Re: [AD] Cross-compiling, and related outstanding issues

[Evert Glebbeek <eglebbk@xxxxxxxxxx>]

> Why so? Normal users don't need to run make depend (explicitly). And
> those that would be toying with the dependancies would need to know what
> they're doing anyway. Indeed, you could even have two targets, 'depend'
> which builds the dependancy file if it doesn't exist and 'regen-dep'
> which forces dependancy regeneration.

True; it just feels redundant. In that case, I would argue for adding
makefile.dep to the library dependencies and having the rule to build it
read `make depend'

> It's non-standard behavior, though. And, unless I missed something in
> the docs about building from CVS, there's nothing saying it needs to be
> done manually,

Actually, there is in the MinGW32 docs:

If you are using a CVS version of Allegro, run 'make depend' to generate
the build dependencies[...]

I suppose the docs would have to be checked to make sure that they are
consistent across platforms.
If I recall correctly, under *nix make depend is run automatically after
configure (or when running make for the first time). I usually run it
manually anyway though, so I could be mistaken here.

> and the eventual errors you get from not calling it
> aren't indicative of what the real problem is (who would think missing
> header files are caused by forgetting to generate dependancy info?).

Well, those particular ones are generated files... so if they're not being
generated during the build process that is a hint that you need to
regenerate the dependency rules. But for that you need to realise that the
files should be generated during the compilation process. I guess it could
be documented in the docs for the cross-compiler.

Evert