I said before that I found with an interpreter vm I could reproduce the crash David described using this in a Workspace:
I find with RC2 it still crashes quite reliably. Linux vm version 4.13.12-3323.
I decided to lay in self halts to look for the point before it crashed. This brought me to MCConfiguration>>depsSatisfying:versionDo:displayingProgress:.
That's salient, because the last thing you see before the crash is a progress bar.
The stepping point for the crash is just after MCReorganizationPreloader preloadMovesBetween: selectedVersions.
If I place a self halt before that I get a debug window. If I place it after that I get a crash.
So I stepped into that and explored #preloadMovesBetween: which looks like this:
preloadMovesBetween: versions
| loader |
loader := self new.
self halt. "location #1"
versions do: [ :version |
loader addVersion: version].
self halt. "location #2"
^ loader preload
If I run with the self halt at location #1 I get a debug window. If I remove self halt location #1 and set it to self location #2 I get the crash.
I don't know why it's crashing, but I posit that it may not be about Monticello and instead may be about nested blocks. #depsSatisfying:versionDo:displayingProgress: has a block within a block. And here's another block.
Perhaps something is going wrong with this nesting?
As I say, I don't really know, but it was interesting to try and find the point where it crashed.
Two last things. Everything I just said applies also when employing the Squeak-4.10.2.2614-linux_i386 vm from
squeakvm.org.
David's first email displayed error output ending with:
> 23038864 WorldState>doOneCycleFor:
> 23038772 PasteUpMorph>doOneCycle
> 5017264 MorphicProject>?
> 5017136 BlockClosure>?
You will see the crash point ends with a question about blocks. The previous listing cites MorphicProject. Hmm, don't we have a known problem with Projects these days? Something about Projects not being able to load?
All this may of no account, but it was interesting to try and isolate the problem.
FWIW,
Chris