[mythtv] [RFC] New Win32 build script

[Lawrence Rust]

On Fri, 2010-12-10 at 17:58 -0800, Robert McNamara wrote:
> On Fri, Dec 10, 2010 at 12:33 PM, Lawrence Rust <lvr at softsystem.co.uk> wrote:
> >
> > I see the problem now.  It's that your mingw prefix is something other
> > than i586-mingw32msvc.  The mkspecs/qmake.conf that's installed by the
> > patch is hard coded to i586-mingw32msvc
> >
> > This needs some thought, probably some sort of macro subsitution in the
> > patch.  I'll keep you posted.
> >
> > -- Lawrence
> 
> Hi Lawrence-- Just taking the script for a test run-- FYI using -j > 1
> can potentially end up causing problems when building the dependencies
> (as it just did for me) that are non-obvious and look like code issues
> (but are not).  In my case, lame compilation failed, and I had to make
> distclean, then manually rm the stamp, then restart the build process
> with a single build thread to continue fine-- there was never any
> actual code issue, just that the compiler got ahead of itself.

Thanks for the testing and report.  What does `make --version` report?
Maybe there's a workaround for specific versions.

To force a single threaded build:
NUMBER_OF_PROCESSORS=1 ./mythbuildw32

Re. ./configure and distclean.  I've noticed this to be a pain so I
added a test in the script for the shell variable [PACKAGE]_CFG e.g.
LAME_CFG and if it's not empty then the script will run distclean
and ./configure appending [PACKAGE]_CFG to the args.  So to force a
re-configure of LAME enter: LAME_CFG=" " ./mythbuildw32 

> That said, it's not a problem with the script, just a potential gotcha
> with the instructions.  Took a few runs through before it would
> complete.  Everything else appears to have behaved.  Nice work.

Thanks for the feedback.  I'll post a revised script later today
incorporating a number of fixes and suggestions.

-- Lawrence