Re: FW: [Xen-devel] test structure missing files

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Wed, 2011-10-26 at 05:36 +0100, Windows Live wrote:


> What is being achieved is the making of the version bump of xen,
> xen-tools and xen-pvgrub, 4.1.2. Having adjusted and repaired a number
> of bug fixes to xen-4.1.1, I couldn't help but wonder why the test
> phase was set restricted.  A test phase ,if present, in a package
> should be incorporated into the ebuild.

FWIW the "test" target appears to be very limited (it tests the python
bindings, not all that thoroughly AFAICT).

> I have added the ebuild, but I don't really think it will supply
> anything useful. To make it simpler, there are NO patches applied.
> Re-setting the path via -I ofcourse we can do. Frankly I doubt the
> 'fix' could have made things worse.  On repeating make test, the
> header was found and it progressed a few dozen lines or so further.

Once that header was installed it would be useless to anyone trying to
compile against it because it now contains paths relative to the xen
source tree layout. In fact I suspect your change would also break other
parts of the in tree build, by virtue of them being at different depths
in the hierarchy.

> The original error is clearly enough displayed in the initial
> description; 

> In file included from xen/lowlevel/xc/xc.c:23:0:
> ../../tools/libxc/xc_dom.h:17:31: fatal error: xen/libelf/libelf.h: No
> such file or directory
> compilation terminated.

In general the interesting stuff will likely be in the few lines before
the actual error, in particular the gcc invocation and sometimes the
make output about which directory it is building in etc. Please be wary
of trimming too much when posting error messages.

I just managed to reproduce the error though, it was non-obvious because
I had to run "make clean" in the tools tree first. I think you need to
have built the tools tree before running the test target, at least this
seems logical and works for me.

I suppose we should find a way to show this in the top-level Makefile
dependencies. Perhaps this is sufficient:

diff -r e6af32ecb448 Makefile
--- a/Makefile  Tue Oct 18 14:15:09 2011 +0100
+++ b/Makefile  Wed Oct 26 07:23:05 2011 +0100
@@ -31,7 +31,7 @@ endif
 # course, many tests require a machine running Xen itself, and these are
 # handled elsewhere.
 .PHONY: test
-test:
+test: tools
        $(MAKE) -C tools/python test
 
 # build and install everything into local dist directory

I get failures now because I didn't build/test on a Xen host. That might
be a show stopper for an ebuild as well?

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel