svn commit: r422981 - in head/dns: bind9-devel bind910 bind911 bind99

Baptiste Daroussin bapt at FreeBSD.org
Tue Oct 4 14:40:22 UTC 2016


On Tue, Oct 04, 2016 at 04:36:25PM +0200, Mathieu Arnold wrote:
> Le 04/10/2016 à 16:35, Baptiste Daroussin a écrit :
> > On Tue, Oct 04, 2016 at 04:27:20PM +0200, Mathieu Arnold wrote:
> >> Le 04/10/2016 à 16:22, John Marino a écrit :
> >>> On 10/4/2016 09:18, Mathieu Arnold wrote:
> >>>> Le 04/10/2016 à 16:16, John Marino a écrit :
> >>>>> On 10/4/2016 09:13, Mathieu Arnold wrote:
> >>>>>> Le 04/10/2016 à 16:04, John Marino a écrit :
> >>>>>>> We build under a very heavy load which flushes out marginally unsafe
> >>>>>>> ports.
> >>>>>> Ok, so make(1) from dragonfly has the same bug make(1) from FreeBSD 9
> >>>>>> has, feel free to fix it in dragonfly's port tree.
> >>>>> master has bmake 20160818 on it (for last 5 weeks)
> >>>>> Release 4.6 has bmake 20141111.
> >>>>> Do you know if make bug only applies to master?
> >>>> I have no idea.  I was told the problem was with make(1) on FreeBSD 9,
> >>>> which seemed to be right, as it does not fail at all on FreeBSD 10/11
> >>>> with -j 2-10.
> >>> The -j number is not the only factor here.  I've seen ports pass under
> >>> very high -j numbers but start failing when the server gets loaded.
> >>>
> >>> DragonFly has the lastest bmake, modern binutils, modern gcc and it
> >>> doesn't have fmake (what freebsd 9 uses).
> >>>
> >>> Why is it so critical to classify bind910 as jobs safe when there
> >>> clearly is a question about it?  Let's not immediately assume DF is at
> >>> fault here.  As I mentioned before, it could easily be the build tests
> >>> you're doing aren't sufficient to flush this out.  It *was* marked
> >>> UNSAFE before, obviously with good reason.  (albeit undocumented)
> >> It works just fine on all supported FreeBSD versions as it is, like I
> >> said, feel free to change it in dragonfly's ports tree.
> >>
> > No it does not see my reply I have reproduced at least 2 times on make -j40.
> >
> > It is not 100% reproducible but happen from time to time.
> 
> Oh, ok, I thought you were saying it was ok for you too.
> 

My bad I replied too quickly and didn't take time to reread myself

Sorry,
Bapt
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/svn-ports-all/attachments/20161004/e4f67fcc/attachment.sig>


More information about the svn-ports-all mailing list