Date: Tue, 20 Nov 2012 18:03:57 +1100 (EST) From: Bruce Evans <brde@optusnet.com.au> To: Eitan Adler <eadler@FreeBSD.org> Cc: svn-src-head@FreeBSD.org, svn-src-all@FreeBSD.org, src-committers@FreeBSD.org Subject: Re: svn commit: r243321 - head/usr.sbin/edquota Message-ID: <20121120172226.R1115@besplex.bde.org> In-Reply-To: <201211200212.qAK2C1sT097754@svn.freebsd.org> References: <201211200212.qAK2C1sT097754@svn.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 20 Nov 2012, Eitan Adler wrote:
> Log:
> Remove unneeded includes.
>
> Tested with "make universe"; there are no conditional features.
"make universe" can't find such features. Except inversely -- when it
doesn't find them, it usually means that there is a bug in the headers.
> Modified: head/usr.sbin/edquota/edquota.c
> ==============================================================================
> --- head/usr.sbin/edquota/edquota.c Tue Nov 20 01:57:21 2012 (r243320)
> +++ head/usr.sbin/edquota/edquota.c Tue Nov 20 02:12:01 2012 (r243321)
> @@ -49,8 +49,6 @@ __FBSDID("$FreeBSD$");
> * Disk quota editor.
> */
>
> -#include <sys/param.h>
> -#include <sys/stat.h>
> #include <sys/file.h>
> #include <sys/mount.h>
> #include <sys/wait.h>
>
This removes used includes.
<sys/param.h> is a documented prerequisite of <sys/mount.h>. sys/mount.h
has some pollution but not that until recently. <sys/mount.h> has included
<sys/ucred.h> for a long time, and someone recently added the following
disgusting pollution to <sys/ucred.h>:
- <sys/ucred.h> includes <bsm/audit.h>
- <bsm/audit.h> includes <sys/param.h> and all of its standard pollution
According to cc -E -MM, this is:
% f.o: f.c /usr/include/bsm/audit.h /usr/include/sys/param.h \
^^^^^^^^^^^^^^^^^^^^^^^^
Polluting headers are underlined.
% /usr/include/sys/_null.h /usr/include/sys/types.h \
^^^^^^^^^^^^^^^^^^^^^^^^
<bsm/audit.h> is of low quality. It also includes <sys/types.h> after
<sys/param.h>. This is a style bug. <sys/types.h> is standard pollution
in <sys/param.h>.
% /usr/include/sys/cdefs.h /usr/include/machine/endian.h \
% /usr/include/x86/endian.h /usr/include/sys/_types.h \
% /usr/include/machine/_types.h /usr/include/x86/_types.h \
% /usr/include/sys/_pthreadtypes.h /usr/include/sys/_stdint.h \
% /usr/include/sys/select.h /usr/include/sys/_sigset.h \
^^^^^^^^^^^^^^^^^^^^^^^^^
Standard pollution in <sys/types.h>.
% /usr/include/sys/_timeval.h /usr/include/sys/timespec.h \
^^^^^^^^^^^^^^^^^^^^^^^^^^^
Standard pollution in <sys/types.h>.
% /usr/include/sys/_timespec.h /usr/include/sys/syslimits.h \
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
% /usr/include/sys/signal.h /usr/include/machine/_limits.h \
^^^^^^^^^^^^^^^^^^^^^^^^^
Standard pollutions in <sys/param.h>.
% /usr/include/x86/_limits.h /usr/include/machine/signal.h \
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
% /usr/include/machine/trap.h /usr/include/x86/trap.h \
^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^
Standard pollutions in <sys/signal.h> and thus in <sys/param.h>.
% /usr/include/machine/param.h /usr/include/machine/_align.h \
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
% /usr/include/x86/_align.h /usr/include/sys/limits.h
^^^^^^^^^^^^^^^^^^^^^^^^^
Standard pollutions in <sys/param.h>.
To remove a single header, you must know all the standard pollutions and
check cc -E -MM output to verify that no nonstandard pollutions are
depended on.
That was for <sys/mount.h>. Now for some others.
> -#include <sys/param.h>
> -#include <sys/stat.h>
> #include <sys/file.h>
<sys/types.h> a is documented prerequisite for <sys/stat.h>. It was
correct to get it via <sys/param.h>. However, <sys/stat.h> is a POSIX
header. POSIX required <sys/types.h> before <sys/stat.h> in 1988, but
removed this requirement in 2001 or before. FreeBSD is slowly catching
up with this:
- FreeBSD <sys/stat.h> has always been massively polluted by including
<sys/time.h> and all if its pollution (which includes <sys/types.h>.
The polluting <sys/time.h> became unnecessary in about 1995 and is
mostly fixed in my version (I include <sys/timespec.h> instead).
- FreeBSD man pages were originally very inconsistent about documenting
the <sys/types.h> prereq. It wasn't ever a prereq in FreeBSD, but was
required for portability. None documented this of course.
- At about the same time that POSIX removed the prereq, lots of man pages
were "fixed" to document the old POSIX prereq.
- man pages (e.g., stat(2)) still document the old POSIX prereq. More code
than before probably doesn't satisfy this.
<sys/file.h> is documented as having no prereqs (in flock(2)). It
satisfies this by polluting itself with <sys/types.h> and other
includes. The pollution is not documented of course.
So this commit is correct for <sys/stat.h> and <sys/file.h> except it
depends on undocumented details for the former.
Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20121120172226.R1115>