[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <533ADC3A.6090800@oracle.com>
Date: Tue, 01 Apr 2014 11:33:14 -0400
From: Sasha Levin <sasha.levin@...cle.com>
To: Josh Boyer <jwboyer@...oraproject.org>
CC: "Linux-Kernel@...r. Kernel. Org" <linux-kernel@...r.kernel.org>
Subject: Re: liblockdep soname versioning
On 04/01/2014 09:35 AM, Sasha Levin wrote:
> On 04/01/2014 09:28 AM, Josh Boyer wrote:
>> On Tue, Apr 1, 2014 at 9:19 AM, Sasha Levin <sasha.levin@...cle.com> wrote:
>>> On 04/01/2014 08:56 AM, Josh Boyer wrote:
>>>>
>>>> Hi Sasha,
>>>>
>>>> We've had a request [1] to package up liblockdep in Fedora. Looking
>>>> things over, I noticed the library isn't actually versioned at all and
>>>> instead just builds a plain .so file. That's likely fine during
>>>> development of it, but if distros are to ship it for broader use then
>>>> it would be a good idea to specify the soname and use a versioned .so.
>>>>
>>>> The makefile already has LIBLOCKDEP_VERSION defined. Would it be
>>>> possible to use this as the soname and version number? Then
>>>> liblockdep.so could be the normal symlink to the versioned .so
>>>> (liblockdep.so.0.0.1 in this case).
>>>>
>>>> Thanks.
>>>>
>>>> josh
>>>>
>>>> [1] https://bugzilla.redhat.com/show_bug.cgi?id=1082763
>>>>
>>>
>>> Sure! I never expected it to live outside the kernel tree as a separate
>>> package, but I'm happy to accommodate for that.
>>>
>>> I think that I'll just match the version number with the kernel version
>>> since what mostly matters is what you have in kernel/lockdep.c, so for
>>> example, right now we'll have 'liblockdep.so.3.15.0'. Sounds good?
>>
>> The only concern I would have is that it would require applications
>> linking to it to rebuild with every kernel release even if nothing
>> else changed. Maybe nothing changing is going to be rare enough that
>> in practice people will need to rebuild anyway. Either way, it's
>> better to be explicit rather than break users silently, so it sounds
>> good to me.
>
> I don't think we ever had a kernel version without changes to lockdep :)
>
> Since lockdep isn't an ABI either, no one promises me it'll work the same
> way between versions either, so I'm kinda happy about just forcing rebuilds.
Hi Josh,
Could you please confirm that the below is what you'd expect it to be:
sasha@...py:~/linux/tools/lib/lockdep$ make
CC FPIC common.o
CC FPIC lockdep.o
CC FPIC preload.o
CC FPIC rbtree.o
BUILD STATIC LIB liblockdep.a
BUILD SHARED LIB liblockdep.so.3.14.0
sasha@...py:~/linux/tools/lib/lockdep$ ls -al liblockdep.so
lrwxrwxrwx 1 sasha sasha 20 Apr 1 11:31 liblockdep.so -> liblockdep.so.3.14.0
sasha@...py:~/linux/tools/lib/lockdep$ readelf -d liblockdep.so | grep SONAME
0x000000000000000e (SONAME) Library soname: ["liblockdep.so.3.14.0"]
Thanks,
Sasha
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists