Re: [gentoo-dev] Re: [gentoo-commits] repo/gentoo:master commit in: eclass/

[Anthony G. Basile]

On 4/16/16 6:36 PM, Rich Freeman wrote:
> On Sat, Apr 16, 2016 at 6:24 PM, Mike Gilbert <floppym@g.o> wrote:
>>
>> And I don't really see the point in the libressl USE flag in this
>> case; I think that was only needed so the slot-operator would resolve
>> correctly.
>>
>
> Somebody else may be better informed, but I thought that there was a
> concern with having both libressl and openssl linked into two
> libraries that are in turn linked to by something else (causing a
> symbol collision), and the USE flag was necessary to switch between
> the two implementations since they aren't virtualized.
>
> If we've solved that problem in some other way then by all means say
> so.  However, I thought this was the main tradeoff in letting it use
> the same soname.
>

I was reasoning along the lines of Mike when I first committed this.
The eclass simply provides a series of functions for generating certs.
It is sufficient that the `openssl` utility be present whether its
provided by dev-libs/openssl or dev-libs/libressl.  You're not linking
against it so the ABI issue is a non-issue.  As far as being able to
tell the difference between the two, you can use `openssl version`.

The slot operator was more subtle.  I'm aware that := only means
something in RDEPENDs but people do RDEPEND="${DEPEND} ..." often and so
as a precaution, I included it.  However, the scope of DEPEND does not
cross the eclass-ebuild boundary and since there is no RDEPEND in that
eclass, its wrong to have :0= there.  :0 is sufficient.  inherit() in
ebuild.sh scrubs the value of DEPEND, so its not as simple as just "source".

-- 
Anthony G. Basile, Ph.D.
Gentoo Linux Developer [Hardened]
E-Mail    : blueness@g.o
GnuPG FP  : 1FED FAD9 D82C 52A5 3BAB  DC79 9384 FA6E F52D 4BBA
GnuPG ID  : F52D4BBA