Hi! On Sat, 2026-02-07 at 17:02:20 -0800, Russ Allbery wrote: > Here is a proposed change incorporating this change and Guillem's > feedback. Does this look good to everyone? (I moved one footnote because > we now try to keep footnotes immediately after the paragraph they refer to > and the added footnote would otherwise be out of order.)
Great, thanks! > >From 19414e709a7dd38f35e621474c94a0931c2f6835 Mon Sep 17 00:00:00 2001 > From: Russ Allbery <[email protected]> > Date: Sat, 7 Feb 2026 16:58:35 -0800 > Subject: [PATCH] Allow *.so files to be linker scripts > > In some cases, *.so files installed by binary development packages > for libraries are linker scripts instead of symlinks. Document that, > and document the constraints imposed by ldconfig on linker scripts. > Based on wording from Aurelien Jarno and feedback from Guillem Jover. > > Closes: #1095039 > --- > policy/ch-sharedlibs.rst | 32 +++++++++++++++++++++----------- > 1 file changed, 21 insertions(+), 11 deletions(-) > > diff --git a/policy/ch-sharedlibs.rst b/policy/ch-sharedlibs.rst > index 69b79d3..cdd2e37 100644 > --- a/policy/ch-sharedlibs.rst > +++ b/policy/ch-sharedlibs.rst > @@ -247,29 +247,39 @@ at a time, librarynameapiversion-dev. Installing the > development package > must result in installation of all the development files necessary for > compiling programs against that shared library. [#]_ > > +.. [#] This wording allows the development files to be split into several > + packages, such as a separate architecture-independent > + libraryname-headers, provided that the development package depends > + on all the required additional packages. > + > In case several development versions of a library exist, you may need to > use ``dpkg``'s Conflicts mechanism (see :ref:`s-conflicts`) to ensure that > the user only installs one development version at a time (as different > development versions are likely to have the same header files in them, > which would cause a filename clash if both were unpacked). > > -The development package should contain a symlink for the associated > -shared library without a version number. For example, the libgdbm-dev > -package should include a symlink from ``/usr/lib/libgdbm.so`` to > -``libgdbm.so.3.0.0``. This symlink is needed by the linker (``ld``) when > -compiling packages, as it will only look for ``libgdbm.so`` when > -compiling dynamically. > +The development package should contain either a symlink or a linker script > +[#]_ for the associated shared library without a version number. For > +example, the ``libgdbm-dev`` package should include a symlink from > +:file:`/usr/lib/libgdbm.so` to :file:`libgdbm.so.3.0.0`, or a linker > +script that achieves the same effect. This symlink or linker script is > +needed by the linker (:command:`ld`) when compiling packages, as it will > +only look for :file:`libgdbm.so` when compiling dynamically. > + > +.. [#] Linker scripts are used when more flexible linking strategies are > + desired, such as conditionally linking with an additional library > + or falling back to a static library for some symbols. > + > +If the package installs a linker script, be aware that :command:`ldconfig` > +requires the linker script be shorter than 32 bytes, or contain either the (As a non-native speaker, I'm not sure whether "script be shorter" is missing a "to" in front of the "be"?) > +``GROUP`` directive or the string ``GNU ld script`` in a comment in the > +first 512 bytes. > > If the package provides Ada Library Information (``*.ali``) files for use > with GNAT, these files must be installed read-only (mode 0444) so that > GNAT will not attempt to recompile them. This overrides the normal file > mode requirements given in :ref:`s-permissions-owners`. > > -.. [#] This wording allows the development files to be split into several > - packages, such as a separate architecture-independent > - libraryname-headers, provided that the development package depends > - on all the required additional packages. > - > .. _s-sharedlibs-intradeps: > > Dependencies between the packages of the same library Seconded, with or without the s/desired/required/ change suggested elsethread and the perhaps missing "to". Thanks, Guillem
