Re: 2.5.66-bk12 causes "rpm" errors

[Robert P. J. Day]

On 6 Apr 2003, Robert Love wrote:

> On Sun, 2003-04-06 at 21:32, Andrew Morton wrote:
> 
> > Does it work OK with earlier 2.5 kernels?
> > 
> > The only change which comes to mind is the below one.  Could you do a
> > patch -R of this and retest?
> 
> This has been happening since 2.5.60-ish.
> 
> It is NPTL-related.  Mr. Day, doing this:
> 
> 	LD_ASSUME_KERNEL=2.2.5 rpm <...>
> 
> should "fix" the problem.

ok, based on messing around this morning with this, here's what
i've found.

(first, apologies to andrew morton; when i said his patch applied
on top of bk12, i was just confused.  it's a "battle tactic". :-)

all of this is based on my RH 9 (shrike) box, running on a dell
inspiron 8100.

first, the rpm flaw exists using all three variations of the
kernel i tested:

  2.5.66
  2.5.66-bk12
  2.5.66-bk12-mm (bk12 minus andrew's filemap patch)

the interesting part is that doing something simple like "rpm -q rpm"
works for a non-root user; it fails only when root tries it, even
though the operation is only a query.  go figure.

next, backing out from rpm-4.2-0.69 to rpm-4.2-0.66 didn't seem to
fix the problem (at least, not for me -- a previous poster claimed
that it fixed it for him, but it didn't solve the problem here).

finally, using:

  LD_ASSUME_KERNEL=2.2.5 rpm -q rpm

solves the problem (at least under the 2.5.66-bk12-mm kernel i'm
running at the moment -- i'll assume it does the same under the
others).

take whatever you can get from this.

rday

p.s.  more 2.5.66-bk12 oddities coming up shortly