That's what I hoped everyone would say. I'll put in a PR for comment. Nick On Tue, 19 Jan 2016 at 18:01, Joan Touzet <[email protected]> wrote:
> I agree we should avoid re-introducing autoconf to the build process. > > Hopefully we can fix this with a simple set of #ifdefs in a header file > somewhere that defines ssize_t (or anything else we need) appropriately, > or dig into the MS SDK for any references available and include compat > header files. > > -Joan > > ----- Original Message ----- > > From: "Alexander Shorin" <[email protected]> > > To: [email protected] > > Sent: Tuesday, January 19, 2016 12:15:25 PM > > Subject: Re: Compiling snappy under Windows > > > > On Tue, Jan 19, 2016 at 8:05 PM, Nick North <[email protected]> > > wrote: > > > I'm not exactly sure what you are proposing here. Are you saying we > > > should > > > run autoconf for the snappy code? There is an earlier commit in > > > couchdb-snappy repository that removed autotools, saying they did > > > not work > > > well with snappy, which makes me a bit cautious about that. > > > Apologies if > > > I've misunderstood your suggestion, > > > > That's the one of the solutions, but will require to: > > 1) integrate autoconf run from rebar > > 2) make autoconf as build dependency > > 3) ... > > and all these just for a few C++ types? > > > > I proposed to add preprocessor conditional to catch Windows case and > > inject the right code to make snappy works. Basically, unfold > > autoconf > > conditions manually. That's the idea. I'm not sure if this will help, > > but worth to start from there. > > > > -- > > ,,,^..^,,, > > >
