[Buildroot] [NEXT 01/26] cpe-info: new make target
arnout at mind.be
Thu Mar 1 20:21:46 UTC 2018
On 28-02-18 05:30, Matthew Weber wrote:
> On Tue, Feb 27, 2018 at 3:40 PM, Thomas Petazzoni
> <thomas.petazzoni at bootlin.com> wrote:
>> On Mon, 26 Feb 2018 20:10:16 -0600, Matt Weber wrote:
>>> Similar to make legal-info, produce a csv delimited
>>> file containing all selected packages CPE
>>> Signed-off-by: Matt Weber <matthew.weber at rockwellcollins.com>
>> A question is: do we need a new make target, or can an external script
>> do the same thing ?
>> After all, "make printvars" gives you pretty much the needed
Indeed. and other things that are currently in the infra could also work that
way. E.g. graph-depends would speed up dramatically if it could make use of
One limitation, however: we currently have quite a few where something like
'make foo-cpe-info' is a useful feature. In the infra, it comes for free; in a
script, it needs special treatment.
On the positive side, however, it should be possible to write a python module
that generalizes all this: gather of packages, doing printvars, filtering the
Oh, one more thing: if we go that route, then I think printvars should gain a
feature to do output a Python dictionary. Or a JSON object. Or something like
that. Now, the QUOTED_VARS is OK for shell but not for Python. And it doesn't
fully work in shell eval, because some variable names (e.g. 4TH_xxx) are not
valid shell variable names.
>> All what is missing is that you can't easily get the list
>> of selected packages in the current configuration, but that would be
>> useful for me for the pkg-stats script as well. So a "make
>> show-packages" or "make list-packages" could be useful.
PACKAGES comes pretty close - but since we don't have full Config.in.host it's
incomplete, of course. So we indeed need a recursive list-packages rule.
>> Perhaps that's how we can make our two different needs converge: by
>> having external scripts rather than adding more stuff to the package
>> infrastructure. A ./utils/cpe-report script could do pretty much what
>> you've done here.
>> Thoughts ?
> I was on the fence when I looked at where to make the change. I went
> the infra route because it ended up being really simple to generate
> that data and the whole process seamed clean.
That is indeed true: make basically has the infra already to do the recursive
expansion we need. We'd have to duplicate that into Python. But as I said, it
can be done once and then used by several scripts.
There is one reason why I would really prefer things to move to scripts: every
additional inner-generic-package variable and rule slows down make a little bit.
It's not by a huge amount, but I think I once tested that removing all those
extra rules and their variables from inner-generic-packages sped up 'make -qp'
by about 10%.
> I could see after
> creating some of the CPE maintaining scripting we could easily convert
> to your suggested approach for the report generation as well.
> I just
> didn't want the creation of the CPE scripting holding up the reporting
> function, as I can get more developers to focus on improving the CPE
> information across a few products over the next quarter if I have the
> We have been looking at the pkg-stats script and thinking about the
> impl steps to take. hoping to get more time on them next month.
> buildroot mailing list
> buildroot at busybox.net
Arnout Vandecappelle arnout at mind be
Senior Embedded Software Architect +32-16-286500
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