Re: [linux-next test] 56810: regressions - FAIL

[Ian Campbell <ian.campbell@citrix.com>]

(I thought I'd hit send on this already, sorry if it is a repeat)

On Fri, 2015-05-22 at 10:17 +0100, Ian Campbell wrote:
> On Thu, 2015-05-21 at 18:05 +0100, Ian Campbell wrote:
> > Assuming this flight produces something useful (i.e. a pass) then I'll
> > look into a bisect.
> 
> With this and a couple more adhoc tests I've narrowed it down to
> somewhere between v3.15 and v3.16, which is 13830 commits. That might be
> a few too many for the bisector to cope but I'll see what it says.

With a minor tweak (below) and telling it to consider 14000 rather than
the default of 1000 commits it did seem to produce an answer.

The resulting bisection.png was 12M and opening it oomed my workstation!

bisection.ps was only 2M and opening it produced a frankly hilarious
graph, but I suppose it will eventually produce an answer so I set
something going.

I've spared you the barrage of mails this would normally generate. You
can see the progress in:
        osstest.xs.citrite.net/~osstest/testlogs/results-adhoc/bisect.linux-linus.test-amd64-i386-freebsd10-amd64..html
but I would recommend looking directly at:
        osstest.xs.citrite.net/~osstest/testlogs/results-adhoc/bisect.linux-linus.test-amd64-i386-freebsd10-amd64..ps
Since the html includes the png...

log2(14000) is 17 and a bit, so I guess we might see an answer by
Tuesday (Monday is a holiday in the UK). I'll try and keep and eye on it
over the weekend and if it looks to be going mad I'll stop it.

Ian.

diff --git a/adhoc-revtuple-generator b/adhoc-revtuple-generator
index 2b04c11..1f5b97a 100755
--- a/adhoc-revtuple-generator
+++ b/adhoc-revtuple-generator
@@ -549,7 +549,7 @@ sub main () {
     my @trees_continuous;
     foreach my $tree (@trees) {
         my $gen= tree_get_gen($tree);
-        my $count= 1000;
+        my $count= $num;
         my $found= 0;
         my $top= undef;
         while ($count-- > 0) {