Re: auto-dep cannot possibly work?

[Mike Shal]

On 11/16/09, Paul Smith <address@hidden> wrote:
> On Mon, 2009-11-16 at 19:55 -0500, Mike Shal wrote:
>  > d! How the heck should you or I know what files need to be modified
>  > just because a new header was added?
>
>
> Are you kidding me?  You BETTER darn well know.  Otherwise, your code
>  will not compile, or even worse it may compile but give you completely
>  incorrect results.  You can't just go around adding headers without
>  knowing what impact that will have on the code!

No, I wasn't kidding about that either.

I guess in my head I was comparing it to the case where you modify an
existing header. Here, I expect make to re-compile every C file that
is affected by the change. That way I know which files don't compile
in case I did something wrong. I don't try to memorize all the C files
that include that header.

Similarly, if I add a new header, I expect make to re-compile every C
file that is affected by the change. That way I know which files don't
compile in case I did something wrong. I don't try to memorize all the
C files that tried to include that header earlier in their include
path, but didn't because it wasn't there.

Ignoring the specifics of make for a moment, I don't see why these two
cases should be so different that the build system is responsible for
handling the first case, but the user is responsible for the second.

>  But, you seem to have discovered a way to do what you want and this
>  discussion has moved away from technical solutions into the murky realm
>  of process and "best practices", so it seems a good place to stop.

Sorry to hijack the thread -- I'll get off my soapbox now :). But for
the record, I wasn't really advocating having a crazy include path
with config.h's strewn all over the place. That would drive me crazy
too...

-Mike