On 18/12/15 15:54, Didier Kryn wrote:
> Le 18/12/2015 15:35, John Hughes a écrit :
>> The list is, of course, spurious.
>>
>> $ cat /etc/debian_version
>> 8.2
>> $ apt-cache rdepends libsystemd0 | wc -l
>> 74
>

>     Sorry, my primary attitude is to believe what people write. So
> it's only 74. Does it include chained dependency?

>

Of course not. Why should it? LKCL's messages were about the enormous
effort needed to fix the 4000 odd packages that "depended" on
libsystemd0, when he only "needed' to modify 74 packages to get rid of
all libsystemd0 dependency.

>> Every single function in libsystemd0 looks like:
>>

>>  if (init_is_systemd) {
>>    do some systemd stuff;
>>  }
>>  else {
>>    carry on as before;
>>  }

>

>     But do they have also a libupstart0, libsysv0, etc, on which all
> these 74 package depend?

Nope. upstart and sysv-init don't provide any interesting features
beyond the simple init subset that packages might like to use. The
point of libsystemd0 is to let packages use systemd features without
depending on systemd.

Which makes LKCL's proposal of making libsystemd0 loaded by dlopen even
funnier -- to avoid depending on libsystemd0 packages would have to
include systemd specific code, or depend on a lib-not-systemd0 package.
But that package would also contain the magic word, so we'd need to
dlopen it, so either programs would include systemd specific code or
we'd have to invent a lib-not-not-systemd0 package, but that package
would...

Does the arrow ever hit the target? Does Achilles catch the tortoise?

(I hate Thunderbird -- resending 'cos idiot program chose wrong sender
address *again*)