Re: [PATCH v3] iotests: Test NBD client reconnection

[Roman Kagan <rkagan@virtuozzo.com>]

On Mon, Nov 11, 2019 at 12:18:48PM +0300, Andrey Shinkevich wrote:
> 
> 
> On 08/11/2019 17:05, Roman Kagan wrote:
> > On Fri, Nov 08, 2019 at 01:49:50PM +0000, Vladimir Sementsov-Ogievskiy wrote:
> >> 01.11.2019 19:54, Andrey Shinkevich wrote:
> >>> +def check_proc_NBD(proc, connector):
> >>> +    try:
> >>> +        exitcode = proc.wait(timeout=10)
> >>> +
> >>> +        if exitcode < 0:
> >>> +            log('NBD {}: EXIT SIGNAL {}\n'.format(connector, -exitcode))
> >>> +            log(proc.communicate()[0])
> >>> +        else:
> >>> +            line = proc.stdout.readline()
> >>
> >>
> >> could we use proc.communicate() for both cases, what is the difference?
> > 
> > In fact if proc produces any non-trivial amount of output you are better
> > off using .communicate() otherwise your child may block on output and
> > never exit.  See
> > https://docs.python.org/3/library/subprocess.html#subprocess.Popen.communicate
> > for how to express the above logic correctly.  The exit code *after*
> > .communicate is available in .returncode.
> > 
> 
> The pattern by the link above does not work (Python3):
> 
> proc = subprocess.Popen(...)
> try:
>      outs, errs = proc.communicate(timeout=15)
> except TimeoutExpired:
>      proc.kill()
>      outs, errs = proc.communicate()
> 
> as 'proc' cannot be used for output after being killed. It results in 
> another exception being raised.

Of course it can't.  You need to use the strings returned by
.communicate().

Roman.