[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Are "incomplete" headers bugworthy?

On Tue, Apr 01, 2003 at 05:27:32PM +0200, Matthias Urlichs wrote:
> Some header files require 
> #include <some_header.h>
> #include <whatever_I_really_want.h>
> instead of a simple
> #include <whatever_I_really_want.h>
> I noticed this while Debianizing my ITP:yapsnmp. It checks a couple of
> libsnmp4.2's include files via AC_CHECK_HEADER. Unfortunately, the
> snmp_include.h file uses in_addr_t without first including the libc header
> where this is defined. Thus it fails.
> IMHO, it is reasonable for all C header files to include any other header file
> they need for compileability. My question is, is there a consensus on this
> reasonableness?

No. A lot of headers are like this; you're gonna have to fix thousands of
source files (some of which are broken in this way for non-trivial
reasons). IIRC, some libc headers won't compile properly if you do this.

Personally, I agree with you that this is ugly; but unless you're willing
to wade through a LOT of code (and I mean a LOT) to fix this, I'd refrain
from mass-filing bugs. Lintian checks might be OK, but it will be annoying
because it isn't the maintainer's fault that headers are screwed up this


Marketing: the art of convincing people to pay for what they didn't need
before and you can't deliver after.

Reply to: