Petrel progress bar problem [was: Re: testing of download from divecomputer needed]

[Jef Driesen]

On 2015-01-14 22:47, Henrik Brautaset Aronsen wrote:
> Henrik Brautaset Aronsen wrote:
>> An individual 0-100% progress for each dive is OK.

Ignoring the fact that it currently never reaches 100%, a progress for 
each individual dive is not as good as a global progress for the entire 
operation. With the individual progress you still don't know how long 
the download will take. Of course that's better than nothing, but not 
optimal either.

>> The problem was that the progress disappeared after a few dives had 
>> been imported.  With Dirk's debug line it becomes obvious why:
> 
> [...]
> 
>> progress 1539 of 1540
>> progress 1540 of 1540
>> progress 0 of 16777219
>> progress 3 of 16777219
>> progress 147 of 16777219
> 
> 16777219  is 2^24 + 3.  Sounds magical.

That 2^24 - 1 (or 0xFFFFFF) is the worst case value I mentioned in my 
previous email. The Petrel protocol contains a 24bit length in the 
request packets. But because a dive has a variable length and that 
length isn't known in advance, you are supposed to set this field to its 
maximum value. The extra 4 bytes are protocol overhead.

Jef