[gnutls-dev] Where to get OpenCDK 0.6.5

Andreas Metzler ametzler at downhill.at.eu.org
Thu Nov 1 10:34:56 CET 2007


On 2007-10-28 Andreas Metzler <ametzler at downhill.at.eu.org> wrote:
> On 2007-10-28 Simon Josefsson <simon at josefsson.org> wrote:
> > Andreas Metzler <ametzler at downhill.at.eu.org> writes:
> [...]
> > > According to libtool's manual this means that 2.0.1 supports
> > > interfaces 13, 14, 15, ..., 21 and that 2.1.5 only supports interface
> > > 14. This does not reflect the actual interface. However AFAICT it is
> > > going to work perfectly on Linux, since libtool cannot express this
> > > correctly just in a soname.
> [...]
> > Is there _any_ system on which libtool can express that semantic?

> That is the 100$ question. ;-)

> > I also think there is something broken in how this mapping works.

> Perhaps we wil get an answer from libtool upstream,
> <http://news.gmane.org/find-root.php?message_id=%3ch5udv4%2d3d9.ln1%40argenau.downhill.at.eu.org%3e>

Ralf Wildenhues has answered:
| * Andreas Metzler wrote on Sun, Oct 28, 2007 at 09:34:25AM CET:
| > 
| > [...] However, the way libtool
| > tries to represent this information in sonames (on Linux) is rather
| > strange, it goes straight from libgnutls.so.13 to libgnutls.so.23. Is
| > this huge jump just bug or is there a reason for it?
| 
| The reason for it is simplicity in the version number calculation.
| There is no requirement for version numbers to be consecutive except
| for "consecutive looks nicer" and the finite number space.  I assume
| its pressure to be far lower than the incompatible change a different
| version number calculation would make.
| 
| The way libtool computes them, it even causes different major version
| numbers on different systems, so the above jump is even specific to the
| linux version_type and not all types.

and in a later mail:
| Date: 2007-10-28 12:37:29 GMT (3 days, 20 hours and 49 minutes ago)
| 
| * Andreas Metzler wrote on Sun, Oct 28, 2007 at 01:28:54PM CET:
| > 
| > Are there libtool-supported archs whose versioning will break if we do
| > this
| > num   c       r       a
| > 2.1.1 22      1       9
| > 2.1.2 14      0       0
| > 
| > instead of
| > num   c       r       a
| > 2.1.1 22      1       9
| > 2.1.2 23      0       0
| > 
| > (I know it would work on linux, resulting in libgnutls.so.14.)
| 
| I think on FreeBSD it will cause a backward jump from libgnutls.so.22 to
| libgnutls.so.14.

> > Anyway, it seems we really should use c=25 then, to avoid any
> > possibility of problems on some systems?  We can make the change in
> > 2.1.5.

Looks like this is the way to go.

cu andreas
-- 
`What a good friend you are to him, Dr. Maturin. His other friends are
so grateful to you.'
`I sew his ears on from time to time, sure'



More information about the Gnutls-dev mailing list