We're probably misunderstading something.
Does the "O => i" change fix JAMES-512 ?
Or we need anyway the full removal of the cache??
I understood that we need the removal of the cache, but maybe I didn't
understand your tests.
I was under the impression that the leak is only fixed removing the 2
hashmaps usage.
Btw I don't care about the discussion issue: I think I already helped
giving my support to this bug.
Now you can discuss how much you like ;-)
I will reply when I think I can add technical value to the discussion
given my knowledge of the codebase.
If you want to write unit tests for this bug I suggest to start from
testing racing conditions between dispose and remove:
String key = mail.getName();
mail.dispose();
repository.remove(mail.getName());
And viceversa.
Imho the caches could have been written for something related to this usage.
I'm sorry for my "too radical diagnosis": I will try to keep my comments
much more on a technical level. Btw I commented that because I
understood that you were testing the "patched" version and I didn't
understood why you was trying to fix the code that was just been
removed. Btw I understood you're trying different ways.
From a technical point I don't think that we should keep a cache if it
is never useful/used in James codebase just to avoid destabilisation.
Stefano
Bernd Fondermann wrote:
Stefano Bagnara wrote:
If you look in the svn history of that file you will see that there
was much worst bug about this. Maybe this was neven called anyway
because the streams were always correctly closed by James.
In fact, currently it gets called. I checked with a debugger and we
notice it from the change in behavior as we change the code.
Imho, the fact that it is (and has been) so buggy is a +1 to remove it
and eventually investigate bugs introduced removing that outdated code
instead of loosing much more time fixing that code itself.
Again, this is a much too radical diagnosis for me (although in the end
it might happen that the whole stuff gets rewritten. but it is not
obvious to me at this point in time). But we know this discussion
already... ;-)
I don't understand this whole reasoning behind 'loosing time',
'discussing too much', 'risk'.
I'd rather understand what's really happening, writing test and all this
old-fashioned stuff and resolve the problem step-by-step.
Bernd
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
