boost problem while packaging SuperCollider 3.8

[Dan S]

2017-09-25 16:36 GMT+01:00 Felipe Sateler <fsateler at debian.org>:
> On Mon, Sep 25, 2017 at 12:22 PM, Felipe Sateler <fsateler at debian.org> wrote:
>> On Mon, Sep 25, 2017 at 11:14 AM, Dan S <danstowell+debmm at gmail.com> wrote:
>>>
>>> 2017-09-25 13:48 GMT+01:00 Felipe Sateler <fsateler at debian.org>:
>>> >
>>> >
>>> > On Mon, Sep 25, 2017 at 7:37 AM, Dan S <danstowell+debmm at gmail.com> wrote:
>>> >>
>>> >> Hi
>>> >>
>>> >> I'm trying to get the new release 3.8 of SuperCollider into the
>>> >> debian-multimedia repository. I've done an import to git here (not
>>> >> pushed yet). The problem is SC's dependency on boost.
>>> >
>>> >
>>> > Please push, so that I can try to reproduce locally.
>>>
>>> OK, done.
>>
>> Oh I'm so terribly sorry. I see now that I forgot to push the last
>> upload, which fixes the FTBFS with gcc 7. I'm now merging your changes
>> with mine to see if things build. I believe this should fix at least
>> some problems with newer gcc.
>
> Indeed, the patches Adrian Bunk sent fix the failure to build. I have
> force pushed the changes to git (so please use git reset instead of a
> simple pull). Some are already upstream but it appears the
> PyrSched.cpp one is not yet.
>
> I'm sorry for causing additional work :(

OK, thanks.

On the latest master then:
* I refreshed patches (because "gbp -S" failing complaining about fuzzy patches)
* Build failed on gcc 7 due to an ICE while compiling some part of supernova
* I'm running a gcc 6 build today, I'll let you know if it completes.


>>> >> The upstream includes boost in the download, and so to be Debian-like,
>>> >> we strip that out and use system boost instead. This works in prev
>>> >> releases, but in 3.8 (which has updated to use boost 1.63) it errors
>>> >> out with confusing compiler/linker problems:
>>> >> https://github.com/supercollider/supercollider/issues/3203
>>> >>
>>> >
>>> > I notice you are using wheezy. I'd first try building in sid. BTW, where did
>>> > you get boost 1.63? Unstable has 1.62...
>>>
>>> "libboost1.63-dev" etc packages
>>
>> OK, why did you bump these deps? I don't see any specific requirement
>> for 1.63 (at least with a quick grep through the sources).
>>
>
> I did not keep this commit in the forced push (but have kept locally
> in case it is still needed), because SC built fine without it.

The reason for the dependency is that in the older boost I was using
last week (1.58?) there was a straightforward build failure because a
boost function's arguments have changed across versions. So there
should be SOME version bump in the dependency but I don't know what
the minimum is. I went for 1.63 because it matches the bundled SC
version and was available.

This has now taken a lot of my time and I'm afraid I cannot do much
more (work commitments). If I had time I could trace that dependency
issue, report the ICE, test that SC runs...! But I must stop, sorry.

Dan