WIP: Good gc

[Ngoguey42]

No description provided.

[Ngoguey42]

@metanivek I have to modify your work a bit in ae2e176 in order to avoid having to make explicit casts when using functions of Io_errors.S

[codecov-commenter]

Codecov Report

Merging #1944 (b0d70ad) into main (2c6185e) will increase coverage by 0.03%.
The diff coverage is 86.00%.

@@            Coverage Diff             @@
##             main    #1944      +/-   ##
==========================================
+ Coverage   64.93%   64.96%   +0.03%     
==========================================
  Files         126      127       +1     
  Lines       15106    15125      +19     
==========================================
+ Hits         9809     9826      +17     
- Misses       5297     5299       +2     

Impacted Files	Coverage Δ
src/irmin-pack/unix/errors.ml	30.00% <0.00%> (ø)
src/irmin-pack/unix/inode.ml	33.33% <ø> (ø)
src/irmin-pack/unix/io_errors.ml	45.00% <33.33%> (-0.46%)	⬇️
src/irmin-pack/layout.ml	73.33% <50.00%> (-8.49%)	⬇️
src/irmin-pack/unix/dispatcher.ml	87.50% <87.50%> (ø)
src/irmin-pack/unix/control_file_intf.ml	100.00% <100.00%> (ø)
src/irmin-pack/unix/ext.ml	74.11% <100.00%> (+0.10%)	⬆️
src/irmin-pack/unix/file_manager.ml	85.59% <100.00%> (+0.57%)	⬆️
src/irmin-pack/unix/gc.ml	83.65% <100.00%> (+0.15%)	⬆️
src/irmin-pack/unix/pack_store.ml	81.72% <100.00%> (-0.30%)	⬇️
... and 1 more

📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

[metanivek]

@Ngoguey42 ah, yep. Making t abstract removed the possibility of having [> t] in function types (IIUC t and [< t] are equivalent). Getting rid of the type coercion is 💯. Thinking about this more, I wonder if we ought to eliminate Errors.Base and move all that functionality into Io_errors since we never (in practice) use just the base error manager. I don't think the separation gains much. This would leave only the error types in Errors.

[Ngoguey42]

Yes, moving this code sounds like a good idea. But IMO, let's keep it as is for now and postpone that cleanup up to the day we have a second refactor needed (e.g. use that new error system in irmin-pack too)

[icristescu]

Included in #1950