You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Rasmus Lerdorf <ra...@lerdorf.on.ca> on 1998/01/25 22:46:52 UTC

Changing httpd.h

Would it make sense to bumb the MODULE_MAGIC_NUMBER on any changes to the
httpd.h file?  To me, since any program that uses the Apache API has to
include httpd, it makes sense to treat httpd.h as part of the API.

This change by Ken breaks any Apache API code:

http://www.apache.org/websrc/cvsweb.cgi/apachen/src/main/httpd.h.diff?r1=1.172&r2=1.173

Or at least, it breaks PHP because I was not adding -I...src/ap to my
build process as I have no need for any of those ap_* routines.

-Rasmus


Re: Changing httpd.h

Posted by ra...@bellglobal.com.
> Bumping the MMN wouldn't have kept your build from breaking; it's a
> run-time control, not a compile-time one.

Bumping the MMN alone never prevents a build from breaking.  But, in my
build process I check MMN for all sorts of stuff, and this could have
been one.  if MMN > whatever, add -I.../ap, for example.

> Of course, this issue would be a complete non-issue if we had
> a single src/include directory.  That's been proposed, and a
> couple of people have said it sounds like a good idea, but
> I'm still suspicious - what will the impact be on Win32, for instance?

This would make life easier on me.  Having to add -I's for src/main, 
src/os/unix and src/ap is a pain.

-Rasmus

Re: Changing httpd.h

Posted by Rodent of Unusual Size <Ke...@Golux.Com>.
Rasmus Lerdorf wrote:
> 
> Would it make sense to bumb the MODULE_MAGIC_NUMBER on any changes to the
> httpd.h file?  To me, since any program that uses the Apache API has to
> include httpd, it makes sense to treat httpd.h as part of the API.
	:
> Or at least, it breaks PHP because I was not adding -I...src/ap to my
> build process as I have no need for any of those ap_* routines.

Bumping the MMN wouldn't have kept your build from breaking; it's a
run-time control, not a compile-time one.

That change didn't break (or even affect) the run-time API at all,
so why should MMN be incremented?

An alternative (for this one case), of course, is to remove ap.h
from httpd.h, and add it to all of the .c files that *do* call
one or another of the libap routines.  That seemed like such a
massive change, though, that I concluded just about anything that
included httpd.h might be using the libap routines at will, so I
put it in a common location.

Of course, this issue would be a complete non-issue if we had
a single src/include directory.  That's been proposed, and a
couple of people have said it sounds like a good idea, but
I'm still suspicious - what will the impact be on Win32, for instance?

#ken	P-)}