Re: [isabelle] Difficult to reproduce bug hanging Isabelle in jEdit session



On Tue, 23 Apr 2013, Denis Lohner wrote:

Am 22.04.2013 16:27, schrieb Lars Noschinski:
On 22.04.2013 15:29, Joachim Breitner wrote:
here at Karlsruhe we are offering a practical Course on Isabelle which
began last week and we were badly surprised that Isabelle/jEdit 2013 had
severe issues on the machines in the student’s pool room: A short while
into editing the highlighting (errors and the light blue) and the output
window are no longer updated.

We could reproduce it most easily by writing, say,
         theory Scratch imports Main begin lemma "foobar"
and then changing "lemma" to "le mma" and back in very rapid succession.

I occasionally had similar problems but did not manage to produce a reproducible example yet. However, I managed to get the system back to working by hitting the "Cancel' button in the "Theories" window a few times.

Unfortunately, hitting "Cancel" doesn't work here at Karlsruhe. However, in the student's computer pool hitting "Check" while a large theory file is being processed, seems to reliabely result in the same freeze of the Output panel.

As there is no error message in the Output panel or at the command line, what are our options to investigate the problem? Is there an option to log the communication between jEdit and the isabelle process? Would this information help to find the cause?

The "Protocol" panel in Isabelle/jEdit gives a full transcript of the two processes communicating, although that slows down things considerably, and usually makes such non-deterministic problems disappear.


There have been reports about several such real-time reactivity problems in the past few months, but so far very few concrete hints of the actual reasons. Problems with Fedora 18 in particular had been reported before, but this does not say much -- recent Linux distributions all tend towards instabilities that were not there 1 or 2 years ago.

In the next round of refinements of Isabelle/Scala and Isabelle/jEdit, I will be more defensive in the use of certain Java 7 GUI / window popup features, and also in the use of Scala actors. There are reasons to believe that these basic building blocks of the interactive framework are not 100% reliable.

Apart from that, any more concrete observation where things don't work in Isabelle2013 will be treated with the usual scrutiny.


	Makarius


This archive was generated by a fusion of Pipermail (Mailman edition) and MHonArc.