On 8 October 2013 13:30, Frank Shearar <frank.shearar at gmail.com> wrote:
> http://build.squeak.org/job/SqueakTrunk/546/console shows a strange
> error during the updating of an image from 4.4-12327 to trunk. I have
> not seen this error when running headful!

For even more kicks, running the scripts on my work machine (Ubuntu
Raring) and my laptop (Windows 8) work just fine.

frank

> MCMcmUpdate >> #updateFromRepositories sends #dependencies twice.
> Judging by the "[] in [] in
> SortedCollection(Collection)>>do:displayingProgress:every:" ancestor
> context, this is the _second_ call site blowing up. The code looks
> like this:
>
> updateList
>     do: [:assoc |
>         ProgressNotification signal: '' extra: 'Processing ' , assoc value.
>         config := r versionNamed: assoc value.
>         "Skip packages that were specifically unloaded "
>         config
>             dependencies: (config dependencies
>                 reject: [:dep | self skipPackages includes: dep package name]).
>             "stuff happens"]
>     displayingProgress: 'Processing configurations'.
>
> (I hope I've reproduced the snippet correctly. Gmail "helpfully"
> removes any formatting, and so I've indented things by hand.)
>
> So the MNU comes because "r versionNamed: assoc value" returns nil. I
> have no idea what assoc looks like.
>
> Has anyone seen an error like this? I'm about to try catch the MNU and
> dump the contents of the contexts at the top of the stack, but I'm
> open to hearing of better ideas. If all else fails I'll try an Eliot
> move and snapshot the busy-failing image off to the side so I can try
> resume the process and see what happens.
>
> frank