Re: databases/py-sqlite3 build on RHEL 7 fails without C99

[Greg Troxel <gdt%lexort.com@localhost>]

"J. Lewis Muir" <jlmuir%imca-cat.org@localhost> writes:

>> So I'll apply Lewis's change in 24h if there are no objections.
>
> Thanks!  Although, I just saw now at
>
>   https://www.netbsd.org/docs/pkgsrc/fixes.html
>
> that it says in 21.4.1. C, C++, and Fortran:
>
>   Language variants like c++11 for USE_LANGUAGES should only be
>   specified if the package does not explicitly set -std=... when
>   compiling (i.e. the package assumes the compiler defaults to C++11
>   or some other standard). This is usually a bug in the upstream build
>   system.
>
> So, maybe I should report it to upstream?  Or for something like this
> that is not an actual patch to the upstream source code, reporting to
> upstream is not needed?  (I of course would rather not take the time to
> report it to upstream, but if that's the recommended approach, I would.)
> But then again, upstream is Python, I think, because this module is
> normally built into Python as of version 2.5, so maybe they wouldn't
> approve of this split and consider it nonstandard, thus decreasing the
> chance of them viewing a bug report from me in a good light.

I would argue that a pedantic computer scientist (I haven't seen any of
those around here :-) would agree that calling a C compiler without
--std=c99 is a bug if C99 mode is required.  However, it is bizarre for
a C compiler not to default to C99, so I would expect most people would
just say that the compiler should be fixed.

We are simply compiling most of python and some parts separately, from
the same tarball.  I suspect that if you build following the upstream
build instructions in your environment, you will see the same problem.