Errors after loading websockets to Pharo 9

[jecisc]

Hi,

I have a project loading Zinc for the websockets.

The dependency is defined like this:

spec
		baseline: 'ZincHTTPComponents'
		with: [ spec
				loads: #('WebSocket');
				repository: 'github://svenvc/zinc:master/repository' ]

After loading this project I seem to get random errors in Pharo. I got one using Iceberg and I from the Process class.
Both errors were from Zinc classes.
I also have a missing dependency warning during the loading.

Sadly I closed the debuggers before copying the stack.
Did the loading of websockets got tested in Pharo 9?

[jecisc]

Hum, I see that my repository was not up to date but at the last release because I was using the release before trying with master.

This might be resolved with the latest zinc. I'll try and reopen if it is not the case. Sorry for the noise.

[jecisc]

I'm reopening because even with the latest Zinc I get errors.

Just letting my image live cause me this error:

Process(Object)>>doesNotUnderstand: #isValid
[ 
		serverSocket isValid 
			ifFalse: [
				"will trigger #ifCurtailed: block and destroy socket"
				^ self listenLoop ].
		self serveConnectionsOn: serverSocket ] in [ [ 
		serverSocket isValid 
			ifFalse: [
				"will trigger #ifCurtailed: block and destroy socket"
				^ self listenLoop ].
		self serveConnectionsOn: serverSocket ] repeat ] in ZnMultiThreadedServer>>listenLoop in Block: [ ...
FullBlockClosure(BlockClosure)>>repeat
[ [ 
		serverSocket isValid 
			ifFalse: [
				"will trigger #ifCurtailed: block and destroy socket"
				^ self listenLoop ].
		self serveConnectionsOn: serverSocket ] repeat ] in ZnMultiThreadedServer>>listenLoop in Block: [ [ ...
FullBlockClosure(BlockClosure)>>ifCurtailed:
ZnManagingMultiThreadedServer(ZnMultiThreadedServer)>>listenLoop
[ self listenLoop ] in [ [ self listenLoop ] repeat ] in ZnSingleThreadedServer>>start in Block: [ self listenLoop ]
FullBlockClosure(BlockClosure)>>repeat
[ [ self listenLoop ] repeat ] in ZnSingleThreadedServer>>start in Block: [ [ self listenLoop ] repeat ]
[ 
			  self value.
			  Processor terminateActive ] in FullBlockClosure(BlockClosure)>>newProcess in Block: [ ...

In ZnMultiThreadedServer>>#listenLoop

[svenvc]

You will have to be more specific and/or isolate the error better.

If you take a clean Pharo 9, load the latest Zinc from GitHub with the WebSocket group and run the unit tests for WebSockets, are they green or not (there could be a test that fails because a public server disappeared) ?

Are you maybe loading into an image that has live server instances ?

There are always some loading warnings because Zinc in Pharo differs too much from the latest official code, most of these are benign, but I should see them to be sure.

[SabineMa]

I have the same problem.

It occurs randomly and it regularly stops our pipelines to run. @svenvc do you have a workaround? I was already thinking about implementing isValid in object ...because it stops our processes...and I was not yet able to find the reason for it

Can also not yet report more than that I load beneath a lot of other stuff this:

spec
	baseline: 'Seaside3'
	with: [ spec
			loads: #('default');
			repository: 'github://SeasideSt/Seaside:v3.4.3/repository' ] 

spec baseline: 'Voyage' with: [ 
	spec
		loads: #( 'mongo' );
		repository: 'github://pharo-nosql/voyage:1.7/mc' ] 

spec
	baseline: 'ZincHTTPComponents'
	with: [ spec repository: 'github://svenvc/zinc:v4/repository' ]
	
		spec
	baseline: 'ZTimestamp'
	with: [ spec repository: 'github://svenvc/ztimestamp:v24/repository' ]

Pharo8, occurs on all platforms unix, windows and Mac

I wonder that at that moment a server is running because it occurs at any time when loading....

after some investigation I found this :-)

https://discord.com/channels/223421264751099906/311419918631305218/708953040794812418

and I see, that I have to change my seaside loading baseline.

Just for completeness, I will not delete this comment. So when I am investigating next year I will find this :-)