erroneous flush() call made to read-only remote DB

[mhammond]

* Start a read-only tcp server on (say) port 8100 * Execute 'python -c "import xapian;xapian.remote_open('localhost', 8100)" * You will see a message logged at the server:

Got exception NetworkError?: Server is read-only

It turns out this is a flush() call being made as the client's 'Database' object is destructing - specifically, the flush is in 'Database::Internal::dtor_called'.

This happens as the database is destructing, so the connection is already being closed. The error doing the flush() is caught and ignored - so as far as I know, this is purely a cosmetic issue (but one that might cause concern)

[olly]

I've changed this error to be InvalidOperationError?, which avoids the worrying message.

The code in dtor_called() still should be sorted out somehow though.

[olly]

I think I saw a way to easily fix this (essentially mark transactions as "UNSUPPORTED" for a read-only database, which is true really). So marking this for 1.0.1.

[olly]

On second thoughts, keep blockage on 120 as well as adding 154 as that seems the more logical approach. Sorry for the extra bug spam.

[olly]

I've just committed a fix to SVN HEAD.

[olly]

Fixed in 1.0.1.