Re: [Acl-devel] [PATCH] attr: refer to ENODATA instead of ENOATTR in man pages

[Mike Frysinger]

On Thu 27 Mar 2014 13:23:09 Florian Weimer wrote:
> On 03/27/2014 07:19 AM, Mike Frysinger wrote:
> > On Tue 21 Jan 2014 14:27:33 Kamil Dudka wrote:
> >> The ENOATTR alias is only provided if <attr/xattr.h> is included, not
> >> with the more standard <sys/xattr.h> header file.  ENODATA is correct
> >> in both cases.
> > 
> > well, that's the case if the OS doesn't support ENOATTR, so the library
> > uses ENODATA in its place.  if the OS does have its own, then we return
> > ENOATTR everywhere, so checking ENODATA won't help.
> > 
> > it might be the case on Linux that ENOATTR == ENODATA, but i think the
> > goal is for libattr to be more than Linux.
> 
> My concern is that it's confusing on Linux that you don't always get
> ENOATTR, so if you write code based on this manual page, it does not work.

sure, your compliant is perfectly reasonable :)

> Hmm, perhaps we should just make sure that glibc header defines ENOATTR,
> too?  That would resolve this issues as well.

i'm not sure how well that'd go over

the man-pages project has dealt with this too:
https://bugzilla.kernel.org/show_bug.cgi?id=51871

and their fix was:
https://git.kernel.org/cgit/docs/man-pages/man-pages.git/commit/man2/getxattr.2?id=2965fbf16539f9442c8a85dc356a50f41bb443dc

 .I errno
 is set to
 .BR ENOATTR .
+.RB ( ENOATTR
+is defined to be a synonym for
+.BR ENODATA
+in
+.IR <attr/xattr.h> .)

wdyt of that solution ?
-mike

signature.asc
Description: This is a digitally signed message part.