On Sun, 16 May 2021 13:14:26 +0200 n952162 <n952...@web.de> wrote: > On 5/16/21 12:53 PM, Andreas Fink wrote: > > On Sun, 16 May 2021 12:49:26 +0200 > > n952162 <n952...@web.de> wrote: > > > >> On 5/16/21 11:28 AM, Neil Bothwick wrote: > >>> On Sun, 16 May 2021 11:26:37 +0200, n952162 wrote: > >>> > >>>>>> There are no use flags defined for any of the packages I did a random > >>>>>> check for, either on the server or the client. I am worried that it > >>>>>> is as you say: that the ebuild has a change of USE flags, which, of > >>>>>> course, has nothing to do with me, the user. > >>>>> As already stated, any USE flag changes would appear in the emerge > >>>>> output, this is most likely caused by --changed-deps. Try with > >>>>> --changed-use but without --changed-deps to see. > >>>>> > >>>>> > >>>> I have introduced that into my build script. But, if it's as you say, > >>>> the one is a subset of the other, it should have no effect on the > >>>> output, right? > >>>> > >>> --changed-use is a subset of --newuse. --changed-deps is separate. > >>> > >>> > >> Ah, I oversaw that. > >> > >> Ah. why would I want to have --changed-deps anyway? That suddenly seems > >> silly. > >> > >> It's unfortunate, if there's no explanatory display if a package got > >> disqualified for that reason. > >> > >> > > Trying to comprehend here... > > > If you want to have a binhost, then --changed-deps will become > > "necessary" at some point. Let me draw you a picture, where a binhost > > would fail to provide the correct package: > > - Binhost builds on day 1 package XYZ(i.e. server updates from internet) > > - computer that would merge with packages from binhost is NOT > > updated(client does NO emerge on that day) > > - the dependencies are changed on day 2(i.e. XYZ is emerged onto server, > > with changed dependencies in the ebuild) > > - Binhost does NOT rebuild, because you do not have --changed-deps > > enabled on day 2*(what is "Binhost" here? The --changed-deps is > > specified on the client)* > > - Computer that merges from the binhost is updated on day 2 but will > > NOT use the binary package from binhost, because the dependencies do > > not match > > There are flags to ignore dependency mismatches, but the default would > > just not use the binary package. > > > > Cheers > > Andreas > > > What does changed-deps mean, actually? > > --changed-deps [ y | n ] > Tells emerge to replace installed packages for which > the corresponding > ebuild dependencies have changed since the packages were > built. ... > > I presume it means that a package needed XYZ before, but now needs > XYZZ. If I don't specify --changed-deps, that I might get a run-time > resolution problem. Changed dependencies means any change in the *.ebuild file with respect to the variables DEPEND/BDEPEND/RDEPEND/PDEPEND, e.g. version of a dependent package has changed, new package was added as dependency, a package was removed as dependency. All are dependency changes. If the changed *.ebuild file is commited to the portage tree WITHOUT a version-bump/revision-bump, then emerge would NOT rebuild the package, unless --changed-deps is given as an argument.
> > Or, does it mean that the package specified XYZ.1 in an excess of > precision and the new version specifies XYZ.3? > > I just ran into this: > > --binpkg-changed-deps [ y | n ] > Tells emerge to ignore binary packages for which the > corresponding ebuild > dependencies have changed since the packages were built. > In order to help > avoid issues with resolving inconsistent dependencies, > this option is auto- > matically enabled unless the --usepkgonly option is > enabled. Behavior with > respect to changed build-time dependencies is controlled > by the --with-bdeps > option. > > But I haven't figured out what it means yet. In particular, what all > the stated implications mean. > This would be the option to ignore dependency mismatches of what the binary package claims its dependencies are (which you could see in $PKGDIR/Packages), and what the resolved dependencies are according to the *.ebuild file as portage is seeing it right now. Cheers Andreas
