clone 471250 -1
reassign -1 g++
thanks
On Sun, Mar 16, 2008 at 11:30:15PM +0100, Kurt Pruenner wrote:
> Package: maildrop
> Version: 2.0.4-3
> Severity: important
>
> I've found lately that whenever I upgraded to a 2.0.4 version of maildrop
> from the
> 2.0.3 one that I was using maildrop would hang forever and chew up what
> little CPU
> there is on my NSLU2 instead of filing mails.
>
> Today, I've finally broken out gdb to take a closer look at what's going on.
> I've
> found that maildrop was endlessly spinning through the loop in
> Message::Rewind in
> message.C because in the second call to Rewind the value assigned to the
> counter
> variable n, maildrop.msginfo.msgoffset, was negative (somewhere around
> -260000).
> Further debugging revealed that it was actually the inlined Message::tell()
> assigning that absurd value to msgoffset.
>
> The mailfilter file I was using consisted of a single
>
> to "./Maildir"
>
> line, and the mail file I used had just a From:, To: and Subject: header and a
> single line of data saying "Test".
>
> Since I was using an unstripped but still optimized package which made things
> hard
> to debug I recompiled maildrop (which only takes a bit over an hour on said
> hardware) with the noopt flag set, only to discover that now suddenly
> everything
> was working perfectly fine.
>
> As I really don't have the time nor the experience to hunt what seems to be an
> GCC ARM compiler bug - could you perhaps forward this compiler bug to the gcc
> maintainer team and in the meantime build maildrop on ARM without
> optimizations?
>
> Unless I'm really unlucky here I'm afraid I'd have to say that maildrop is
> currently non-functional on ARM... :(
Hm, okay, let's clone the bug report to g++...
--
2. That which causes joy or happiness.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
