On 11/22/2013 09:59 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange@redhat.com>
The python binding now lives in
http://libvirt.org/git/libvirt-python.git
that repo also provides an RPM which is upgrade compatible with the old libvirt-python sub-RPM.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com> --- .gitignore | 11 - Makefile.am | 7 +- autobuild.sh | 6 +- cfg.mk | 7 +- configure.ac | 86 -
These changes are important...
libvirt.spec.in | 36 +- m4/virt-compile-warnings.m4 | 8 - mingw-libvirt.spec.in | 1 -
and these...
run.in | 9 -
and this (hmm, do we need a ./run counterpart added to libvirt-python to make it easy to import a built but uninstalled python bindings for testing purposes?)
37 files changed, 7 insertions(+), 13369 deletions(-)
The rest of the patch is mechanical 'git rm'; it might have been nice to use git's -O/path/to/file to order the patch to float the interesting parts up top and trim the rest of the email. grepping for python still had hits after your patch; here's my analysis: ChangeLog-old - good to keep; we aren't changing details of historical releases bootstrap.conf - unsure: Does bootstrap.conf still need to worry about finding python-config, or is that a thing of the past? cfg.mk - keep; we still have .py files for docs docs/Makefile.am docs/apibuild.py docs/apps.html.in docs/bindings.html.in docs/errors.html.in docs/hooks.html.in docs/index.py docs/library.xen docs/news.html.in docs/python.html.in docs/sitemap.html.in docs/testsuites.html.in docs/windows.html.in - good to keep all docs stuff for doc generation libvirt.spec.in - good to keep; only mentions the historical %changelog src/esx/esx_vi_generator.py src/hyperv/hyperv_wmi_generator.py src/util/virconf.c src/util/virkeycode-mapgen.py - good to keep all the C code generators We also need to scrub the code base for references to 1.1.5 and instead refer to 1.2.0.
diff --git a/.gitignore b/.gitignore index 6b024e7..faabd33 100644 --- a/.gitignore +++ b/.gitignore @@ -91,17 +91,6 @@ /mkinstalldirs /po/* /proxy/ -/python/generated.stamp -/python/generator.py.stamp -/python/libvirt-export.c -/python/libvirt-lxc-export.c -/python/libvirt-lxc.[ch] -/python/libvirt-qemu-export.c -/python/libvirt-qemu.[ch] -/python/libvirt.[ch] -/python/libvirt.py -/python/libvirt_lxc.py -/python/libvirt_qemu.py
Oddly enough, removing these lines will mean that someone with an incremental tree that likes to switch branches between master and any earlier branch (say v1.0.5-maint) will now see git complaining about untracked files. I personally tend to avoid removing lines from .gitignore (if we've ever ignored a file in the past, then branch switching could leave the file around to still be ignored); but I won't complain too much if you make the deletion (I can always re-add the lines in my .git/info/exclude for my own environment).
+++ b/Makefile.am
@@ -74,8 +73,6 @@ check-local: all tests
tests: @(cd docs/examples ; $(MAKE) MAKEFLAGS+=--silent tests) - @(if [ "$(pythondir)" != "" ] ; then cd python ; \ - $(MAKE) MAKEFLAGS+=--silent tests ; fi)
Did configure.ac clean up the setting of $(pythondir)? /me looks some more Wow - we never set $(pythondir) anywhere pre-patch, so it only ever did something if you invoked 'make pythondir=...', and since autobuild wasn't doing it, I think this was dead code.
+++ b/autobuild.sh @@ -86,8 +86,7 @@ if test -x /usr/bin/i686-w64-mingw32-gcc ; then --prefix="$AUTOBUILD_INSTALL_ROOT/i686-w64-mingw32/sys-root/mingw" \ --enable-expensive-tests \ --enable-werror \ - --without-libvirtd \ - --without-python + --without-libvirtd
Hmm. Was --without-python disabling _just_ the python bindings, or was it _also_ disabling our attempts to use $(PYTHON) for building things in our docs/ subtree? We may STILL want to keep this setting for controlling whether to attempt to build generated docs (vs. using the pre-generated version shipped in the tarball).
