Re: Rejection of Orthanc
Hi Sébastien,
On Tue, Sep 15, 2020 at 08:22:40AM +0200, Sébastien Jodogne wrote:
> > I agree but need to work on this package structure. May be I also
> > check hor to provide a .a static lib for the -dev package.
>
> I had a look at the CMake stuff, and it visibly doesn't allow the
> simultaneous generation of the .so shared lib together with the .a
> static lib.
I previously did some patches in other libraries to be able to follow
library packaging guidelines but in this case we should probably keep
things as simple as possible. So I will not write this (usually simple)
cmake patch for this purpose.
> As a consequence, one has to build twice civetweb, once with the
> "-DBUILD_SHARED_LIBS=ON" option (to get the .so), and once without this
> option (to get the .a).
That's not worth the effort here.
> One could also consider building the C++ library by setting
> "-CIVETWEB_ENABLE_CXX=ON", which produces the "libcivetweb-cpp.(a|so)"
> C++ library together with the "libcivetweb.(a|so)" C library. Note
> however that the C++ library is not used by Orthanc.
>
> The "include/CivetServer.h" corresponds to the C++ header file, whereas
> the "include/civetweb.h" corresponds to the C header (the latter being
> the only one used by Orthanc).
I think we should do the bare minimum for this package to fulfill your
requirement for orthanc. In case someone might ask for further features
later we can add those (and at least will learn that our packaging
effort had some additional use ;-) ).
> If you prefer, I can try and write the "d/rules" given these hints, once
> you have imported the Git repository and created the skeleton of the
> package.
Please check my packaging *thoroughly*. I'm not sure whether we should
rename the lib package according to
libcivetweb1: package-name-doesnt-match-sonames libcivetweb1.11.0
No idea how many parallel installations of this lib with different
versions is to be expected.
Otherwise the package should be OKis and I issued an ITP. I'll wait
for your final confirmation before uploading.
> > Please note: I removed the Git repository since I had to rebuild the
> > tarball to get rid of several unneeded code copied of third party
> > software. I simply wanted to save space inside the Git repository. (I
> > also need to get rid of the remaining jquery.js so the current tarball
> > is not yet final but I will not recreate the repository again.)
>
> Actually, the core of civetweb is quite small wrt. to the size of the
> tarball: The "src" and "include" directories are the only mandatory
> subfolders, as long as the tests and the extensions are not built.
I simply removed everything that might make hazle license-wise - but
please review this as well. I'm in favour to strip anything that might
create extra problems. I admit I'm hesitating a bit about the tests
since I'm a fan of testing whatever can be tested - but the log says:
dh_auto_test
cd obj-x86_64-linux-gnu && make -j4 test ARGS\+=-j4
make[2]: Entering directory '/build/civetweb-1.12+dfsg/obj-x86_64-linux-gnu'
Running tests...
/usr/bin/ctest --force-new-ctest-process -j4
Test project /build/civetweb-1.12+dfsg/obj-x86_64-linux-gnu
No tests were found!!!
> > Thank you for your comments and I hope I can contribute to the
> > enhancement of orthanc by packaging this.
>
> I'm extremely grateful for your help, that clearly stimulates me to
> continue the Debian maintenance.
Exactly this was my intention! ;-) Nice to hear that it seems to
work.
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: