Advisory on past uses of TamiFlex with Soot

[Eric Bodden]

Dear all,


I am very sad to say that I have recently accidentally discovered a
serious mistake that happened in a Soot commit about three years ago:

https://github.com/Sable/soot/commit/db26a897fd9231ebed7f0aa40a6ce1409ebd761f


The commit was meant to simply rename a pack in Soot, but accidentally
I had committed also another change that I had presumably made for
debugging purposes. That change seems to have disabled the
TraceBasedReflectionModel that Soot/Spark normally uses to pick up the
effects of during call-graph construction. In other words, after this
change, performed on 2010-07-13, the reflection log should actually
have shown no effect whatsoever within Soot's call-graph construction.
I just noticed this when actually starting to use TamiFlex myself
again.


Now what really startles me is how this problem can have gone
unnoticed for so long. From what I can see, there are several research
papers that have used TamiFlex in combination with Soot in the past.
Surely in this setting one should have noticed if the log files show
now effect whatsoever. I am thus currently interested in finding out
who has previously used the TamiFlex/Spark combination during this
period of time. Any advice would be appreciated.


The problem is fixed now, and hopefully will not reappear ever again.


Best wishes,

Eric