Thanks for the report Helmut, very sorry for the delay in getting back to you.
You asked: The invocation however says "compile". Does it actually run a C compiler? I don't believe so: the "compile" step should just be converting a (textual) schema into some target source language (usually C++). The multiarch hinter says that the libcapnp-* packages should be > Multi-Arch: same. > I'm confused whether I should be adding Multi-Arch: foreign or Multi-Arch: same. :) I'm also a little unclear of the implications of adding either -- if you can give a high-level summary it'd be appreciated. I'm not particularly familiar with multi-arch/cross-compilation as it pertains to Debian at the moment. Thanks a ton for your patience -- I'm in the process of prepping 0.7.0 for release, so if this is easy enough to wrap up we should be able to get it out in the next upload. On Fri, Aug 3, 2018 at 6:45 AM Helmut Grohne <hel...@subdivi.de> wrote: > Package: capnproto > Version: 0.6.1-1 > Control: affects -1 + src:sonic-visualiser > User: helm...@debian.org > Usertags: rebootstrap > > cross building sonic-visualiser fails, because running capnp fails with > an "Exec format error". That usually indicates that the relevant package > (capnproto in this case) should be marked Multi-Arch: foreign to select > the build architecture version rather than the host architecture one. > But is such a marking actually correct? > > Unfortunately, answering that question is usually difficult as it > requires both multiarch-specific knowledge and package-specific > knowledge. Most people lack either or both. I for one lack the latter. > Can you help me understand whether the marking is correct? > > The capnproto package contains a few programs. When thinking about > Multi-Arch, one thinks about the interface of a package and these > programs certainly are the main interface. The question to ask is: Do > they behave any different when run on a different architecture? > Sometimes one can quickly assert that both input and output formats are > textual rather than binary. Is that the case for capnproto? If yes, it > seems very likely that the marking can be correct. The invocation > however says "compile". Does it actually run a C compiler? Since the > output of a C compiler is architecture-dependent, running it voids and > Multi-Arch: foreign marking. > > Please enter a discussion with me such that we can find out the correct > Multi-Arch marking for capnproto. The actual marking is a rather simple > step, once we know how it should be marked. The multiarch hinter says > that the libcapnp-* packages should be Multi-Arch: same. Please consider > adding that. If in doubt, please ask me. > > Thanks for your assistance > > Helmut > -- *Tom Lee */ http://tomlee.co / @tglee <http://twitter.com/tglee>
