Re: [dpdk-dev] [PATCH v3 5/7] build/pkg-config: output driver libs first for static build

[Bruce Richardson <bruce.richardson@intel.com>]

On Wed, Jul 01, 2020 at 05:36:07PM +0200, Thomas Monjalon wrote:
> 01/07/2020 17:16, Bruce Richardson:
> > On Wed, Jul 01, 2020 at 04:42:33PM +0200, Thomas Monjalon wrote:
> > > 30/06/2020 16:14, Bruce Richardson:
> > > > When calling pkg-config --static --libs, pkg-config will always output the
> > > > regular libs first, and then the extra libs from libraries.private field,
> > > 
> > > s/libraries.private/Libs.private/
> > > 
> > > > since the assumption is that those are additional dependencies for building
> > > > statically that the .a files depend upon.
> > > > 
> > > > However, for DPDK, we only link the driver files for static builds, and
> > > > those need to come *before* the regular libraries. To get this result, we
> > > > need two pkgconfig files for DPDK, one for the shared libs, and a second
> > > > for the static libs and drivers, which depends upon the first. Using a
> > > > dependency means that the shared libs are printed only after the
> > > > libraries.private field rather than before.
> > > 
> > > s/libraries.private/Libs.private/
> > > 
> > > > Without this patch, the linking works in DPDK because in all cases we
> > > > specify the libraries after the drivers in the Libs.private line, ensuring
> > > > that the references to the libs from the drivers can be resolved. The
> > > > current output is therefore of the form, "(shared)libs, drivers,
> > > > (static)libs", while after this patch the output is, "drivers,
> > > > (static)libs, (shared)libs". The former case will not work if we use the
> > > > --whole-archive flag on the static libs as it will lead to duplicate
> > > > definitions due to some references having been previously resolved from the
> > > > shared libraries. By ensuring the shared libraries come last in the link
> > > > link, this issue does not occur, as duplicate references when linking the
> > > > shared libs will be ignored.
> > > > 
> > > > Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > > > Acked-by: Luca Boccassi <bluca@debian.org>
> > > > Acked-by: Sunil Pai G <sunil.pai.g@intel.com>
> > > > ---
> > > > +# When calling pkg-config --static --libs, pkg-config will always output the
> > > > +# regular libs first, and then the extra libs from libraries.private field,
> > > > +# since the assumption is that those are additional dependencies for building
> > > > +# statically that the .a files depend upon. However, for DPDK, we only link
> > > > +# the driver files for static builds, and those need to come *before* the
> > > > +# regular libraries. To get this result, we need two pkgconfig files for DPDK,
> > > > +# one for the shared libs, and a second for the static libs and drivers, which
> > > > +# depends upon the first. Using a dependency means that the shared libs are
> > > > +# printed only after the libraries.private field rather than before.
> > > 
> > > This is not obvious. In order to avoid messing it up in future,
> > > I suggest this longer reword:
> > > 
> > > # When calling pkg-config --static --libs, pkg-config will always output the
> > > # regular libs first, and then the extra libs from Libs.private field,
> > > # since the assumption is that those are additional dependencies for building
> > > # statically that the .a files depend upon. The output order of .pc fields is:
> > > #   Cflags   Libs   Libs.private   Requires   Requires.private
> > > # The fields Requires* are for package names.
> > > # The flags of the DPDK libraries must be defined in Libs* fields.
> > > # However, the DPDK drivers are linked only in static builds (Libs.private),
> > > # and those need to come *before* the regular libraries (Libs field).
> > > # This requirement is satisfied by moving the regular libs in a separate file
> > > # included in the field Requires (after Libs.private).
> > > # Another requirement is to allow linking dependencies as shared libraries,
> > > # while linking static DPDK libraries and drivers. It is satisfied by
> > > # listing the static files in Libs.private with the explicit syntax -l:libfoo.a.
> > > # As a consequence, the regular DPDK libraries are already listed as static
> > > # in the field Libs.private. The second occurences of DPDK libraries,
> > > # included from Requires and used for shared library linkage case,
> > > # are skipped in the case of static linkage thanks to the flag --as-needed.
> > > 
> > > # Link order summary:
> > > #   libdpdk.Libs.private: whole-archive(static drivers/libs), drivers deps flags
> > > #   libdpdk.Requires: libdpdk-libs package
> > > #   libdpdk-libs.Libs: as-needed(shared libs)
> > > #   libdpdk-libs.Libs.private: libs deps flags
> > > #   libdpdk.pc.Requires.private: deps packages
> > > 
> > > 
> > > If you agree, I could change this comment while merging.
> > > I would add my Signed-off ;)
> > >
> > 
> > This seems generally ok, but probably should just be added as part of patch
> > #7 when all parts of the above have been applied.
> > 
> > Couple of comments:
> > * One small nit is that cflags are not output as part of the --libs call, so
> > you can remove them from the list on line 5 of the comment. They aren't
> > really relevant to this comment/essay.
> 
> Yes, I will remove Cflags.
> 
> > * I find the link-order summary to actually be more confusing than helpful.
> > I think the text block is explanatory enough. It just confuses things
> > introducing the extra details of what goes in the requires-private or
> > libs-private of the libdpdk.pc file. That's just regular stuff, unrelated
> > to the changes or to DPDK special-case of needing private libs first.
> 
> The intent of this summary was to help navigating
> for future changes in this area.
> Personnaly it helps me, but it is maybe more a developer note
> that can be deduced from the rest.
> I can remove it.
> 
Ok thanks. If you are happy to make these comment changes on apply, please
feel free to do so, thanks.

[BTW: I think the shebang line on the new python script needs a "python"
changed to "python3" also. I missed that in the latest revision, sorry!]