On Fri, 2006-10-13 at 14:34 -0400, Matthew Miller wrote: > On Fri, Oct 13, 2006 at 02:35:11PM -0400, seth vidal wrote: > > The thing I can't figure out is that at the point where the transaction > > occurs we've handed off all the control to rpmlib. > > So it's likely an rpmlib bug? I don't know. It _could be_ but I can't say that with any amount of certainty. > > I can think of some instrumentation we could add to find out what the > > tsInfo object looks like before the transaction if you can get it to > > reoccur. > > I can't get it to reoccur on purpose. But, if the instrumentation can be > added in a semi-transparent manner, I'm willing to put it on all our systems > here so that if/when it happens next we'll have that result. I can probably give you a patch which writes out a bunch of debug data to a file. after it happens I'd just need that file. -sv
