Fix Lwt_fmt.stderr to actually point to stderr instead of stdout (#850)

[vog]

Fix Lwt_fmt.stderr to actually point to stderr instead of stdout.

This fixes #850.

[Drup]

👍

(I'm slightly ashamed 🥺)

Maybe add a test ?

[raphael-proust]

Maybe add a test ?

A test might be good, but it's not obvious how to add it. The only tests that touch stdout/stderr are for Lwt_io and they are tests for servers and connections.

I'd be ok merging this without a test. I'll leave this open for now so the discussion can continue for a bit on that topic.

[raphael-proust]

If you have the time, can you add the following diff? (I couldn't push to your repo/branch directly.) If not it's ok, I'll add it when doing the release.

diff --git a/CHANGES b/CHANGES
index 67f9ad454..aa7540df0 100644
--- a/CHANGES
+++ b/CHANGES
@@ -4,6 +4,10 @@

   * Lwt_seq: a Seq-like data-structure with Lwt delayed nodes (#836, Zach Shipko).

+====== Bugfixes ======
+
+  * Fix Lwt_fmt.stderr to actually point to stderr (#852, #850, Volker Diels-Grabsch).
+
 ===== 5.4.0 (2020-12-16) =====

 ====== Installability ======

[vog]

Will there be an immediate bugfix release of Lwt as soon as this fix ran successfully through CI/CD, or do you collect for fixes before releasing? In the latter case: When will the schedule for the next bugfix release?

[raphael-proust]

I'm tempted to make a bugfix release for that. I'll check with the other maintainers and come back to you.

[avsm]

Agreed, let's do a bugfix release for this.

[raphael-proust]

I'll make the release this week.