Hmm, seems there's more to it than that. It looks like any attempt to build for i386 (I tried setting a few different target/sdk combinations) results in a failure to locate libiconv during the build. Telling jhbuild to skip libiconv during bootstrap doesn't fix this; I'm guessing maybe telling jhbuild to skip libiconv in all circumstances may cause different problems, but I'll have another look at this once the current "native" build finishes.
On a similar thread, I've tried building with Xcode 5, and jhbuild initially bombs out because it uses a nonexistent compiler; adding os.environ settings for CC and CXX to point to /usr/bin/cc and /usr/bin/c++ respectively gets the build going, but it throws a bunch of errors elsewhere that I've not yet looked into. I'll leave this to one side for the moment since I have a working build environment to play with on the 10.7 machine.
On 20 Sep 2013, at 14:41, John Ralls <jralls@...> wrote:
> On Sep 20, 2013, at 3:34 AM, Ronan Waide <waider@...> wrote:
>> On 09/20/2013 10:39 AM, Vassilii Khachaturov wrote:
>>> On 20.09.2013 12:29, Ronan Waide wrote:
>>>> I can put the nightly Mac builds on my own webserver on a trial basis
>>>> to see what sort of cost it incurs. I'm using Amazon Web Services, so
>>>> if it turns out to be too expensive I'll figure out some way of
>>>> restricting it. We could probably drop nightly builds into a
>>>> directory on sourceforge, mind you.
>>> Sounds great, thanks for willing to try it out!
>> As this is the first one of these I've put together, a couple of notes
>> are in order - I should include these in an extra README file, possibly
>> one that opens automatically when you mount the disk.
>> * The build is from trunk, SVN revision 22913, as indicated in the
>> filename. The build script should allow me to provide arbitrary branch
>> builds, but I haven't tested that.
>> * The build is for Intel, and may possibly be 64-bit only - I've not
>> checked this.
>> * This was built on MacOS 10.7 to avoid the problems we've seen with
>> low-level pygobject crashes using XCode4 on 10.8. I don't yet have a
>> working 10.8.5 environment to see if this bug still exists, but I know
>> that a 10.7 build will at least not crash on startup!
>> * I'm not a registered Apple developer, so the binary isn't signed,
>> which means you have to right-click and select 'Open' to launch the app,
>> unless you've configured MacOS to ignore application signatures.
>> * I've done no real testing on this other than verifying that I can run
>> it and open an existing tree (with accompanying warnings about upgrading
>> the database, etc.)
>> * I've made some minor changes to the build to allow it to build; this
>> includes updating the bundler file, and fixing the resourcepath issue. I
>> know John's aware of the latter, not sure about the former since it only
>> matters when you're trying to build the finished application. In any
>> case, I can file a bug and offer a patch.
>> I'd appreciate if any Mac users lurking on the list, developer or
>> otherwise, could give this a spin and let me know if it works for them.
>> DO NOT FEED THIS ANY DATA YOU CARE ABOUT WITHOUT BACKING UP FIRST. It's
>> a developer release, it will possibly contain bugs.
>> Oh, and as noted, if my hosting costs suddenly skyrocket I'll be
>> removing this build. In any case, since the URL is based on the SVN
>> revision it has a limited lifetime :)
> You can enable 32-bit builds with SDK 10.6 by calling
> setup_sdk(10.6, 10.6, [i386])
> from your .jhbuildrc-custom. That will allow a somewhat broader audience to use the builds.
> John Ralls
waider@... / It's about as impersonal as you can get.