[libvirt] [PATCH] safe{read, write}: Don't lie on nonblocking FD

[Michal Privoznik]

On 16.01.2013 19:40, Daniel P. Berrange wrote:
> On Wed, Jan 16, 2013 at 07:39:53PM +0100, Michal Privoznik wrote:
>> On 16.01.2013 19:31, Daniel P. Berrange wrote:
>>> On Wed, Jan 16, 2013 at 07:27:46PM +0100, Michal Privoznik wrote:
>>>> Currently, whenever somebody calls saferead() on nonblocking FD
>>>> (safewrite() is totally interchangeable for purpose of this
>>>> message) he might get wrong return value. For instance, in the
>>>> first iteration some data is read. The number of bytes read is
>>>> stored into local variable 'nread'. However, in next iterations
>>>> we can get -1 from read() with errno == EAGAIN, in which case the
>>>> -1 is returned despite fact some data has already been read. So
>>>> the caller gets confused.
>>>>
>>>> Moreover, the comment just above the functions says, they act
>>>> like regular read() with nicer handling of EINTR. Well, they
>>>> don't now.
>>>
>>> I think that it is correct that these APIs return -1 on EAGAIN.
>>> These APIs should *not* be used on non-blocking FDs.
>>>
>> In that case I think we have to note it explicitly in the comments.
> 
> BTW, what code did you encounter that was using this with non-blocking
> fds ?
> 
> Daniel
> 

My new code which I am working on. Basically, from the event loop I was
trying to read from a FD (hence a nonblocking FD) and I used
saferead(fd, ...) instead of read(fd, ...). It took me a while to find
out why am I not getting anything else than -1/EAGAIN.

Michal