Re: CVS commit: pkgsrc/databases/py-apsw

[David Brownlee <abs%netbsd.org@localhost>]

On Mon, 2 Oct 2023 at 19:47, Jonathan Perkin <jperkin%mnx.io@localhost> wrote:
>
> * On 2023-10-02 at 19:23 BST, David Brownlee wrote:
>
> >On Mon, 2 Oct 2023 at 17:27, Jonathan Perkin <jperkin%mnx.io@localhost> wrote:
> >>
> >> * On 2023-10-02 at 16:55 BST, David Brownlee wrote:
> >>
> >> >Module Name:   pkgsrc
> >> >Committed By:  abs
> >> >Date:          Mon Oct  2 15:54:46 UTC 2023
> >> >
> >> >Modified Files:
> >> >       pkgsrc/databases/py-apsw: distinfo
> >> >Added Files:
> >> >       pkgsrc/databases/py-apsw/patches: patch-src_apsw.c
> >> >
> >> >Log Message:
> >> >Fix build on systems without alloca.h
> >> >
> >> >alloca.h has been deprecated for a long time, and no modern system
> >> >should need it (and some do not even have it). While there are some
> >> >references to alloca() in the source, none are in src/apsw.c
> >>
> >> SunOS-based systems still require alloca.h for alloca(3C), please at
> >> least conditionalise this on NetBSD just in case.
> >
> >Thanks for the heads up.
> >
> >In this case the py-apsw author intends to restrict alloca() usage to
> >Windows, so it should not be an issue?
>
> Yeh hopefully, I just don't want it removed if there's a chance they
> will blindly add alloca() usa back at some point and ignore the header
> because it happens to work without on some OS.

I think my change was "safe" in that src/apsw.c included alloca.h, but
never actually called alloca(), while two other source files called
alloca() but did not include alloca.h, so if alloca.h was actually
needed, it wasn't going to be working anyway.

However the author looks to have a better long term solution, so
that's hopefully all OK :)

David