Re: [gentoo-user] ebuild log suggests to remove old libraries

[Michael Mol]

On Mon, May 21, 2012 at 2:16 PM, Thanasis <thanasis@××××××××××.org> wrote:
> on 05/21/2012 08:49 PM Michael Mol wrote the following:
>> On Mon, May 21, 2012 at 1:28 PM, Thanasis <thanasis@××××××××××.org> wrote:
>>> After a recent update of dev-libs/libffi from version 3.0.10 to 3.0.11
>>> the ebuild log suggests to run:
>>>
>>>  # revdep-rebuild --library '/usr/lib64/libffi.so.5'
>>>
>>> and once finished running revdep-rebuild, it should be safe to
>>> delete the old libraries, like so:
>>>
>>>  # rm '/usr/lib64/libffi.so.5'
>>>
>>> However by querying:
>>>
>>> equery b /usr/lib/libffi.so.5
>>>  * Searching for /usr/lib/libffi.so.5 ...
>>> dev-libs/libffi-3.0.11 (/usr/lib64/libffi.so.5)
>>>
>>> we see that /usr/lib64/libffi.so.5 is reported as belonging to
>>> dev-libs/libffi-3.0.11. Is that normal?
>>
>> I think so. It might be clearer if equery omitted the version number,
>> or if it tracked which versions of a package a file belonged to.
>
> So, are you saying that libffi.so.5 does *not* actually belong to
> dev-libs/libffi-3.0.11 ?

I would be shocked if it were generated by that absolute atom, if
that's what you mean. I think it's valid to expect it was generated by
an old version of that package.

If anything, it's probably most precise to say that libffi.so.5
belongs to dev-libs/libffi, but not to any version either in your
world file nor pulled in as a dependency by something else. (Saying
"it doesn't belong to any currently-installed version of a
currently-installed package" is ambiguous, depending on whether you
count the file's presence as meaning that the older version is
installed.)

All said, though, I've never bothered to run "equery b" on something
portage told me was an obsolete version of a library. I always run
revdep-rebuild, and then remove the old version, as the instructions
say.

-- 
:wq