[pology] Pology bug where non-exact compendium matches lose information on previous strings

[Chusslove Illich]

 > [: Karl Ove Hufthammer :]
 > But when testing, I found a behaviour which I find somewhat confusing,
 > but which may be intended. What should happen when a string is matched
 > with a similar *fuzzy* string in the same file? [...] Note that there
 > is no longer a ‘#|’ comment indicating the msgid of the suggestion. So
 > it looks like the original msgid is the present msgid.

I think that we should change the angle we're looking at this, as
follows.

If in a PO file there is a fuzzy message (in any context whatsoever)
that doesn't have original msgid, then the relationship between present
msgid and present msgstr has to be taken as unknown, and the translation
has to be checked manually in full. Would you agree to this convention?

Then, if some tool is intentionally fuzzing a translated message, it
should make sure not just to set the fuzzy flag, but also to set the
original msgid to the present msgid. In Pology this is the case with
pomerge with compendium (which is what triggered your analysis), and
also one other obscure case with posummit scatter. I attach the patch
which now does set original msgid to present upon setting fuzzy flag.
Can you check it a bit in action and report if anything is amiss?

-- 
Chusslove Illich (Часлав Илић)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pology-merge-compendium-fuzzy-with-previous-01.diff
Type: text/x-patch
Size: 1317 bytes
Desc: not available
URL: <http://lists.nedohodnik.net/pipermail/pology-nedohodnik.net/attachments/20220526/dd1fc5ca/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 203 bytes
Desc: OpenPGP digital signature
URL: <http://lists.nedohodnik.net/pipermail/pology-nedohodnik.net/attachments/20220526/dd1fc5ca/attachment.sig>