Re: Compiled results don't match for CURRENT and 7.0-RELEASE on amd64 / gdb weirdness with CURRENT

On Sun, Nov 23, 2008 at 3:08 PM, Garrett Cooper <yanefbsd_at_gmail.com> wrote:
> Hi guys,
>
>      I'm running into a weird bug where the following compiled
> results don't match (and this could potentially be a regression).
>
>      System 1: FreeBSD 8-CURRENT, synced up as of 3 days ago. Machine
> uses ULE and runs off a dual-core Intel proc.
>      System 2: FreeBSD 7.0-RELEASE, no patches, SMP enabled. Machine
> is a single-proc VMware Fusion image.
>
>      I've made a few changes to hexdump in trying to triage the bugs
> under my name, and I noticed a difference in behavior between
> 8-CURRENT and 7.0-RELEASE, in particular the fact that 8-CURRENT
> doesn't appear to assert a != NULL statement, like on the 7.0-RELEASE
> machine, but instead chokes on an earlier statement. The patch can be
> found at: <http://pastebin.com/f410b88ca>. OSX does the same thing as
> the 7.0-RELEASE machine as well, which is interesting, but there are a
> lot of points of variability with that machine, compared to the
> CURRENT box and the 7.0-RELEASE VM (compiler version, binutils
> versions, libc version, etc), and I don't want to go compare a vanilla
> Mac with FreeBSD.
>      I'm syncing up the 7.0-RELEASE's sources to 7-RELEASE and we'll
> see if that fixes anything, but I'd be really curious if anyone else
> notes this behavior with different synced versions of CURRENT.

The behavior's still the same between 7.0-RELEASE and 7.1-PRERELEASE,
so there appears to be an issue with 8-CURRENT. Suggestions for
tracking down this issue?
-Garrett