On Tue, Jan 22, 2013 at 6:13 AM, Robbie Gemmell <robbie.gemm...@gmail.com>wrote:
> Most of the times I download and compile a release (something even us Java
> developers still do on occasion :P), the first thing that is required is
> changing into a directory, which is true with the proton-c releases.
> Changing down 2 directories doesnt sound overly taxing to me, and proton-c
> seems like a fairly obvious directory to change into if I want to compile
> the C version of proton and dont find a makefile/src dir in the initial
> version directory.
> That said, looking at the 0.3 proton-c source tar I notice that it is
> already more complex than a simple export of the proton-c subdir in the
> repo, since the contents of the two dont match directly. As there is
> already some processing going on somewhere (release.sh), it is really that
> complex to export two extra directories from the top level into the release
> tar instead of the one extra that is being done already?
The only processing that goes on now for the C tarball is to include the
tests, and this has no impact on the behaviour of the build system. Doing
this with the java api directory would be more complex as I suspect there
would need to be two different paths through the build system depending on
whether you're building from SVN or building from a release artifact.
Even this little level of processing isn't great though, and there used to
be even more than this. When we had the examples at the top level they were
also copied into the C source tarball. I eliminated this processing and
moved them in there directly after getting bug reports on the README. It
turns out the README was written/modified to reflect the svn layout and as
this was different from the release artifact layout the READMEs ended up
referencing locations that didn't exist.