Comment 11 for bug 910708

Specific to your reported reproducer:

And rpmdb is stateful and persistent: the root cause may
have occurred long before your reproducer.

And rpmdb issues come in waves and pulses when, say, someone
has a "brain fart" on some patch. With the number of patches
that Mandriva (and Per Oyvind) have added, its likely time
to start the process of integration. I.e. I haven't any clue
what issues Mandriva <-> ROSA are seeing because of the number
of patches that are not integrated, and because RPM as used by
Mandriva/ROSA is seriously different from the code I have released
with lots of changes under
     #ifdef RPM_VENDOR_MANDRIVA
patching.

There is a blueprint here to integrate and/or reject per-vendor
"Have it our own way!" patches that I (as project leader @rpm5.org)
am actually forbidden (technically) to comment on or change.

The reality of "rpm therapy" is still the same:
    Show me the backtrace and/or -vv output please.

I actually prefer valgrind because if there is one bug, there's often
others, and other root causes, that need to be considered. But
I have no time to instruct developers on the usage of valgrind while
doing "rpm therapy".