STATUS
------
Approved with changes.
REQUIRED
--------
1. Fix spelling errors, please.
COMMENTARY
----------
I only tested it on Mac.
DETAILS
-------
> Bug#45737 - BACKUP: ASSERT failure in stream_v1_transport.c
>
> It is possible to arrive at a situation where backup stream's output
> buffer is empty (buf.begin == buf.header == buf.pos) and at the same
> time is at the very end of the current output block (buf.pos == buf.end).
>
> If bstream_write_part() is called in such situation, it calles
> bstream_flush() expecting that after that there will be free space in
> the buffer (buf.pos < buf.end). This is checked with an assertion.
>
> Before this fix, bstream_flush() did nothing in the above situation
> leading to failure of the assertion in bstream_write_part().
>
> After this patch, it will be ensured that after bstream_flush() there
> always is some free space in the output buffer.
> @ sql/backup/stream_v1_transport.c
> Call reset_output_buffer() if it is empty but at the end of the current
> output
> block. This will move stream to the next output block and prepare it for
> accepting more data.
> @ unittest/backup/bug45737-t.c
> A unit test which provokes the situation which lead to assertion failure.
[1] I think you meant 'led to the assertion failure'. You could also use
'caused the assertion failure', or 'resulted in an assertion failure'.
Or did you mean, 'can lead to an assertion failure'?
...
> --- a/unittest/backup/bug45737-t.c 1970-01-01 00:00:00 +0000
> +++ b/unittest/backup/bug45737-t.c 2009-07-14 12:54:20 +0000
> @@ -0,0 +1,123 @@
> +/*
> + @file
> +
> + Provoke a situation which lead to assertion failure reported in BUG#44569.
[1] s/led/lead
...
> + if (ret != BSTREAM_OK)
> + {
> + diag("Writting error %d when preparing final stram position", ret);
[1] s/stream/stram ... or is 'stram' a new word for backup stream? :P
...
Chuck
