> And your next step is to determine how the sort orderings are different (by > looking at the log results), determine why a certain sort causes a problem,
OK, but I have absolutely no idea what you are talking about. :) More importantly, I am not sure if I have described the test results well enough. Let me try again: Succinct Problem Symptoms (to date) 1. The FK error is generated 100% of the time (not intermittently) when I tried to add a Product entity. In addition, simply logging into to my app (accessing the admin table), trigger the FK error for the Product - which was totally illogical since logging into admin does not INSERT a Product entity. 2. I restarted tomcat and did a recursive "touch" on the JSP dir each time I performed the test to presumably clear out the memory and cache. 3. The ONLY condition that results in a successful test is to use my development tomcat (on my OSX laptop), or set Cayenne logging to DEBUG on the production machine. Confusion: 1. I do not know how these symptoms might point to a "sort ordering" issue based on the available test results. 2. If the problem *NEVER* occurs while logging is set to DEBUG, then how would I check the logs as you suggest? This is why I have suggested that this problem is "insane"; This is sort of like a perverted Heisenberg Uncertainty Principle. :) Question (Based on the "DEBUG Uncertainty Principle"): 1. If the FK problem ONLY occurs with DEBUG set to FALSE then I have limited visibility into the problem. 2. If I set DEBUG to TRUE all the time, then I am concerned I am simply hiding a bug. 3. Of course, if I set DEBUG to TRUE, then it is "Miller Time" :) I have the feeling like some configuration file parameter might be corrupted. If nobody has any other ideas (give the visibility constraints), then I guess I will try to rebuild the project cayenne config files from scratch and hack it into submission. :) (Not looking forward to this option.) Thanks Joe On Feb 8, 2012, at 11:18 AM, Mike Kienenberger wrote: > Oops. Didn't finish. > > And your next step is to determine how the sort orderings are > different (by looking at the log results), determine why a certain > sort causes a problem, and then subclass the sorter to insure it sorts > in a consistently-valid manner each time. > > On Wed, Feb 8, 2012 at 11:15 AM, Mike Kienenberger <[email protected]> wrote: >> I think what Michael is saying is that the ashwood sorter isn't >> guaranteed to return the same results for the same commit operation on >> the next run after the application is restarted. So setting the >> logging wasn't what caused the change -- it was restarting the app. >> >> On Wed, Feb 8, 2012 at 11:02 AM, Joe Baldwin <[email protected]> wrote: >>> Well, if it is not "insane", then what should I do? I mean if the app only >>> works correctly when logging is set to DEBUG then I would think that >>> indicates a problems. I could run the app with DEBUG on all the time, but >>> that would just be covering up what I would assume is a serious problem. >>> What should I do next to debug this problem? >>> >>> More info: >>> Each time I update my project jar file, I do a recursive "touch" (which I >>> have found forces tomcat to recompile all the jsp's and recognize the new >>> jar file). After this, I restart tomcat. There was a problem with the >>> webhost's configuration of tomcat caching static variables from the >>> previous jar file into jsp static non-variables. I would *presume* this >>> would clear out everything, but maybe not. >>> >>> On Feb 8, 2012, at 9:14 AM, Michael Gentry wrote: >>> >>>> On Tue, Feb 7, 2012 at 6:03 PM, Joe Baldwin <[email protected]> >>>> wrote: >>>>> 2. So I found their log4j.properties file and inserted >>>>> "log4j.logger.org.apache.cayenne.access.QueryLogger=DEBUG" (I hope this >>>>> is what you were thinking of) >>>> >>>> I actually think you want INFO instead of DEBUG. >>>> >>>> >>>>> 4. Rod Serling moment: when I tried to add the Product entity - for the >>>>> first time it did not fail >>>>> - detail: it added the Product, there was no Exception >>>>> - I verified the Product by listing all of my products >>>>> >>>>> So, is this 1. Insane, or 2. totally Insane? >>>> >>>> I've seen AshwoodEntitySorter produce different insert orderings >>>> before from different starts. Since you shutdown the application, AES >>>> would have to produce a new dependency graph on startup. I'm not >>>> saying this is responsible, but it is a possibility. So: 0. Maybe not >>>> insane. :-) >>>> >>>> mrg >>>
