On Oct 21, 2007, at 12:05 PM, Jeff Johnson wrote:

>
> Digging, only 2311 lines in the diff, so I should have a fix  
> soonishly.
>

I've got a patched tree (with time stamp) that is functional from about
4 days ago, so I will eventually find the bug. There's been way too many
code changes to easily sort out where the flaws are.

Meanwhile, it's gonna take a few more days to finish using rpmTagData
everywhere, which should help make the bleeping problem disappear  
forever.
A pointer to a container will be passed, which will carry data to and  
from
headers.

The fundamental flaw in rpm is that headerGetEntry and headerAddEntry  
(et al)
share underlying code paths with "void *" in prototypes.

headerGetEntry actually passes a "void **", while the other methods
pass a "void *", and so there's a level of indirection on the common  
cpde paths
that gets really twisty and nasty.

Meanwhile use HEAD from last Sunday 10/21/2007.

Apologies for the hassle. For whatever reason, I cannot easily get my  
head
around pointers to unions of void pointers.

73 de Jeff