[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#501274: state of #501274

On Fri, Dec 12, 2008 at 09:20:39PM +0200, Eugene V. Lyubimkin wrote:
> Kurt Roeckx wrote:
> > On Fri, Dec 12, 2008 at 02:05:19PM +0200, Eugene V. Lyubimkin wrote:
> >> Thanks, I've looked into. Unfortunately, diff covers almost all the
> >> disassemble, can't do anything more.
> >>
> >> In this moment I believe that it's build problem, so I am
> >> reassigning this bug to buildd.debian.org (at least, temporarily).
> > 
> > Even if rebuilding the package fixes it, I don't see any good reason to
> > assign this bug to the buildd package.  It's more likely that this is
> > either a bug in the package or one of it's (build) dependencies.
> > 
> > If this is a powerpc specific issue, I suggest you contact the powerpc
> > list about it.
> I have no idea. CC'ing this letter to powerpc list and powerpc buildd team.

What do you want the buildd maintainers to do? If you suspect that a
rebuild will fix the issue, we can trigger a binNMU and see whether it
does; otherwise, please direct porting questions to the debian-powerpc
mailinglist only.

(it might be nice to analyze a backtrace of a segfault to figure out
where the crash comes from, though)

> > The only case I can think of that a bug might be assigned to the buildd
> > package is that one of the buildds itself is broken.  And I see no sign
> > of this.
> I tend to disagree, but I understand that I have no proofs. Marked
> 'unreproducible'.

"The buildd is broken" would mean something like "the buildd has a
corrupt /var/lib/dpkg/status file" or "the compiler was incorrectly
installed on the machine", or something similar. If the compiler is
installed correctly but generates broken code, that would mean a bug in
the compiler. This I find extremely unlikely, however; it's much more
likely that the bug is in apt, but that it is a porting issue that is
powerpc specific.

If that is the case, there are two things you should look at:
- endianness: i386 is little endian, powerpc is big endian. If your code
  makes assumptions about endianness in some places, verify that code.
- assumption about char signedness: 'char' is unsigned on powerpc, and
  signed almost everywhere else. If you have a 'char' datatype
  somewhere, verify whether the code tries to assign a negative value to
  that variable; if it does, then that's probably your bug and you
  should either change the code to not make assumptions, or change the
  declaration to say 'signed char'.

<Lo-lan-do> Home is where you have to wash the dishes.
  -- #debian-devel, Freenode, 2004-09-22

Reply to: