diff mbox

[Xen-devel,OSSTEST] mfi-common: Only override the pvops kernel repo for linux-arm-xen branch (Was: Re: [linux-linus test] 24817: regressions - FAIL)

Message ID 1392028633.5117.29.camel@kazak.uk.xensource.com
State New
Headers show

Commit Message

Ian Campbell Feb. 10, 2014, 10:37 a.m. UTC
On Sun, 2014-02-09 at 16:04 +0000, xen.org wrote:
> 
>  build-armhf-pvops             4 kernel-build                 fail   never pass 

+ git checkout 494479038d97f1b9f76fc633a360a681acdf035c
fatal: reference is not a tree: 494479038d97f1b9f76fc633a360a681acdf035c

This is because it is using git://xenbits.xen.org/linux-pvops.git
instead of the tree it should be testing...

The following fixes it for me, but although the results are as I wanted
I'm not 100% sure about this override in the first place. In my
experiments with cr-daily-branch I see:

        Branch		$TREE_LINUX		$TREE_LINUX_ARM
        
        xen-unstable	pvops			pvops
        linux-linus	torvalds		pvops
        linux-arm-xen	arm-xen			arm-xen

        Key:
        [arm-xen] git://git.kernel.org/pub/scm/linux/kernel/git/sstabellini/xen.git
        [torvalds] git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
        [pvops] git://xenbits.xen.org/linux-pvops.git
        
IOW $TREE_LINUX is always correct.

When invoking make-flight directly both are always "pvops".

8<---------------------------------------

From 344b0ca5e623d8212ee9d5452f19aba2df2c97fe Mon Sep 17 00:00:00 2001
From: Ian Campbell <ian.campbell@citrix.com>
Date: Mon, 10 Feb 2014 09:33:00 +0000
Subject: [PATCH] mfi-common: Only override the pvops kernel repo for
 linux-arm-xen branch

Otherwise e.g. linux-linus tries to use the wrong tree and fails.

I have confirmed that for flights on xen-unstable, linux-linus and
linux-arm-xen the only difference in the runvars is
linux-linux.build-armhf-pvops.tree_linus which changes from
git://xenbits.xen.org/linux-pvops.git to
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git as
expected.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
---
 mfi-common | 20 ++++++++++++--------
 1 file changed, 12 insertions(+), 8 deletions(-)

Comments

Ian Campbell Feb. 10, 2014, noon UTC | #1
On Mon, 2014-02-10 at 11:51 +0000, Ian Jackson wrote:
> Ian Campbell writes ("[PATCH OSSTEST] mfi-common: Only override the pvops kernel repo for linux-arm-xen branch (Was: Re: [Xen-devel] [linux-linus test] 24817: regressions - FAIL)"):
> > On Sun, 2014-02-09 at 16:04 +0000, xen.org wrote:
> > > 
> > >  build-armhf-pvops             4 kernel-build                 fail   never pass 
> 
> Surely this is correct.  Linus's tip doesn't pass this test.
> 
> > + git checkout 494479038d97f1b9f76fc633a360a681acdf035c
> > fatal: reference is not a tree: 494479038d97f1b9f76fc633a360a681acdf035c
> > 
> > This is because it is using git://xenbits.xen.org/linux-pvops.git
> > instead of the tree it should be testing...
> 
> That's exactly what it should be using, isn't it ?  After all, in
> principle, eventually, Linus's tip should work.  In the meantime it
> doesn't and this is a real bug.

Linus' tip is git.kernel.org/...torvalds/xen.git. This linux-pvops.git
tree contains the output of this pushgate.

The error is that it appears to be looking in the output tree for the
new revision, which obviously only exists in the input tree.

