<LnL>
I took a stab at that at some point, but it isn't in the "this compiles" type of code
<LnL>
the change was pretty straightforward but not sure what the side effects are, since a bunch of stuff later on expects the previous eval to be there for diffing and such
<cole-h>
Maybe if the target branch fails to eval, we only eval the PR and drop some of the other statistics/checks that rely on the target branch's successful eval?
* cole-h
is just spitballing
<LnL>
yeah, there's no way to check those if eval is failing
<LnL>
but the code uses things like Option<Stdenvs> which get filled in and the later code assumes it's Some
<cole-h>
Hm
<LnL>
so the type checking doesn't help much, needs proper testing
<gchristensen>
yeah that code is nasty
<gchristensen>
I'm sosorry
<LnL>
(or fails if it's None)
<cole-h>
ofborg 2.0 Eventually™
<gchristensen>
ofborg 2.0 one deployed patch at a time
<cole-h>
Does RiiR apply even if the base project is already written in Rust? 🤔
<gchristensen>
RiiGR
<cole-h>
"Rewrite it in Gooder Rust"
<gchristensen>
exactly
<cole-h>
Love it
<LnL>
if that had some tests it would be relatively fun to refactor into more of a statemachine
<LnL>
but currently it's kind of the most scary part for me
<cole-h>
The entire codebase is scary to me x)
<LnL>
heh :)
<gchristensen>
make changes, push it live, see what happens