Re: bzr commit into mysql-6.0-backup branch (svoj:2638) Bug#37012

[Øystein Grøvlen]

Rafal Somla wrote:
> Øystein,
> 
> I think it is better to test for a pipe in the open() method. One could 
> want to create Output_stream instance even when the underlying pipe does 
> not yet exist or is a regular file. Only when we open the stream, it is 
> important that the path points at valid pipe (or no such file exists). 
> This is remote possibility but why to restrict flexibility without a 
> reason?

I have a reason: Reduced code complexity by not manipulating m_flags in 
several locations.  Why increase complexity before the flexibility is 
needed?

An alternative that is OK with me, it to move all m_flags handling to 
open.  That is, do not set m_flags at all in the constructor.

--
Øystein

> 
> Rafal
> 
>  Grøvlen wrote:
>> Sergey,
>>
>> I think a test for this functionality needs to be included.
>>
>> I also wonder whether it would be better to put the new code in the 
>> constructor.  I think putting the m_flags logic in one place instead 
>> of spreading it, would make the code easier to understand.
>>