Re: Failed assertion in buildworld on armv7

From: bob prohaska <fbsd_at_www.zefox.net>
Date: Fri, 12 Sep 2025 13:37:10 UTC 
On Thu, Sep 11, 2025 at 04:35:12PM -0600, Warner Losh wrote:
> On Thu, Sep 11, 2025 at 4:09 PM bob prohaska <fbsd@www.zefox.net> wrote:
> 
> > On Thu, Sep 11, 2025 at 10:52:44AM -0600, Warner Losh wrote:
> > > On Thu, Sep 11, 2025 at 10:50 AM bob prohaska <fbsd@www.zefox.net>
> > wrote:
> > >
> > > > A Pi2 (armv7) self-hosting -current is failing buildworld with
> > > > ...
> > > > Building
> > > >
> > /usr/obj/usr/src/arm.armv7/lib/clang/libllvm/Target/ARM/ARMMachineFunctionInfo.pico
> > > > <jemalloc>:
> > > > /usr/src/contrib/jemalloc/include/jemalloc/internal/ehooks.h:184:
> > Failed
> > > > assertion: "p[i * (size / sizeof(size_t) / nchecks)] == 0"
> > > > ...
> > > >
> > > >
> > > >
> > >
> > > I committed this weeks ago, and did a armv7 build...  Did it work when
> > you
> > > restarted?
> > >
> > > Warner
> >
> > It's past the previous stopping point. I'll make a noise if it fails.
> >
> > Thanks for writing!
> >
> 
> Yea, the cross build worked for me earlier today.
> 
Buildworld stopped again last night, with:

....
Building /usr/obj/usr/src/arm.armv7/lib/clang/libllvm/Target/ARM/ARMLegalizerInf
o.pico
<jemalloc>: /usr/src/contrib/jemalloc/include/jemalloc/internal/ehooks.h:170: Fa
iled assertion: "p[i] == 0"
PLEASE submit a bug report to https://bugs.freebsd.org/submit/ and include the c
rash backtrace, preprocessed source, and associated run script.
Stack dump:
....

The offending assertion looks like a simplified version of the previous
version.

I've used
rm  /usr/src/contrib/jemalloc/include/jemalloc/internal/ehooks.h
and 
git restore  /usr/src/contrib/jemalloc/include/jemalloc/internal/ehooks.h
to see if that cleans things up. Buildworld has been restarted using
WITH_META_MODE.

If there are better things to try please indicate.

Thanks for writing!

bob prohaska