+++ b/configure.ac @@ -2012,83 +2012,6 @@ fi AM_CONDITIONAL([WITH_HYPERV], [test "$with_hyperv" = "yes"])
- AC_MSG_NOTICE(Found python in $with_python/bin/python) - PYTHON="$with_python/bin/python" - with_python=yes - else - if test -x "$with_python" - then - AC_MSG_NOTICE(Found python in $with_python) - PYTHON="$with_python" - with_python=yes - else - if test -x "$PYTHON" - then - AC_MSG_NOTICE(Found python in environment PYTHON=$PYTHON) - with_python=yes - fi - fi - fi - - if test "$with_python" = "yes" ; then - AM_PATH_PYTHON(,, [:])
Ouch. Deleting this may break uses of $(PYTHON) when building our docs.
- - if test "$PYTHON" != : ; then - PYTHON_CONFIG="$PYTHON-config"
Meanwhile, this part is probably safe to drop, which goes back to my question on getting rid of the check for python-config in bootstrap.conf.
- fi - else - AC_MSG_ERROR([You must install python to build Python bindings]) - fi - else - AC_MSG_NOTICE([Could not find python in $with_python, disabling bindings]) - with_python=no - fi -fi
So it looks like pre-patch, --without-python attempted to disable only the bindings, but may have also accidentally disabled the setting of $(PYTHON)...
-AM_CONDITIONAL([WITH_PYTHON], [test "$with_python" = "yes"]) -AC_SUBST([PYTHON_VERSION]) -AC_SUBST([PYTHON_INCLUDES]) - dnl Allow perl overrides AC_PATH_PROG([PERL], [perl])
...maybe all we need is a simple AC_PATH_PROG([PYTHON], [python]) for the sake of still using $(PYTHON) during doc generation? And some of this goes back to whether deleting --without-python from autobuild makes sense.
--- a/examples/domain-events/events-python/event-test.py
Giant snip :)
+++ b/libvirt.spec.in @@ -126,7 +126,6 @@ %define with_libssh2 0%{!?_without_libssh2:0}
# Non-server/HV driver defaults which are always enabled -%define with_python 0%{!?_without_python:1} %define with_sasl 0%{!?_without_sasl:1}
@@ -425,7 +424,6 @@ BuildRequires: gettext-devel BuildRequires: libtool BuildRequires: /usr/bin/pod2man %endif -BuildRequires: python-devel
Ouch. Don't we still need a buildreq on 'python' for doc generation purposes, possibly conditionally based on whether we have patches applied to a downstream rpm that warrant doc regeneration?
-%if %{with_python} -%package python -Summary: Python bindings for the libvirt library -Group: Development/Libraries -Requires: %{name}-client = %{version}-%{release}
All of this transferred over to the libvirt-python.git, so good to see it go here.
-%if ! %{with_python} - %define _without_python --without-python -%endif - %if ! %{with_libvirtd} %define _without_libvirtd --without-libvirtd %endif @@ -1378,7 +1359,6 @@ of recent versions of Linux (and other OSes). %{?_without_sasl} \ %{?_without_avahi} \ %{?_without_polkit} \ - %{?_without_python} \
Same question as in autobuild on whether --without-python still matters for suppressing doc regeneration, or whether it was intended only for suppressing bindings.
--- a/python/Makefile.am +++ /dev/null
Another giant snip :)
+++ b/run.in @@ -58,15 +58,6 @@ export LIBVIRT_LOCK_MANAGER_PLUGIN_DIR="$b/src/.libs" export VIRTLOCKD_PATH="$b/src/virtlockd" export LIBVIRTD_PATH="$b/daemon/libvirtd"
-# For Python. -export PYTHON=@PYTHON@ -if [ -z "$PYTHONPATH" ]; then - PYTHONPATH="$b/python:$b/python/.libs" -else - PYTHONPATH="$b/python:$b/python/.libs:$PYTHONPATH" -fi -export PYTHONPATH
Fine. If you can answer the comments I raised above, and if everyone is okay with your libvirt-python.git patches, then I'm okay with this patch. Let's get it polished and in before freeze! I also just kicked off a 'make distcheck' and will report back later if I saw anything fishy that needs fixing. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org