Hi Tim,

On Sun, 28 Jun 2020, Tim Johnson wrote:

> Hi all,
>
> Since applying this fix to my server image, it has stayed running and 
> available for far longer (~17 days) than it ever had previously (2-6 
> days).  I, for one, recommend inclusion in Trunk.

Done.

>
> A more specific/holistic evaluation of the server image after this fix:
>
> (Process allSubInstances select: [:ea | ea name includesSubstring: 
> 'WebServer']) size    " => 2 "

Do you have 2 WebServer instances as well?

>
> Process allSubInstances select: [:proc |
> 	proc suspendedContext
> 		ifNil: [ false ]
> 		ifNotNil: [:context |
> 			context closure
> 				ifNil: [ false ]
> 				ifNotNil: [:closure | | receiver |
> 					receiver := closure outerContext 
> receiver.
> 					(receiver respondsTo: #outerContext)
> 						ifTrue: [ receiver 
> outerContext receiver class == WebServer ]
> 						ifFalse: [ false ] ] ] ] 
> " size => 4"
>
> WebServer allInstances first connections size   " => 3"

That should be 0 unless there are debuggers open or there are actual 
connections to your server.
Perhaps the Seaside error handler does something that doesn't let 
connections to be removed.


Levente

>
> (Process allSubInstances select: [:p | p isTerminated]) size " => 10"
>
> WASession allSubInstances size "=> 1"
>
> And leftover Sockets seem healthier than they did before this fix:
>
> (Socket allInstances collect: [:ea | ea statusString asSymbol]) 
> asBag   "=> a Dictionary(#connected->6 #destroyed->1 
> #invalidSocketHandle->60 #otherEndClosedButNotThisEnd->6 
> #waitingForConnection->2 )"
>
> Maybe WebServer listener sub-processes are still not getting GCed 
> until I click somewhere in the image, which is strange (?), but at 
> least they're /able/ to go away now.  :)
>
> Thanks,
> a Tim
>
>