[OpenAFS-devel] hudson success (i.e. automatically building)

Russ Allbery rra@stanford.edu
Mon, 12 Jul 2010 16:52:03 -0700


Jason Edgecombe <jason@rampaginggeek.com> writes:

> The hudson machine is running Debian Lenny on i386 and building with the
> command "sh regen.sh && ./configure && make"

What I do when verifying patches is:

./regen.sh
headers=`ls /usr/src | grep 'linux-headers.*-686-bigmem' | tail -n 1`
./configure --enable-disconnected --enable-unix-sockets \
    --enable-supergroups --enable-demand-attach-fs --enable-checking \
    --with-linux-kernel-headers="/usr/src/$headers" \
    --with-linux-kernel-packaging --with-xslt-processor=xsltproc \
    --with-html-xsl=/usr/share/xml/docbook/stylesheet/docbook-xsl/html/chunk.xsl
make -j4
(cd doc/man-pages && ./check-pod)
(cd doc/xml/AdminGuide && make)
(cd doc/xml/QuickStartUnix && make)
(cd doc/xml/UserGuide && make)

I think you don't have to pass the explicit kernel header path any more.
You do need to have the kernel headers installed.  Building the kernel as
well is very useful; there are a lot of build failures in that code.

--enable-checking is important for build testing.

Checking the documentation as above would be nice.

> 2. It builds all changes, regardless of the author or approvers.

In the long run, we would want to do something about this, probably by
having some kind of whitelist of people whose changes are always
autobuilt.  We could then also have a whitelist of reviewers (or reuse
that same whitelist) and only build changes from other people if they have
at least a +1 from someone on the reviewer list.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>