[lfs-dev] perl tests loop at 100% CPU in current SVN
zarniwhoop at ntlworld.com
Mon Aug 8 16:27:15 PDT 2016
On Mon, Aug 08, 2016 at 04:45:12PM -0500, Bruce Dubbs wrote:
> Ken Moffat wrote:
> > On Mon, Aug 08, 2016 at 09:26:55PM +0100, Ken Moffat wrote:
> > >
> > > At the moment I suspect some of it might be down to an inopportune
> > > kernel config, but I suppose I got lucky in the manual run (test
> > > terminated after division by zero). I'll give that another go, then
> > > think about changing my kernel config to be closer to the haswell.
> > >
> > Well, the test in chroot is still in lib/Benchmark, but the tests on
> > the host finished in 16m27 with only op/time.t failing. Weird.
> I don't know what to say Ken. From my log:
> All tests successful.
> Elapsed: 483 sec
> u=2.27 s=0.91 cu=207.90 cs=7.71 scripts=2242 tests=850411
> Run with -j1.
> -- Bruce
Yes - I did get through the tests in the scripted chroot build of
20160527 with a 4.6.0 kernel - but after almost 2h45, and with the
same three failures. The build then continued, until I noticed it
had got through perl - by that stage a segfault in /bin/sh during
the automake tests showed on the screen.
The times on this box will be much longer (it's AMD, and old with
slow memory and only SATA-2).
I've now booted the revised 4.7 config but I haven't yet tried
testing perl on the host. I've got a few things to do in
spreadsheets, then I'll dig out a memtest86 CD.
This is the box which for years has segfaulted when compiling with
-j4, even building the kernel after a cold boot. I have/had a theory
that perhaps the BIOS screws up the initialisation, so at the moment
I'm dropping caches before doing any builds, and -j4 has been fine.
Except, the repeat of perl was using -j1, but ISTR we force -j4 for
the automake tests so I might have blown that theory.
`I shall take my mountains', said Lu-Tze. `The climate will be good
for them.' -- Small Gods
More information about the lfs-dev