On Tue, Oct 24, 2017 at 05:16:23PM +0200, Michael Haggerty wrote:
> But in fact the problem is more general; it can happen whenever a
> reference deletion within a transaction is processed successfully, and
> then another reference update in the same transaction fails in
> `lock_ref_for_update()`.
In [1], Peff described a bug that he found that could cause a
reference transaction to be partly carried-out and partly not, with a
successful return. The scenario that he discussed was the deletion of
one reference and creation of another, where the two references D/F
conflict with each other.
Bu
2 matches
Mail list logo
