adding traces in SpMorphicMillerAdapter to trace problems in CI

[adri09070]

Trying to fix #664

[Ducasse]

do you need it in Pharo or just your PR is enough to get the trace?

[adri09070]

This is just for me to investigate so I think the PR should be enough (and if it should be in Pharo, I should do a PR in pharo-spec/Spec instead ).
I discovered what causes problem here but I don't think this is a debugger problem. This is a problem in Morphic/SpecMorph adapters.

In SpMorphicMillerAdapter>>#scrollToShowLastPage, a bloc evaluation is defered which show a submorph of its inner widget. However, sometimes its inner widget has no submorph and the adapter tries to show the morph at index 0, which causes this random error on the CI.
I don't think this is something I can control in the debugger. Maybe @estebanlm has an idea as to how the SpMorphicMillerAdapter's inner widget happens to have no submorph ?

[adri09070]

Example of CI run that went wrong with traces: https://github.com/pharo-spec/NewTools/actions/runs/7611736592/job/20727806873?pr=665

Example of CI run that went smoothly with traces: https://github.com/pharo-spec/NewTools/actions/runs/7611658534/job/20727565575

Also another example of CI run that went wrong and in which I don't know what the hell happened because there are completely different errors with identical (but less numerous) traces: https://github.com/pharo-spec/NewTools/actions/runs/7611490325/job/20727060848

[adri09070]

An ugly solution that could at least fix the CI is to add an innerWidget submorphs ifNotEmpty: guard but I'm pretty sure someone familiar with Morphic would be better suited to investigate and fix it

[Ducasse]

Let us fix it in a ugly way because there is no Morphic expert and we are working for killing morphic :)

[estebanlm]

the problem seems to be a race condition, not directly a morphic problem, heh.
in anycase, I will merge if you promise to revert as soon as you have the information you need :)

[estebanlm]

An ugly solution that could at least fix the CI is to add an innerWidget submorphs ifNotEmpty: guard but I'm pretty sure someone familiar with Morphic would be better suited to investigate and fix it

you have Morph>>#hasSubmorphs method, no need to break demeters law.

[estebanlm]

btw, you know you can run smalltalkci locally, isn't?
that may help you to find the problem without charging the real CI

[adri09070]

btw, you know you can run smalltalkci locally, isn't? that may help you to find the problem without charging the real CI

How do you do that?

[estebanlm]

btw, you know you can run smalltalkci locally, isn't? that may help you to find the problem without charging the real CI

How do you do that?

you checkout https://github.com/hpi-swa/smalltalkCI and from terminal execute something like:

./bin/smalltalkci -s Pharo64-12 /home/esteban/dev/repo/pharo-vcs/iceberg/.smalltalk.ston

[adri09070]

I think I fixed this error, but I now have a different (but similar) one (that was already there: https://github.com/pharo-spec/NewTools/actions/runs/7603649968/job/20705636397?pr=660)

[adri09070]

btw, you know you can run smalltalkci locally, isn't? that may help you to find the problem without charging the real CI

How do you do that?

you checkout https://github.com/hpi-swa/smalltalkCI and from terminal execute something like:

./bin/smalltalkci -s Pharo64-12 /home/esteban/dev/repo/pharo-vcs/iceberg/.smalltalk.ston

Well, when I do that, I get an error because the CI image cannot find Sindarin ...

[adri09070]

Everything should be good once pharo-spec/Spec#1511 and pharo-project/pharo#16016 will be merged