Bug#578458: still fails to build

Niko Tyni ntyni at debian.org
Thu Sep 9 18:47:40 UTC 2010


clone 578458 -1
retitle -1 libanyevent-perl: architecture specific constants in an arch:all package
reassign -1 libanyevent-perl 5.251-1
severity -1 grave
found -1 5.271-1
block 578458 by -1
submitter -1 !
notforwarded -1
thanks

On Mon, Aug 23, 2010 at 10:47:25AM +0300, Niko Tyni wrote:
> On Fri, Jun 18, 2010 at 03:03:38PM +0200, gregor herrmann wrote:
> > On Thu, 20 May 2010 18:05:09 +0200, gregor herrmann wrote:
> > 
> > > 5.230-1 still FTBFS on some architectures; and now in an earlier test
> > > than the one I patched to get more informations :/

> It consistently fails for me on albeniz, and so does 5.210-1 so this
> seems to affect squeeze as well.
> 
> It looks like the test broke with libanyevent-perl_5.251-1. I don't have
> the time to investigate more right now. FWIW, the current libanyevent-perl
> version does pass its own test suite on albeniz, so it doesn't seem to
> be totally broken.

Finally got round to this. The problem is this change in AnyEvent 5.25:

        - do not load Fcntl at runtime, saving memory and loading time.
        - precompile a number of constants and use them instead of runtime
          detection and eval.

This doesn't work for an Architecture:all package: in this case,
libcoro-perl broke because Fcntl::O_NONBLOCK is 2048 on amd64 (where
the package was probably built) and 4 on (at least) alpha and sparc.

Building against a modified libanyevent-perl_5.271-1 with a fixed
O_NONBLOCK value makes the test pass for me on both smetana (sparc)
and albeniz (alpha).

I think libanyevent-perl should be turned into an Architecture:any package
and at least AnyEvent/constants.pl needs to go in /usr/lib/perl5.

> BTW, I think the current debian/rules backend selection is potentially
> buggy as it's looking for substrings in the arch list, meaning "mips"
> matches "mipsel" and "i386" matches "kfreebsd-i386".
> 
> The backend selection doesn't seem to make a difference though; I tried
> all the sensible configuration options on albeniz with no luck.

IMHO the backend selection in libcoro-perl should be cleaned away for squeeze
(assuming the above fix works for all architectures.)
-- 
Niko Tyni   ntyni at debian.org





More information about the pkg-perl-maintainers mailing list