Hello Dmitry, Gianfranco,

I did some research and testing into this bug...

On Sat, Apr 14 2018, kact...@gnu.org wrote:

> Hello. I am a bit lost about state of this RFS, but it seems I did
> stupid thing with format=1.0; complicating sponsoring.
> Let me settle things, provide sane package with format=3.0 (quilt), with
> pristine tar. I will ping once I am ready. Sorry for noise.

That is not the problem.  The orig.tar is already in the archive so
no-one should need pristine-tar.  I can obtain something that works like

    % git clone salsa.debian.org:iu-guest/inotify-tools
    % cd inotify-tools
    % origtargz
    % sbuild            # or similar

Gianfranco, please try that :)

Now as to the FTBFS I was seeing, I investigated further.
dh_auto_configure works fine, but dpkg-buildpackage does not:

    configure:3383: ./conftest
    ==28639==ASan runtime does not come first in initial library list; you 
should either link runtime to your application or manually preload it with 
    configure:3387: $? = 1

I found a bug report[1] which says that this error is bogus when
triggered by cowbuilder or fakeroot.  dpkg-buildpackage doesn't use
fakeroot when it invokes dh_auto_configure, though ...

I would suggest just disabling address sanitising for now.  My sbuild
setup is v. similar to the buildds, so I suspect we are going to see an
FTBFS if this is left turned on.

[1]  https://github.com/google/sanitizers/issues/786

Sean Whitton

Attachment: signature.asc
Description: PGP signature

Reply via email to