(This entry was imported from the savannah tracker, original
location: https://savannah.nongnu.org/bugs/index.php?16471)
We are trying to move towards the server doing less and less
checking for things like disk corruption, and letting the client do
it instead (since the client has to do it anyway, no point in the
server doing it too). Some people have pointed out that they
perhaps do not trust casual users doing initial pulls to figure out
what has gone wrong when they receive corrupted data, inform the
admins of the problem, etc.
Therefore, it would be nice if the client somehow could send the
server a message letting them know that corruption had been
detected, before the client disconnected. This could then be logged
on the server, where the admins could see it.
This is somewhat complicated by the fact that we don't generally
have specially separated out "consistency checking code"
-- inconsistencies are simply signaled by hitting invariants. One
solution would simply arrange that if we hit an invariant while we
are talking on the network, we dump a description of the invariant
to our peer before dying.
Reported by Unknown User, May 1, 2006