Re: [E-devel] ecore_pipe again

[Peter W. <pet...@we...>]

Sebastian Dransfeld schrieb:
> Peter Wehrfritz wrote:
>   
>> Hi all,
>>
>> I took Vincent's last ecore_pipe approach and changed ecore_pipe_write 
>> to accept variable long data and not only a pointer. I also added the 
>> possibility to pass a data pointer to the handler callback to avoid the 
>> usage of globals. The win version does probably not work.
>>
>> Please read it carefully, because I'm not very familiar with pipes. 
>> Attached you'll find a little test app, too.
>>     
>
> Shouldn't you check that you actually read the number of bytes the 
> system says is in the pipeline

Right, sorry I forgot to mention, that I skipped them, because I'm not 
sure how I should handle these cases. Of course that doesn't mean that 
we have not to handle them.

1. What shall we do if the return size is smaller (but > 0). If I 
understood it right, that can happen if the writing process writes more 
bytes then PIPE_BUF. Should we then retry it, to get the rest of the 
data. Or should we buffer the already read data and try it to get the 
rest when _ecore_pipe_read is called again.

2. What should we do if we get an error messages? And when can this 
happen? If we can loose some bytes then we are out of sync with the 
lenght and data portions and the results would be randomly, but in any 
case wrong.

3. In Vincent's version he called read as long as he gets pointer from 
the pipe, should I also loop as long as there is data in the pipe?

I hope I covered here all possible errors.

Peter