[isabelle-dev] An experience report on the testboard

[Lars Hupel]

> The problem I was referring to is that in order to get a deeper idea
> what's going on you want to have a look at the topology of the history.
>  However I guess this is closely related to my surprise that not both
> patches to distro and AFP pushed in proximity have been considered as
> one build job.  So maybe this is not important.

Sorry, I don't understand. You push some changeset and Jenkins will tell
you exactly which Mercurial id is the current tip. What else do you need?

> A notification to the committer(s) of the respective changesets.  Maybe
> Jenkins has already support for that?

It does, but currently neither of the following conditions apply:

1) both repositories observe the DVCS standard of including a valid
email address in the commit author string; or
2) we have a "testboard_submit" command which asks for an email address
or reads it from a configuration or uses desktop notifications; or
3) we can figure out which user pushed a change

> The workflow I am alluding to is as follows:
> a) You make a change to the distro.
> b) You have a rough idea what the consequences are and check the
> involved sessions after an analysis.
> c) You do some kind of plausibility checking (e.g. building HOL-Library)-
> d) You push optimistically to the testboard.
> e) Hence you get a rough idea what has still to be done in certain sessions.
> f) You check the corresponding sessions.
> g) Now you are ready for the next iteration.

Right. That would be easier under the presence of a good
"testboard_submit" command.

As I alluded to in my earlier email this is all well within the realm of
possibility (even platform-independent desktop notifications upon
completion), but it requires some dedicated implementation effort.

Cheers
Lars