Git conversion: fixing email addresses from ChangeLog files

[Joseph Myers]

On Sun, 29 Dec 2019, Joseph Myers wrote:

> I've now made those changes to the checked-in list so it's pure UTF-8, and 
> thus easier to review and edit.  We still need to implement code in 
> bugdb.py to use that list to pick the preferred form from each list of 
> variants (and people may wish to change the preferred forms in some 
> cases).

I've now implemented that code in bugdb.py.

Given those fixes, I'm planning to compare author names from the 
reposurgeon conversion and Maxim's conversion, as I think cases where they 
find different authors (not just different email addresses) are good cases 
for manual review (we already have various such manual author fixups for 
individual commits in bugdb.py).  In fact that manual review may show up 
*other* commits that should be reattributed.  One example Maxim gave of a 
missing author was Aymeric Vincent.  That was a commit on 
premerge-fsf-branch where the reposurgeon heuristic "don't use 
attributions from ChangeLog for a ChangeLog-only commit" applied.  But 
whether or not the commit just adding the ChangeLog entry should be 
reattributed to the person named in that ChangeLog entry, the real changes 
that ChangeLog entry relates to are two previous commits (each file 
committed separately), so it shows up that those two previous commits 
ought to be reattributed.

-- 
Joseph S. Myers
joseph@codesourcery.com