Hi @FAKERINHEART thanks for looking into this. Our workflow requires approvel for first-time contributors. I've just approved your CI workflow run so let's see how the CI tests run through.

Cc: @jzmaddock

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 94.1%. Comparing base (83d5844) to head (716eb0e).
Report is 2 commits behind head on develop.

@@           Coverage Diff           @@
##           develop    #657   +/-   ##
=======================================
  Coverage     94.1%   94.1%           
=======================================
  Files          279     279           
  Lines        28979   28979           
=======================================
  Hits         27253   27253           
  Misses        1726    1726           

Oh Also, @FAKERINHEART, do you have a test case that exercises this change to the for-loop?. That would be nice because then we could add the test case as well for the new lines.

Actually... is there any reason not to just delete the whole outer loop? If we enter the loop then the condition that leads to the break will always be taken?

Actually... is there any reason not to just delete the whole outer loop? If we enter the loop then the condition that leads to the break will always be taken?

Do you mean to remove the outer while-loop at line $848$?

I agree @jzmaddock.

• But then also remove the break; statement at line $853$ that, at the moment, breaks from the outer while-loop, right?
• Do we need a range-check on offset + 1?

Cc: @FAKERINHEART

Oh Also, @FAKERINHEART, do you have a test case that exercises this change to the for-loop?. That would be nice because then we could add the test case as well for the new lines.

No, I think that it may be no needs to add a new test case. Please look into mpfr.hpp:431, and its commit log: 485fa7c. It was fixed for the mpfr_float_imp<digits10, allocate_dynamic> three years ago, but it forgot to fix mpfr_float_imp<digits10, allocate_stack> which is for this merge case. They are in the same condition, and the previous has been add tests in the 485fa7c.

Actually... is there any reason not to just delete the whole outer loop? If we enter the loop then the condition that leads to the break will always be taken?

Do you mean to remove the outer while-loop at line 848 ?

I agree @jzmaddock.

• But then also remove the break; statement at line
  853
  that, at the moment, breaks from the outer while-loop, right?
• Do we need a range-check on offset + 1?

Cc: @FAKERINHEART

If we wanna remove the whole loop, please take its effect both to mpfr_float_imp<digits10, allocate_dynamic> and mpfr_float_imp<digits10, allocate_stack>.

CI has been passed.
Could we merge this PR please?
@ckormanyos

Bare with me, I'm not convinced this is the correct fix (or indeed that there is an issue, as the whole loop appears to be a mistake that can never actually loop anyway), I need to investigate some more....

My apologies: you're quite correct, I was completely misreading the code until I saw it with highlighting in my editor! My bad, will merge...

I'll merge this into master for 1.88 since we're still safe to do so

thx~