From: <aa...@po...> - 2001-04-30 16:16:29
|
On Saturday 28 April 2001 02:13, you wrote: > > Yes, edit the code and insert a delay. I would suggest instead that = you > > build the jat file somewhere else and copy (or better yet, move) it i= nto > > place when it is done. > > I would suggest the latter... adding a delay would be the wrong thing > to do IMHO. I have experienced similar problems. I use 'scp' to copy the jar's over t= he=20 network. Every once in a while, the AutoDeployer starts reading the archi= ve=20 before everything is in place. I understand the AutoDeployer uses modified date (from the file system) t= o=20 determine whether a file has changed. If this date is set when the copy=20 procedure starts to write to disk, as it appears to do (I'm running Linux= ,=20 kernel 2.4.0/ReiserFS), there is a discomforting possibility for a race=20 condition. In my humble opinion, either the AutoDeployer, or ReiserFS is broken. If this is something that the AutoDeployer should fix, a simple delay=20 wouldn't guarantee anything, the network could be slow, just to mention o= ne=20 possible situation. One simple solution comes to mind: If ZipException causes deployment to=20 fail; wait; retry. Repeat until OK, or a number of retries. I have never done any patches to the source-code before, so unless someon= e=20 asks me, I won't give it a try.... =C5smund Hjulstad |