Unit tests of internal functions

[Ken Raeburn]

On Dec 28, 2009, at 19:17, Sam Hartman wrote:
> Option 3 seems reasonable in terms of result, but I'm not convinced  
> the
> build system complexity is justified.
>
> If my input is going to be seriously considered it's a good idea to
> sanity check that  the build system complexity of our current practice
> isn't more than that of option 3.

I think it would be easier if we were using a script to build shared  
libraries given a list of OBJS.SH files and the other data, instead of  
sticking everything in shell commands directly in the makefiles, but  
Tom has objected to the script approach in the past.  (Bad experiences  
with libtool?)  Having to fit every new bit of functionality into one  
long command line makes things pretty ugly, IMO.  There may be ways to  
break things apart more sanely, though, like creating a makefile  
target to build a temporary file to hold some of the command-line  
options we need to pass to the linker on platform X, etc.

Ken