[Buildroot] [PATCH 50/50 v2] support/scripts: don't draw most of busybox' dependencies
Arnout Vandecappelle
arnout at mind.be
Sun Jul 8 11:25:47 UTC 2018
On 08-07-18 11:46, Yann E. MORIN wrote:
> Since most dependencies of busybox' are there to guarantee that busybox
> does not install applets that are already provided by other packages,
> they are not really functional dependencies, and mostly clutter the
> dependency graph.
The thing is, they still are dependencies. I often use graph-depends to do a
poor man's TLPB by starting builds of individual packages in separate shells,
but then I first have to check that their dependency chains are non-overlapping.
If you're going to hide some dependencies, that's not going to work any more...
I also sometimes use it to understand why make decides to execute things in a
particular order. Again, hiding dependencies doesn't help there.
On the other hand, it's like the toolchain: you know everything depends on the
toolchain, and similarly you know that buildroot depends on everything. So maybe
it's not that bad.
>
> Only the dependencies on libraries are interesting, so that is all we
> keep.
>
> Even though that function is only called with pkg set to 'busybox', we
> still pass it as an argument, for symetry with the other functions that
> removes the dependencies of a package.
>
> Signed-off-by: "Yann E. MORIN" <yann.morin.1998 at free.fr>
> Reviewed-by: Matt Weber <matthew.weber at rockwellcollins.com>
> Cc: Carlos Santos <casantos at datacom.com.br>
>
> ---
> Changes v1 -> v2:
> - carry tag from Matt
> - clarify linux-pam case (Carlos)
> ---
> support/scripts/graph-depends | 11 +++++++++++
> 1 file changed, 11 insertions(+)
>
> diff --git a/support/scripts/graph-depends b/support/scripts/graph-depends
> index 621e603278..0a3bc11be7 100755
> --- a/support/scripts/graph-depends
> +++ b/support/scripts/graph-depends
> @@ -181,6 +181,15 @@ def remove_mandatory_deps(pkg, deps):
> return [p for p in deps[pkg] if p not in ['toolchain', 'skeleton']]
>
>
> +# This function removes all dependencies of busybox, except its
> +# dependencies on libraries (of which libpam, from linux-pam) and
> +# host-pkgconf.
> +def remove_busybox_deps(pkg, deps):
> + busybox_dep_whitelist = ['linux-pam', 'host-pkgconf']
> + return [p for p in deps[pkg]
> + if p.startswith('lib') or p in busybox_dep_whitelist]
This feels horribly fragile to me...
I would feel more comfortable if we would introduce _INSTALL_DEPENDENCIES, as
discussed in v1 of the series, and then simply exclude the _INSTALL_DEPENDENCIES
from the graph. Or maybe even add a graph-depends option to do that.
Note: I have *not* marked this patch as changes requested, since it's actually
good to go, except that I have more fundamental concerns about it.
Regards,
Arnout
> +
> +
> # This function will check that there is no loop in the dependency chain
> # As a side effect, it builds up the dependency cache.
> def check_circular_deps(deps):
> @@ -211,6 +220,8 @@ def check_circular_deps(deps):
> # This functions trims down the dependency list of all packages.
> # It applies in sequence all the dependency-elimination methods.
> def remove_extra_deps(deps, transitive):
> + if 'busybox' in list(deps.keys()):
> + deps['busybox'] = remove_busybox_deps('busybox', deps)
> for pkg in list(deps.keys()):
> if not pkg == 'all':
> deps[pkg] = remove_mandatory_deps(pkg, deps)
>
--
Arnout Vandecappelle arnout at mind be
Senior Embedded Software Architect +32-16-286500
Essensium/Mind http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint: 7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
More information about the buildroot
mailing list