Hi! On Tue, 2025-08-12 at 11:35:21 +0200, Ferenc Wágner wrote: > What I miss here is when dpkg sets the DPKG_ROOT variable to something > non-empty. From the text I understand that dpkg sets DPKG_ROOT to the > empty string when --root and --instdir are not set (so "normally"). > Furthermore, dpkg also sets DPKG_ROOT to the empty string when --root or > --instdir is set, *unless* --force-script-chrootless is also specified > (because the maintainer scripts are run in a chroot then). So it would > seem sensible if dpkg set DPKG_ROOT to the value of --instdir (or > --root) if either of them is specified, *and* --force-script-chrootless > is also specified. But the text does not say that, instead it contrasts > DPKG_ROOT being empty with instdir being non-empty. This is unexpected, > and I don't even understand instdir in the "Internal environment" > context where this paragraph resides. > > I'm not sure I managed to express my difficulty any better now...
Ah! I think I see what you mean now. Would the attached patch fix the confusing text? (I also noticed that there's no documentation for dpkg honoring DPKG_ROOT and DPKG_ADMINDIR being set when it runs, which was a recentish behavior change, will document that too.) Thanks, Guillem
diff --git i/man/dpkg.pod w/man/dpkg.pod index 8349f0afd..001373b75 100644 --- i/man/dpkg.pod +++ w/man/dpkg.pod @@ -1438,8 +1438,8 @@ During normal operation, this variable is empty. When installing packages into a different B<instdir>, B<dpkg> normally invokes maintainer scripts using L<chroot(2)> and leaves this variable empty, but if B<--force-script-chrootless> is -specified then the L<chroot(2)> call is skipped and B<instdir> -is non-empty. +specified then the L<chroot(2)> call is skipped and +this variable is set to B<instdir> (which will be non-empty). =item B<DPKG_ADMINDIR>
