Sizes of message for paging tests

[timfox]

The paging tests make hardcoded assumptions about the message size.

Since the message size differs on 64 bit and 32 bit systems, "fixing" the hardcoded size for a 64 bit machine might make it fail on a 32 bit machine and vice versa.

Instead the tests should be more sensible and calculate number of messages by taking into account the real size.

[timfox]

E.g. the pagingtest you just "fixed" already passed on my machine (32 bit) but didn't on yours (64 bit)

[clebert.suconic]

That test is writing 512 body on each message. my expectation when writing is.. the message could have a little memory overhead beyond 512. But those small messages are consuming more than 1K on the server due to dynamic buffer.

Shouldn't we be using smaller sizes on the dynamic buffer for small messages on the server?


I will look into calculating the size later.

[timfox]

I will look into calculating the size later.

That's not a priority for now. Getting tests passing is.