Re: [Xen-devel] [PATCH/RFC OSSTEST] Debian PV netboot guest test

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Mon, 2013-11-25 at 16:02 +0000, Ian Jackson wrote:
> Ian Campbell writes ("Re: [PATCH/RFC OSSTEST] Debian PV netboot guest test"):
> > On Mon, 2013-11-25 at 14:44 +0000, Ian Campbell wrote:
> > > I've been working on this on the odd occasion, I think it mostly works,
> > > or it did last I tried which was a while back. I'm sure it is too hacky
> > > in places. My plan was to clean it up on the next test day.
> > > 
> > > I'm mostly just sending this for Wei's benefit since he is independently
> > > looking at adding Debian HVM guest tests for OVMF purposes.
> > 
> > FWIW I was planning to introduce a separate "distros" flight to osstest,
> > which would semiregularly test a bunch of distros (of which Debian
> > happens to just be the one I was interested in).
> > 
> > My lashed up skanky thinking here is appended..
> 
> This is quite nasty, isn't it.

Yeah, it was mostly just for completeness rather than because I thought
it needed commenting on, I should have said so, sorry. Thanks anyway.

> 
> > diff --git a/make-distro-flight b/make-distro-flight
> > new file mode 100644
> > index 0000000..b02281d
> > --- /dev/null
> > +++ b/make-distro-flight
> > @@ -0,0 +1,201 @@
> > +#!/bin/bash
> ...
> > +stripy () {
> > +        local out_vn="$1"; shift
> > +        local out_0="$1"; shift
> > +        local out_1="$1"; shift
> > +        local out_val=0
> > +        local this_val
> > +        local this_cmp
> > +        while [ $# != 0 ]; do
> > +                this_val="$1"; shift
> > +                this_cmp="$1"; shift
> > +                if [ "x$this_val" = "x$this_cmp" ]; then
> > +                        out_val=$(( $out_val ^ 1 ))
> > +                fi
> > +        done
> > +        eval "$out_vn=\"\$out_$out_val\""
> 
> make-flight is hardly the most well-structured area of osstest, but
> at the very least, much of this common code needs to be in a common
> script sourced by both make-flights.

That was along the lines of what I was thinking of doing eventually.

> 
> As an alternative to this, perhaps you want to tangle make-flight some
> more by making its contents conditional :-/.  Hmm.  Not sure that
> appeals.
> 
> > diff --git a/sg-report-flight b/sg-report-flight
> > index 69ef252..65733d1 100755
> > --- a/sg-report-flight
> > +++ b/sg-report-flight
> > @@ -95,7 +95,9 @@ while (@ARGV && $ARGV[0] =~ m/^-/) {
> ...
> >      my $flightinfo= $dbh_tests->selectrow_hashref(<<END);
> >          SELECT * FROM flights
> > -            WHERE flight=$flight
> > +            WHERE flight="$flight"
> 
> This needs to be done with ?-substitution, because in the actual
> database the flight columns are INTEGER.

Ah, makes sense.

> 
> > diff --git a/standalone-reset b/standalone-reset
> > index 8be7e86..b045dba 100755
> > --- a/standalone-reset
> > +++ b/standalone-reset
> > @@ -156,7 +156,10 @@ fi
> >  export BUILD_LVEXTEND_MAX
> >  
> >  OSSTEST_FLIGHT=$flight \
> > -./make-flight "$branch" "$xenbranch" play $buildflight >/dev/null
> > +mainflight=`./make-flight "$branch" "$xenbranch" play $buildflight` 
> > >/dev/null
> > +
> > +OSSTEST_FLIGHT=${flight:-standalone}-distros \
> > +./make-distro-flight distros "$xenbranch" play $mainflight >/dev/null
> 
> This really doesn't make any kind of sense.

What I wanted to do eventually was have the distro flight pickup the
build results from the most recent successful normal test flight (and/or
avoid duplicating those bits of make-flight).

Perhaps I should just rebuild everything though. And the duplciation can
be handled via the library scriptlet as you suggested.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel