[Shotwell] Getting 0.3.0 to build

[Matthias Clasen]

On Wed, Nov 4, 2009 at 3:15 PM, Jim Nelson <jim at yorba.org> wrote:


> * I'm not thrilled about running the cache updaters from make install,
> but we do want people to be able to install Shotwell without using a
> package manager.  (We have nothing against package managers, of
> course.)  Are you suggesting we merely copy the schema/desktop files
> and omit the update step, and assume that the caches will update at
> some point in the future?  My concern here is that without the cache
> updates the install doesn't "feel" complete (i.e., the icon doesn't
> display in the Applications menu).  Or, are you saying Makefile should
> respect GCONF_DISABLE_MAKEFILE_SCHEMA_INSTALL to decide if it should
> run the cache updater?

I think having one or more env vars to suppress anything beyond the
putting down of files would be more than sufficient to make packagers
happy, while still allowing
make install to just work in a self-built scenario

> * I'm also not thrilled about sending a HUP signal in a Makefile.
> We'd found docs online suggesting this was necessary.  We'll probably
> just forgo the step ... it smells of patchwork and hackery.
>
> * In your Makefile patch the gcc options use -O2 for a debug build.
> Is this the norm?  I find -O0 useful when, well, debugging, especially
> under gdb.

Oh, sorry. That was an oversight. I didn't mean to send that along.

> * -m64 makes Shotwell impossible to build on 32-bit based installs
> (i.e. my own machine).  Shouldn't this just be omitted?
>
> * I'm having trouble finding good docs on -fexceptions.  This seems to
> be more appropriate for C++.  You do know that Vala generates its own
> exception handling and stack unwinding in straight C?
>
> * Also, your gcc options remove some of the floating-point
> optimizations we included for the color transformation code.  Is there
> a problem with leaving these in?

See above. Please ignore the CFLAGS hunk. These are just the default
CFLAGS that are used in Fedora builds. I'll investigate to possibly
build the color transformation code with special flags. But it seems
plenty fast enough on my T61 anyway...