[oe] [meta-networking][PATCH] strongswan: Split plugins

David Vincent freesilicon at gmail.com
Thu Apr 27 07:45:12 UTC 2017


Sorry for the late reply, I've seen the patch has been merged into master-next 
but I let this thread unanswered. Here are some thoughts to explain it.

On mercredi 26 avril 2017 16:57:29 CEST Joe MacDonald wrote:
> [Re: [oe] [meta-networking][PATCH] strongswan: Split plugins] On 17.04.20 
(Thu 16:36) Martin Jansa wrote:
> > I think currently the preferred way is only one -dev/-dbg/-staticdev
> > package per recipe.
OK, I didn't know about this rule but I can send a v2 to remove the -dev/-
dbg/-staticdev split if it is preferred.
> > 
> > Why not set RDEPENDS_${PN} based on plugins packaged in first
> > do_split_packages call?
> > 
> > Just my 2c - I'm not using strongswan and Joe has last word for
> > meta-networking.
> 
> Yeah, I've been giving this some thought and I'm inclined to agree.  The
> RDEPENDS_${PN} being set with a canned list that doesn't reflect what
> was configured for the build seems non-intuitive.
The list comes from strongswan default configuration. I agree the RDEPENDS 
seems to come out of nowhere and for now it just looks like the same list as 
the result of the first do_split_packages call.

My main reasoning was to not break compatibility with existing use cases while 
still offering the possibility to refine it in the near future if some plugins 
are not required.
There is also the possibility to enable the compilation of all plugins with 
'--enable-all' and only install the default set. Therefore, it will be up to 
the user to select its own set through IMAGE_INSTALL (or like perl does)
One last possibility will be to use PACKAGECONFIGs for all plugins and build 
the RDEPENDS list automatically.
> 
> As for not breaking compatibility with the current configuration, if
> RDEPENDS_${PN} is set based on the build, won't that come essentially
> for free?
> 
> -J.
> 
Please, let me know if modifications are required.

--
David



More information about the Openembedded-devel mailing list