[Debian-med-packaging] Bug#973612: I'm afrait I did some kind of sabotage by upgrading civetweb ... [karsten.hilbert at gmx.net: Bug#973612: orthanc: libcivetweb version mismatch]

[Andreas Tille]

[including the list again]
Hi Sébastien,

On Mon, Nov 02, 2020 at 04:39:31PM +0100, Sébastien Jodogne wrote:
> To be fair, I am unsure to understand the issue and how to fix it...
> 
> Should I simply modify the "Build-Depends" of the orthanc package to the
> fix the version of civetweb to 1.13 (i.e. "libcivetweb-dev (= 1.13)")? Or
> set "Depends" to "libcivetweb1 = 1.13"?

I do not think so.  The packaging process will usually set versioned
depends correctly.  Moreover at least in theory orthanc should work with
both libcivetweb1 versions (1.12 and 1.13) since the soversion was not
bumped.  To verify this I added a symbols file to 1.12 first and after
checking that there are only some new symbols in 1.13 I considered it
harmless to upload (which was obviously wrong).
 
> This sounds very strange to me: I would have expected that, as the orthanc
> package was built against civetweb 1.13 (which is implied by the fact that
> ldd on "/usr/sbin/Orthanc" indicates "libcivetweb.so.1.13.0"), its
> "Depends" should have automatically been set to "libcivetweb1 >= 1.13"
> (whereas ">= 1.12" is found in the .deb package).

I confirm that the build really injects libcivetweb1 (>= 1.12).  I need
to admit that I do not know those internals but I would assume that as
long as the soversion is unchanged the smallest possible version number
is used here.  But I'm unsure.  I think if for whatever reason >= 1.13
is needed than this has to be specified explicitly.  But please note
that I never faced this before and clarifying this on debian-mentors
might be advisable.

> Please could you explain it to me?

Not better than above, sorry. :-(

Kind regards

       Andreas.
 
> On Mon, 2 Nov 2020 at 15:36, Andreas Tille <tille at debian.org> wrote:
> 
> > Hi Sébastien,
> >
> > hopefully that can easily be fixed - sorry for the inconvenience
> >
> >      Andreas.
> >

-- 
http://fam-tille.de