On 07/25/2016 01:35 PM, Rick Moen wrote:
> Quoting Simon Hobson (linux@???):
>
>> Thanks, a bit heavy going for me at this time in the morning !
>
> Well, if you want to learn the subject, there's an irreducible minimum
> of complexity, you know, but it was mostly a citation I gave as an
> accuracy cross-check on my ultra-quick extemporaneous description.
> I.e., you needn't take my word for this bit; here's a decent write-up.
>
>> OK, that's what I thought, which is at odds with some comments that have been made.
>
> Well, if you're referring to 'comments that have been made' about
> libsystemd0, the more useful (IMO) comments characterised what is
> actually present in that library, that it contains just interface code
> that absent systemd does[1] nothing -- and the way one knows that is to
> either read the source code or rely on the characterisations of those
> who have. The fallback paranoia position then inevitably gets trotted
> out, of 'Yes, but evil nasty upstream in collusion with evil nasty
> distro packagers could _in the future_ add code that steals my lunch,
> sends my e-mails to the FSB, and opens a subspace channel to V'ger.'
>
> The same is of course possible for the contents of every other Linux
> distribution package. And the distro installers. And maybe even the
> documentation, etc.
>
> Far be it from me to recommend less paranoia, but I might make the
> modest and mild suggestion that unfocussed paranoia wastes time.
>
> Upthread, I quite seriously suggested libsystemd0 package dependency
> should have long ago been FAQed, and, fellahs, you really ought to.
> This topic should have gotten put to rest years ago, rather than
> rehashed over and over.
>
> [1] Someone disputed this characterisation by citing the GNOME gvfs
> code in XFCE4 providing or not providing 'drive icons' depending on
> whether libsystemd0 is present or not. The poster claimed this was
> libsystemd0 'doing something'. To me, it looked like GNOME gvfs 'doing
> something', and further proof of GNOME being a fragile dependency
> hairball, as if that were needed. But make up your own mind.
>

Yeah, that was me, and it was based on partially incorrect testing. I set
the permissions to 000 on the wrong target. The test with the dummy
libsystemd0 package worked great to fulfill the package dependency and
allowed me to install gvfs, but gvfs wouldn't make the drive icons.

I repeated the permissions test on the correct target with the real
libsystemd (chmod 000 /lib/i386-linux-gnu/libsystemd.so.0.3.1)
and I got the same result. If libsystemd is not readable, gvfs won't show
the drive icons.

So yes, I agree with you that it looks like it's gvfs that's doing
something, and the something it's probably doing is using code in
libsystemd. Or maybe it's telling something else to do it. Either way, it
looks like libsystemd is passively providing code for something else to
use. If the code is being used by some program, that's doing something.

Is there another interpretation of these results?

-fsr