Once this is merged I plan on adding the test added here to https://github.com/stan-dev/performance-tests-cmdstan/tree/master/optimizer-stress-models -- since this is a runtime error, testing the result of actually running it will be more useful

I couldn't tell from the PR title what was being done here. I hope the intention is to fix the bug so that the program you list works as intended. In the end, x[:] is equivalent to x[] and x, so if the proposal's just to avoid generating a needless multi-index, that sounds like a great idea.

I'd look at not just x[:], but also x and x[], all of which are synonyms here.

P.S. I had no idea that there is a zeros_vector(n) function! I see rep_vector(0, n) everywhere in my own and others' code.

so if the proposal's just to avoid generating a needless multi-index, that sounds like a great idea.

Yes. If the left-hand-side of an assignment is indexed only with what we call "omni" indexes (blank or :, they're the same by this part of the compiler), we can leave them out of the generated C++ rather than pass them and end up in a different code path

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.56%. Comparing base (e511e97) to head (16f05b4).
Report is 11 commits behind head on master.

@@            Coverage Diff             @@
##           master    #1508      +/-   ##
==========================================
+ Coverage   89.55%   89.56%   +0.01%     
==========================================
  Files          66       66              
  Lines        9695     9698       +3     
==========================================
+ Hits         8682     8686       +4     
+ Misses       1013     1012       -1