> Hello world,
> the attached patch fixes the PR by not trying to determine the length
> of the base symbol when there is a substring with a length which
> is not constant. This might make the length of temporary strings
> generated by the front end optimization pass shorter, while inserting
> an additional call to determine length of the substring.
> No test case because the failure was only exposed by an instrumented
> Regression-tested. OK for trunk?