Current setup timestamp 1590343308

[Fergus Daly]

>> It looks to me that /etc/setup/timestamp is updated by the last successful setup
>> (upgrade?) run, and contains a copy of the selected mirror's setup.ini
>> setup_timestamp field as of the last successful setup (upgrade?) run, a few
>> hours earlier than the last successful setup (upgrade?) run.

Thank you for this.
I can't really think of anything to say other than repeat my observation, today, of altered practice.
Since as long as I can remember (and I mean - a matter of years) a successful upgrade of Cygwin32
(implementing the current version of setup-x86.exe currently 2.904
and pulling in the latest version of setup.ini found at {source}/Cygwin/x86/)
has concluded with
(i) a glitch-free accounting in /var/log/setup.log(.full), together with
(ii) a revision of the file /etc/setup/timestamp
which reflects the timestamp line in the aforementioned setup.ini file.
(And other things too, quite apart from any updates to the Cygwin provision - e.g. an updated /etc/setup/installed.db file.)

What I have observed, twice now (earlier today using setup.ini incorporating timestamp 1590343308 and now using the latest setup.ini incorporating timestamp 1590407755) is that event (ii) is failing: the file /etc/setup/timestamp is not updated - and if it isn't there at all, it is not created.

Maybe this does not matter. The file is not referred to during any upgrade session  (or not obviously - I think the contents of /etc/setup/installed.db are what dictate any necessary upgrades to an individual user's Cygwin provision) but it has been, for me at least, a useful rapid check of whether "what I've got" matches "what is now available". 

All I was trying to do was draw attention to (what I perceive to be) suddenly altered practice, with no change to the setup executable that might have explained it or caused it. So (a) is it a real change? (easily confirmed, or not, by anybody with /etc/setup/timestamp reporting < 1590407755 then running an upgrade, and seeing what happens to /etc/setup/timestamp); and (b) if so, is it intended? or is it a trivial glitch introduced goodness-knows-how that can be corrected; and (c) if intended, to provide what improvement? because it seems to me a bad move.

Fergus