On Sun Aug 28 10:16:31 2016, 1parr...@gmail.com wrote: > > I think you're getting an end user's-eye view of what internal > > backtraces look like. > > Right! If the problem is simply a user error, (an invalid entry in an > sprintf format string), the messages are irrelevant and confusing. > > I thought I was reporting some problem in the processing or creation > of the error message. In that case P6 is broken in some way, and the > messages presumably help the language's developers fix that. > Improved things a good bit in dc7f27988. The sprintf format strings are parsed using a grammar in NQP. When we can't parse it, we call the panic method (a standard name used by the Perl 6 grammar also). This is all fine. What wasn't so fine is that:
* We let that leak into the userspace backtrace, where it...well, apparently makes people panic * We accidentally trimmed the userspace backtrace below that line due to some not so smart logic to avoid showing unhelpful stuff. I've fixed those up so that you get a decent backtrace of where the sprintf was now and it doesn't mention the panic method - which is an implementation detail. We should probably write some kind of test for this; I'll tag it testneeded. :-)