Ian.
Ian Campbell Feb. 10, 2014, 12:01 p.m. UTC | #2
On Mon, 2014-02-10 at 13:48 +0200, Pasi Kärkkäinen wrote:
> On Mon, Feb 10, 2014 at 10:37:13AM +0000, Ian Campbell wrote:
> > On Sun, 2014-02-09 at 16:04 +0000, xen.org wrote:
> > > 
> > >  build-armhf-pvops             4 kernel-build                 fail   never pass 
> > 
> > + git checkout 494479038d97f1b9f76fc633a360a681acdf035c
> > fatal: reference is not a tree: 494479038d97f1b9f76fc633a360a681acdf035c
> > 
> > This is because it is using git://xenbits.xen.org/linux-pvops.git
> > instead of the tree it should be testing...
> > 
> > The following fixes it for me, but although the results are as I wanted
> > I'm not 100% sure about this override in the first place. In my
> > experiments with cr-daily-branch I see:
> > 
> >         Branch		$TREE_LINUX		$TREE_LINUX_ARM
> >         
> >         xen-unstable	pvops			pvops
> >         linux-linus	torvalds		pvops
> >         linux-arm-xen	arm-xen			arm-xen
> > 
> >         Key:
> >         [arm-xen] git://git.kernel.org/pub/scm/linux/kernel/git/sstabellini/xen.git
> >         [torvalds] git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
> 
> Isn't that the "old" repo (but a symlink to the new one)? The new one being:
> git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
> 

Presumably. I don't think it really matters here though.
Ian Campbell Feb. 10, 2014, 12:10 p.m. UTC | #3
On Mon, 2014-02-10 at 12:05 +0000, Ian Jackson wrote:
> Ian Campbell writes ("Re: [PATCH OSSTEST] mfi-common: Only override the pvops kernel repo for linux-arm-xen branch (Was: Re: [Xen-devel] [linux-linus test] 24817: regressions - FAIL)"):
> > Linus' tip is git.kernel.org/...torvalds/xen.git. This linux-pvops.git
> > tree contains the output of this pushgate.
> > 
> > The error is that it appears to be looking in the output tree for the
> > new revision, which obviously only exists in the input tree.
> 
> Oh, I see, yes.
> 
> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
> 
> Unless you have more bugfixes in the pipeline, you should probably
> push that right away so that it doesn't get entangled with the
> forthcoming change to use FreeBSD RELEASE instead (as prompted by
> Roger).

I don't have anything else pending[0], but I'm about to borrow one of
the marilith machines and do some local manual runs to get a quicker
turnaround on this stuff. I wouldn't expect that to result in any
patches before tomorrow though.

Ian.

[0] actually not quite true, I have a bunch of older minor patches:
        [PATCH OSSTEST] README: Add some core concepts and terminology
        [PATCH OSSTEST] standalone: add blessing to flights table.
        [PATCH OSSTEST] cri-args-hostlists: Allow environment to control OSSTEST_CONFIG

But these aren't urgent or tied up in anything else.

Ian.
diff mbox

Patch

diff --git a/mfi-common b/mfi-common
index 8f56092..f7f981e 100644
--- a/mfi-common
+++ b/mfi-common
@@ -44,6 +44,11 @@  create_build_jobs () {
 
     if [ "x$arch" = xdisable ]; then continue; fi
 
+    pvops_kernel="
+      tree_linux=$TREE_LINUX
+      revision_linux=${REVISION_LINUX:-${DEFAULT_REVISION_LINUX}}
+    "
+
     case "$arch" in
     armhf)
       case "$branch" in
@@ -59,10 +64,13 @@  create_build_jobs () {
       xen-4.1-testing) continue;;
       xen-4.2-testing) continue;;
       esac
-      pvops_kernel="
-        tree_linux=$TREE_LINUX_ARM
-        revision_linux=${REVISION_LINUX_ARM:-${DEFAULT_REVISION_LINUX_ARM}}
-      "
+
+      if [ "$branch" = "linux-arm-xen" ]; then
+        pvops_kernel="
+          tree_linux=$TREE_LINUX_ARM
+          revision_linux=${REVISION_LINUX_ARM:-${DEFAULT_REVISION_LINUX_ARM}}
+        "
+      fi
       pvops_kconfig_overrides="
         kconfig_override_y=CONFIG_EXT4_FS
       "
@@ -71,10 +79,6 @@  create_build_jobs () {
       case "$branch" in
       linux-arm-xen) continue;;
       esac
-      pvops_kernel="
-        tree_linux=$TREE_LINUX
-        revision_linux=${REVISION_LINUX:-${DEFAULT_REVISION_LINUX}}
-      "
       ;;
     esac