@eryksun
If I was to report this bug in the GNU readline's tracker, what should I have called the title?

I'd title it "errno not returned for some errors in write_history, append_history, and history_truncate_file". In particular these calls return -1 when the internal histfile_restore call fails because rename fails. It's fine for histfile_restore to return the result from rename, but this should be checked for failure (e.g. -1) and handled appropriately by the caller in history_do_write and history_truncate_file. For example, in history_do_write they do the following:

if (rv == 0 && histname && tempname)
    rv = histfile_restore (tempname, histname);

if (rv != 0)
    {
      if (tempname)
	unlink (tempname);
      history_lines_written_to_file = 0;
    }

This needs a simple fix to update the value of rv when histfile_restore fails:

if (rv == 0 && histname && tempname)
    rv = histfile_restore (tempname, histname);

if (rv != 0) {
    rv = errno;
    if (tempname)
        unlink(tempname);
    history_lines_written_to_file = 0;
}

Codecov Report

Merging #18299 into master will decrease coverage by 0.07%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #18299      +/-   ##
==========================================
- Coverage   82.20%   82.12%   -0.08%     
==========================================
  Files        1957     1954       -3     
  Lines      589079   583367    -5712     
  Branches    44401    44401              
==========================================
- Hits       484247   479093    -5154     
+ Misses      95174    94636     -538     
+ Partials     9658     9638      -20     

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 58a4054...087f9da. Read the comment docs.

I still want to notice the users that the .python_history file is not writable, so that they wouldn't guess there's something wrong somewhere else.