> On 2020-06-16, at 11:37 PM, Beckmann, Tom <Tom.Beckmann at student.hpi.uni-potsdam.de> wrote:
> 
> sorry to hear that you encountered troubles during your demo.

And apologies for sounding a bit cross. I *was* a bit cross at the time of course but that isn't a reason to be rude to you.


> 
> If I understand correctly we have already moved the blame from Metacello to a later step in the loading process, or am I mistaken? Could you send the do-its you were trying to execute so I can try to reproduce the problems here?

What happened is basically - 
I had a simple script to load Seaside/Postgresql/osprocess/cryptography
Ran it as part of a meeting meant to show off how easy it is to use Squeak for development
So many things failed
 - loading metacello left complaints in the Transcript and things in Undeclared
 - loading Seaside seemed to go ok but see later
 - loading Postgresql looked ok
 - loading Cryptography just blew up
 - attempting to show off that at least Seaside worked... didn't

So inevitably the first reaction is that the apparent problems with installing metacello were likely to have cascaded into problems correctly loading the rest. Some digging showed the inappropriate usage of the Zinc net connection stuff and that implied issues big enough to be concerned about.

After Tobias suggested that the Zinc stuff might be a red herring I did more digging in the Seaside behaviour and noticed the failure to even handle the error walkback correctly; which eventually lead to discovering that the Seaside code is using some deprecated messages and that the walkback handling code has probably not worked properly in at least a decade!

After that, the puzzle was why load OSProcess worked immediately but Cryptography failed so spectacularly; which turned out to be an error in handling mcm files by the newly installed squeaksource upgrad.

All in all a really annoying storm of multiple small issues cascading to make a perfectly infuriating and very embarrassing demo. I'm hoping to rescue the situation today.


tim
--
tim Rowledge; tim at rowledge.org; http://www.rowledge.org/tim
Long computations that yield zero are probably all for naught.