path can be affected not only by the Msys installation location
(and version in the future), but also by the filesystem mount points in
use. I know of at least one major PostgreSQL hacker who keeps his Msys
in a non-standard location that would likely run into problems if we
tried to convert the paths somehow.

I'm open to ideas on how this might be done, but given that the vast
majority of Windows/PostgreSQL users run the pgInstaller distribution, I
suspect it's not worth the effort it would take - and those few users
that might compile from scratch, with non-standard paths, would probably
have no trouble overriding default locations on the configure command
line anyway.

Regards, Dave.


More information about the Slony1-general mailing list