Yeah, I was just looking into what that change was all about - plenty long ago enough to be a mystery. My notes suggest it was fixing duplications in the incoming list and so using a Set was pretty obvious. I'm astonished it's taken almost 2 years for anyone to notice any problem...

I think the real problem here is trying to use a simple listing UI to present too-sophisticated information. A message trace browser could probably do much better with something akin to a folding-list UI. Or even a cool!Graphical!web! (like https://vega.github.io/vega/examples/edge-bundling/ or similar)

Lots of work though. But then all the browsers are pretty naff when you look at the structures; subclassing from StringHolder? Really?

> On 2022-06-22, at 12:03 AM, Marcel Taeumel <marcel.taeumel at hpi.de> wrote:
> 
> Hi Tim --
> 
> This introduced a regression, which I now fixed via Tools-mt.1165.
> 
> Yes, the name "MessageSet" indicates a Set semantic. However, tools in the past have relied on the sort order. We must preserve that. I added a simple "Does it have indentation?" test to assume pre-defined sort order.

... and having to use that sort of test it why we really should do better with the broader structure. But still, we can't solve that at this stage.

tim
--
tim Rowledge; tim at rowledge.org; http://www.rowledge.org/tim
He who hesitates is probably right.