On 05/25/2011 04:10 PM, Eric Blake wrote: > On 05/15/2011 08:51 PM, Wen Congyang wrote: >> At 05/12/2011 04:30 AM, Eric Blake Write: >>> Anything generated that must end up in the tarball must either >>> have unconditional rules for generation (remote_protocol.c) or >>> must live in libvirt.git for the case where the person running >>> 'make dist' has disabled the configure options that control the >>> rebuild of the generated file (remote_protocol-structs). >>> >>> * src/Makefile.am (remote_protocol-structs): Add a dependency and >>> document why it must live in git. >>> ($(srcdir)/remote/%_protocol.c, $(srcdir)/remote/%_protocol.c): >>> Unconditionally generate. >>> --- >>> >>> This should be the last piece of the puzzle for fixing an issue >>> that I first pointed out here: >>> https://www.redhat.com/archives/libvir-list/2011-May/msg00343.html > > This patch still isn't in. Should it be?
Actually, I've going ahead and pushing it under the build-breaker rule
(after fixing a nit below), since I was able to confirm that:
mkdir build
cd build
../configure --without-remote
make distcheck
failed without this patch, because remote_protocol.h was not in the
tarball and that resulted in make trying to (re)generate the file into a
read-only directory (make distcheck is time-consuming, but it's nice for
proving issues like this). Furthermore, there was talk on IRC about the
daily snapshots not being generated lately, precisely because they are
being generated from a build configured --without-remote, so this should
help snapshot builds be more useful.
>
>>> .PHONY: remote_protocol-structs
>>> -remote_protocol-structs:
>>> +if WITH_REMOTE
>>> +remote_protocol-structs: libvirt_driver_remote_la-remote_protocol.$(OBJEXT)
>>
>> This line will introduce another problem for ./configure --with-remote:
>> make[1]: *** No rule to make target
>> `libvirt_driver_remote_la-remote_protocol.o', needed by
>> `remote_protocol-structs'. Stop.
>> make[1]: Leaving directory `/home/wency/source/libvirt-nodaemon/src'
>> make: *** [distdir] Error 1
>
> I can't reproduce that failure on Fedora 14. What system were you on?
> Perhaps it is dependent on libtool version or automake version used to
> generate the Makefile? Should I isolate this change from the rest of
> the patch, so that 'make dist' works better from a --without-remote setup?
Oh, I think I see it now. libvirt_driver_remote_la-remote_protocol.o is
a side-effect of the libtool rule for creating
libvirt_driver_remote_la-remote_protocol.lo; so while the subsequent
pdwtags call operates directly on the .o file, the Makefile dependency
has to be on the witness .lo file that signifies whether the build has
been done.
I squashed this in:
diff --git i/src/Makefile.am w/src/Makefile.am
index 82cf6c4..c3d3102 100644
--- i/src/Makefile.am
+++ w/src/Makefile.am
@@ -207,7 +207,9 @@ r2 = /\* <[[:xdigit:]]+> \S+:\d+ \*/
.PHONY: remote_protocol-structs
if WITH_REMOTE
-remote_protocol-structs: libvirt_driver_remote_la-remote_protocol.$(OBJEXT)
+# The .o file that pdwtags parses is created as a side effect of running
+# libtool; but from make's perspective we depend on the .lo file.
+remote_protocol-structs: libvirt_driver_remote_la-remote_protocol.lo
$(AM_V_GEN)if (pdwtags --help) > /dev/null 2>&1; then \
pdwtags --verbose libvirt_driver_remote_la-remote_protocol.$(OBJEXT) \
| perl -0777 -n \
--
Eric Blake [email protected] +1-801-349-2682
Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
-- libvir-list mailing list [email protected] https://www.redhat.com/mailman/listinfo/libvir-list
