[Pkg-acpi-devel] Bug#653496: Duplicated: FTBS: Debian 6.0.3 - Security Fix - acpid_2.0.7-1squeeze3

[Adam D. Barratt]

On Thu, 2011-12-29 at 14:17 -0500, John L. Males wrote:
> This issue has been duplicated exactly as reported using Knoppix
> 6.7.1 DVD.

No, what you've demonstrated is a completely different failure - with a
completely different message referencing a completely different line of
code - and one that's not unexpected.  The error from your original
report was

cc1: warnings being treated as errors
libnetlink.c: In function ‘addraw_l’:
libnetlink.c:521: error: comparison between signed and unsigned integer expressions
make[1]: *** [libnetlink.o] Error 1

From your Knoppix log:

libnetlink.c
libnetlink.c: In function 'rtnl_from_file':
libnetlink.c:444:17: error: variable 'type' set but not used [-Werror=unused-but-set-variable]
cc1: all warnings being treated as errors

How is that "exactly as reported", other than affecting the same file?

Your new issue is due to your Knoppix system containing:

ii  gcc                 4:4.6.1-2           GNU C compiler

Using different versions of toolchain / library packages is likely to
cause such issues in a number of cases.

However, as I stated in my original response, I strongly believe that
this issue is being caused by your inclusion of packages from backports
in your build environment, specifically:

pi  linux-libc-dev                    2.6.39-3~bpo60+1                  Linux support headers for userspace development

That package provides /usr/include/linux/netlink.h.  The line number
quoted in the error message references "NLMSG_ALIGN", which is provided
by netlink.h.  NLMSG_ALIGN, in turn, uses NLMSG_ALIGNTO.  I don't have
the exact version referenced above to hand to check, but certainly the
definition changed between 2.6.32 (in squeeze) and 3.1 (in wheezy):

-#define NLMSG_ALIGNTO	4
+#define NLMSG_ALIGNTO	4U

which looks to me quite likely to cause the error you're seeing.

As I said before, this problem does /not/ occur on a pure Squeeze
system, and FTBFS on mixed systems are not RC (arguably they're <
normal).  If anyone can demonstrate that such an issue occurs on a
standard Squeeze system, then that should be fixed; all the evidence so
far suggests that that does not occur, hence the "unreproducible" tag.

If nothing changes soon, I propose downgrading the severity of the bug
and letting the maintainers consider whether it's worth making any
changes to the package to accommodate it.  With my Stable Release hat on
though, I'm dubious about the merits of fixing it in stable, unless it
can be shown there's a real problem in that environment.

Regards,

Adam