[libvirt] [PATCH 0/2] First round of out-of-tree build fixes

[Michal Privoznik]

On 11/11/19 2:42 PM, Daniel P. Berrangé wrote:
> On Mon, Nov 11, 2019 at 02:22:19PM +0100, Michal Privoznik wrote:
>> Some low hanging fruit here.
>>


>> 3) I've tried to update Gentoo live ebuild, but it failed to compile:
>>
>> Making all in src
>> make[2]: Entering directory '/var/tmp/portage/app-emulation/libvirt-9999/work/libvirt-9999_build/src'
>> ...
>> /usr/bin/perl /var/tmp/portage/app-emulation/libvirt-9999/work/libvirt-9999/src/access/genpolkit.pl < /var/tmp/portage/app-emulation/libvirt-9999/work/libvirt-9999/src/access/viraccessperm.h > access/org.libvirt.api.policy || rm -f access/org.libvirt.api.policy
>> /usr/bin/perl -w /var/tmp/portage/app-emulation/libvirt-9999/work/libvirt-9999/src/rpc/gendispatch.pl --mode=aclheader \
>>    remote REMOTE /var/tmp/portage/app-emulation/libvirt-9999/work/libvirt-9999/src/remote/remote_protocol.x \
>>    > access/viraccessapicheck.h
>> /bin/sh: access/org.libvirt.api.policy: No such file or directory
>>
>> The libvirt-9999_build/src/access dir doesn't exist (I'm surprised it
>> wasn't created automatically by autotools since src/ was). I'll have to
>> investigate further.
> 
> I hit this problem when making the RPM uses a build!=src dir
> originally.  I can't recall what I did to fix it though - it
> was something annoyingly stupid though IIRC. Like a parallel
> build happening when it shouldn't or something like that.

I've identified the issue. It's ebuild/eclass who disables 
dependency-tracking whenever possible. Gentoo developers claim it's for 
faster configure (because by default, gentoo does in-tree builds) and 
they are right. But what dependency tracking has to do with creating 
directories is beyond me.

Michal