Re: [PATCH 1/2 v3] tst-fanotify: new simple test

[Mike Frysinger <vapier at gentoo dot org>]

On Tuesday 03 September 2013 18:40:20 Joseph S. Myers wrote:
> On Tue, 3 Sep 2013, Mike Frysinger wrote:
> > On Tuesday 03 September 2013 17:08:47 Joseph S. Myers wrote:
> > > On Tue, 3 Sep 2013, Mike Frysinger wrote:
> > > > short of doing a AC_CHECK_HEADER, i don't see a solution here.
> > > 
> > > Increase the required headers version (would need proposing in a
> > > separate thread with its own subject).
> > 
> > i think adding an AC_CHECK_HEADER is better than bumping the min build
> > kernel version to 2.6.37 (i think that's the first release to include
> > this).  seems unreasonable to move that far ahead for a single (minor)
> > test.
> 
> The benefits of a more recent version go beyond the one test - they mean
> much more generally that code doesn't need #ifdef conditionals on __NR_*
> if all architectures are sure to have the macro defined when a newer
> kernel headers version is in use, for example.

true, but i don't think the current overhead is terribly burdensome.  i also 
suspect that many of the existing __NR_ checks in the tree are already 
obsolete, but no one has gone through to clean them up.  so judging the number 
of ifdef checks we can drop due to moving from 2.6.19 to 2.6.37 is hard to 
judge.

also just grepping the tree for "#ifdef __NR_xxx" doesn't tell the whole 
story.  some of the checks are to pick between two different styles arches have 
adopted rather than there being one answer that everyone agrees on.
-mike

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