It isn't as big a change as it might appear (some of its size is simply due to duplication, which Keith has since separately refactored out on trunk).
The first change amounts to doing a single call to retrieve a time value instead of two seperate calls which might span a change in the value. The second change is resetting a value so that it is not compared in cases where it should not have been (whereby a stale value from an older operation was incorrectly being used for a comparison). The changes are fairly minor, and only really have any effect if the associated feature is enabled (which it isnt by default). They have been on trunk for a few weeks without issue in CI, and even before then were already given to the lucky users who experienced the issues prompting the change. They are not a regression from 0.18 however. Robbie On 7 January 2013 21:04, Justin Ross <[email protected]> wrote: > Hi, Keith. This one is big enough that I want to double check. Is this > still a good change for 0.20? (For example, have subsequent CI runs on > trunk been clean?) Also, is it a regression? > > Thanks, > Justin > > > On Thu, 27 Dec 2012, Keith W wrote: > > Hi Justin, >> >> Can I request the inclusion of QPID-4503 in 0.20? It is a small >> changs in a narrow area of code to address a race condition in a >> feature to detect producer transaction timeouts. >> >> http://svn.apache.org/viewvc?**view=revision&revision=1421884<http://svn.apache.org/viewvc?view=revision&revision=1421884> >> http://svn.apache.org/viewvc?**view=revision&revision=1424427<http://svn.apache.org/viewvc?view=revision&revision=1424427> >> >> Kind regards, Keith. >> >> > ------------------------------**------------------------------**--------- > To unsubscribe, e-mail: > [email protected].**org<[email protected]> > For additional commands, e-mail: [email protected] > >
