Re: [TSVN] Re: TSVNCache deadlock

SteveKing wrote:
>> Cons:
>> - It's a hell of a lot harder to find the bug!
>
> Why? It still 'hangs' for 5 seconds.

5 seconds is not a lot to fire up a debugging tool and find out where it hangs.

In fact, it takes more than 5 seconds to just enter the stack trace
view in process explorer.

I'd much prefer if the cache just hung, it makes debugging a hell of a
lot easier.

>> - TSVNCache cycles endlessly, creating new InstanceThread's and
>> killing the old ones.
>
> That's how it should work in such a case.

Except that the old ones are not always killed, and you end up with
more and more threads...

>> All in all, a very nasty change, if you ask me..
>> Increases confusion and doesn't make the cache work better ;-/.
>
> No, it doesn't.

Yes it does, otherwise I wouldn't have said so :-)

> But: many people don't know what to do if the explorer
> hangs and just reboot. And a reboot while you have apps open means data
> loss. Yes, you could kill the cache process, but how many people would
> even know how to kill a process, or even know which process is causing
> the lockup?

Hm, ok.
That's fair enough.
Still it would be nice to know that something is wrong.

Otherwise users will just think "wow, my explorer just hung for 5
seconds, but now it's okay again. It does this all the time now! Man
this TSVN is a piece of junk software", instead of "OK, my explorer is
completely dead, something is wrong. I better file a bug report."

>> Attached is test output using rev 2933.
>
> Thanks for that, but I still can't see why there's a deadlock :(

Any chance you could make this new pipe / thread / what not timeout
configurable in the registry so that I can bump it to, say, 1 billion
years or so? :-)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@tortoisesvn.tigris.org
For additional commands, e-mail: dev-help@tortoisesvn.tigris.org
Received on Wed Mar 30 08:50:41 2005