[OE-core] Lots of [host-user-contaminated] warning

[Max Krummenacher]

Am Sonntag, den 17.01.2016, 19:55 -0700 schrieb Christopher Larson:
> On Fri, Nov 20, 2015 at 4:02 AM, Max Krummenacher <
> max.oss.09 at gmail.com>
> wrote:
> 
> > Am Dienstag, den 17.11.2015, 09:46 +0100 schrieb Andreas Müller:
> > > Hi,
> > > 
> > > this bugs me for a while and I wonder if I am again the only one
> > > facing this: I get floods of warnings for sourcecode files as
> > > 
> > > WARNING: QA Issue: qtserialport:
> > > /qtserialport-dbg/usr/src/debug/qtserialport/5.5.99+5.6.0-alpha1
> > +gitAUTOINC+2575d33fab-r0/git/src/serialport/qserialport_p.h
> > > is owned by uid 1000, which is the same as the user running
> > > bitbake.
> > > This may be due to host contamination [host-user-contaminated]
> > > WARNING: QA Issue: qtx11extras:
> > > /qtx11extras-dbg/usr/src/debug/qtx11extras/5.5.99+5.6.0-alpha1
> > +gitAUTOINC+d64ee96f0d-r0/git/src/x11extras/qx11info_x11.cpp
> > > is owned by uid 1000, which is the same as the user running
> > > bitbake.
> > > This may be due to host contamination [host-user-contaminated]
> > > WARNING: QA Issue: qtxmlpatterns:
> > > /qtxmlpatterns-dbg/usr/src/debug/qtxmlpatterns/5.5.99+5.6.0
> > > -alpha1
> > +gitAUTOINC+94136d4280
> > -r0/git/src/xmlpatterns/type/qbuiltintypes.cpp
> > > is owned by uid 1000, which is the same as the user running
> > > bitbake.
> > > This may be due to host contamination [host-user-contaminated]
> > > 
> > > I checked PACKAGE_DEBUG_SPLIT_STYLE is not set globally - few
> > > recipes
> > > set PACKAGE_DEBUG_SPLIT_STYLE to debug-without-src.
> > > 
> > > Help appreciated
> > > 
> > > Andreas
> > 
> > I see this also in my builds. The warnings clutter the build output
> > in a
> > way that makes the warnings mostly useless. Who will spot the one
> > important new warning in the hundreds of host-user-contaminated
> > ones?
> > A fresh build of core-image-minimal today gave me 54 of these
> > warnings.
> > All but one (glibc-locale) complaining on source files.
> > 
> > I checked a few recipes and most of them were not even using a
> > custom
> > do_install but autotools generated 'make install' with something
> > like
> > 'cp -p src/* dest/' or so.
> > 
> 
> Yes, this is the most common case. Usage of cp -a or cp -p means the
> build
> ownership is retained, and no chown is run to fix it.
> 
> 
> > Even if I know how to do it I think we should not patch upstream
> > sources
> > to suppress a warning introduced by the downstream build system
> > with
> > loads of false positives and then have to maintain the patches as
> > upstream is progressing.
> > 
> 
> Either we're calling the upstream buildsystem and not running cp
> directly,
> in which case it's a bug in the upstream buildsystem, since make
> install
> will result in non-deterministic behavior in the ownership of the
> installed
> files, or it's a cp -a case, which doesn't require altering upstream
> with a
> patch at all, only recipe modification.
> 
> 
> > I haven't looked into how one does it but I'm inclined to switch of
> > the
> > warning globally, e.g. in local.conf.
> 
> 
> Blindly disabling them will just result in ignoring a problem. Files
> in the
> target filesystem are owned by your build user, who most likely
> doesn't
> even exist on target.

Don't you get me wrong, fixing the issues are a good plan. My point is
that currently there are so many [host-user-contaminated] warnings,
more or less all with -dbg packages, that I fear to overlook a warning
which is more likely to indicate a bug that is going haunt me.

Regards
Max