Compile PostGIS under MingW64 - windows 64-bit offering

[robe]

This we have started to work on, but realized I don't think we have a ticket in place for it.

Our starting work is here: http://trac.osgeo.org/postgis/wiki/DevWikiWinMingW64

Unfortunately we haven't gotten too far with it yet. Mostly because it takes a bit of a mindset change and also siwthcing to b\my 64-bit laptop to do it. Still shooting for PostGIS 2.0.0 to have a 64-bit windows offering.

[robe]

Side note — looks like someone might be willing to help out. We'll see.

​http://www.postgresonline.com/journal/archives/178-PostGIS-1.5.2-is-out.html#c3155

[robe]

Mark et al, We are not having much luck getting geos to compile under mingw64. Some guidance would be nice.

We have successfully compiled proj 4.7 using Paul's hack (have to revisit that though), compiled PostgreSQL 9.1beta1 (probably because Andrew Dunstan (​http://people.planetpostgresql.org/andrew/index.php?/archives/176-Windows-64-bit-work.html) put a lot of work into that. Our geos issue is outlined in ​http://trac.osgeo.org/geos/ticket/447. We are using GCC 4.4.7 as I recall FWIW.

Anyrate we'll try next trying to get the other pieces compiled and then revisit geos.

[mcayland]

Well I don't mind taking a look except for the fact that the only Windows VM I have access to is an old 32-bit XP virtual - I really don't have much call for Windows these days…

Part of me is wondering if I should start resurrect my CMake proof-of-concept from a while back. Unfortunately it would probably require an RFC because it will involve removing the C preprocessor from the build process and various other tweaks. What are people's thoughts on this? If we go CMake, we should remove autotools entirely as it makes absolutely no sense to support more than one build system. Also does CMake work under Solaris?

[robe]

Can't you use MingW-64 under Linux. The link I have on Geos ticket points to Linux binaries for it. So you would at least be able to trouble-shoot - Or is that not really baked in.

Given why everyone is saying — why are you doing this. We are beginning to wonder ourselves now. Perhaps few people care as much as we think they did.

Strk I think said he tried MingW-64 and couldn't get "hello world" to work so gave up on that. Well if it is that no one is interested in a 64-bit windows we could just give up on that dream. The main reason we want to is to compare the performance of our 64-bit Windows running true 64-bit PostgreSQL against our Linux 64-bit to see if it really does make much of a difference. One of our clients is running parallel. They have a build on 64-bit windows (with 32-bit PostGIS) and one on 64-bit CentOS. Both running more or less the same database.

Performance wise - Linux is slightly better but not by all that much. For some larger queries it is much better. That might have more to do with the ability to run higher shared memory and so forth on Linux than on windows and also that fact that windows is doing the webserver processing as well as some other things (though shutting that didn't make things better or worse).

The main issue why this is important to them is they have some processes — e.g. specialize reporting and so forth that won't work under Linux since they require windows. So its sort of a question of "Do we pay twice for hosting to have Linux DB server and Windows webserver" or can we do everything on windows.

aside from that — it's just the nagging non-paying windows users whining about how they can't use 64-bit PostgreSQL. We care much less about them of course until they put their money or time where their mouth is.

[robe]

We did this in 2.0