[squeak-dev] Installer ensureRecentMetacello is now broken

Tobias Pape Das.Linux at gmx.de
Tue Jun 16 21:42:54 UTC 2020


> On 16.06.2020, at 22:43, tim Rowledge <tim at rowledge.org> wrote:
> 
> After a deeply embarrassing attempt to demonstrate loading up an image with seaside/postgres/etc this morning I've been trying to track down the source of the error. Given that doing this just a couple of weeks ago worked perfectly, this is annoying.
> 
> The initial problem is that somebody has broken the stuff we get loaded by
> ` Installer ensureRecentMetacello.`
> 
> It now includes a bunch of code that relies upon Zinc, FileLocator and FileSystem. So far as I can tell the problems have been introduced by changed to -
> BaselineOfMetaCello (tobe.98) {last working version is Esteban Lorenzo.97}
> Metacello-MC (tobe.738) {last working version was Esteban Lorenzo.735}
> Metacello-Platform (tobe.60) {last working version topa.45}
> 
> Clearly we have a problem here in that the changes were not actually tested adequately, probably the #ensureRecentMetacello method should actually use some form of certified set of packages, and anyone in reach of 'tobe' should give them a slap on the wrist and stop their chocolate privileges for a month.


'tobe' is a very nice and cautious person and has written just some 10 days ago:

>please excuse the issues. In our recent changes to Metacello for Squeak we missed issues with the bootstrapping process for Metacello. These issues should now be resolved.

> Through this, we also got helpful input from Dale on the bootstrapping process of Metacello in Squeak and will investigate making the process faster and more resilient in the future. See here for some more details: https://github.com/squeak-smalltalk/squeak-ston/pull/1

>Best,
>Tom

To solve this without fingerpointing, lets start with some infogathering:
- did you try on a 5.3 or a trunk?
- Can you give the transcript output during the load?


:D
-topa

> I can't even see a way to fix this locally right now because the stupidity around git appears to prevent finding the last-working versions. Just why exactly is anyone enamoured of this?
> 
> Any solutions anyone can suggest here?
> 
> tim





More information about the Squeak-dev mailing list