Thanks for the analysis Gregor,
  It explains why when I built it by hand it seemed to work ok. There was a
patch to fix parallel build where something needed to depend on something
else. I suspect that there needs to be another (or more) of these
dependencies.
Odd how if it doesn't find the library it just doesn't abort instead of
merrily making a bad module, but there you go.

Probably also explains why the reproducible build check always fails.

It gives me a place to look, thanks!


 - Craig


On Tue, 15 Oct 2019 at 06:07, gregor herrmann <gre...@debian.org> wrote:

> On Mon, 14 Oct 2019 10:34:41 +1100, Craig Small wrote:
>
> > > libsnmp-perl is broken.
> > Ouch, I don't use the module (or Perl much for that matter) but that's
> very
> > broken. No idea what's going on but it worries me that
> > netsnmp_ds_get_boolean is the first function in that module which means a
> > coincidence or all functions are not available.
>
> Some notes:
>
>     % apt-cache --no-all-versions show libsnmp-perl | grep Depends
>     Depends: perl (>= 5.30.0-6), perlapi-5.30.0, libc6 (>= 2.15)
>
> So the perl package doesn't depend on any snmp libraries. The
> previous version had
>     Depends: perl (>= 5.30.0-5), perlapi-5.30.0, libc6 (>= 2.15),
> libsnmp30 (>= 5.7.3+dfsg)
>
> Digging around in the git repo shows that the perl build system has
> been changed. (No idea if this is related.)
>
> Looking at the (local) build log we see:
>
> … well different things at each build, but once I saw the perl
> modules built first in dh_auto_build with warnings about missing C
> libraries …
>
> Upstream ChangeLog says:
>
>     The Perl module build problems were caused by a concurrent build
>     (make -j<n>) and not by a problem in any of the Makefile.PL files.
>
> When I try
>
>     %:
>         dh $@ --no-parallel
>
> I end up with
>
>     Depends: perl (>= 5.30.0-6), perlapi-5.30.0, libc6 (>= 2.15),
> libsnmp35 (>= 5.8+dfsg)
>
> and the problem is gone:
>
>     % perl -MSNMP -e 1
>     %
>
> So it looks like the "whole" problem is that net-snmp is not
> parallel-build safe …
>
> Not sure if --no-parallel is an acceptable solution but at least the
> area of the problem is a bit clearer now. (Fixing the many
> Makefile.*s might be cleaner but that's too much for me now.)
>
>
> Cheers,
> gregor
>
>
> --
>  .''`.  https://info.comodo.priv.at -- Debian Developer
> https://www.debian.org
>  : :' : OpenPGP fingerprint D1E1 316E 93A7 60A8 104D  85FA BB3A 6801 8649
> AA06
>  `. `'  Member VIBE!AT & SPI Inc. -- Supporter Free Software Foundation
> Europe
>    `-
>

Reply via email to