Re: [PATCH v1] Add a way to supply stdin to TAP tests

On 5/6/19 10:42 PM, David Fetter wrote:
> On Tue, May 07, 2019 at 11:05:32AM +0900, Kyotaro HORIGUCHI wrote:
>> Hi.
>>
>> At Sun, 28 Apr 2019 17:07:16 +0200, David Fetter <david(at)fetter(dot)org> wrote in <20190428150716(dot)GP28936(at)fetter(dot)org>
>>> Our test coverage needs all the help it can get.
>>>
>>> This patch, extracted from another by Fabian Coelho, helps move things
>>> in that direction.
>>>
>>> I'd like to argue that it's not a new feature, and that it should be
>>> back-patched as far as possible.
>> The comment for the parameter "in".
>>
>> +# - in: standard input
>>
>> Perhaps this is "string to be fed to standard input". This also
>> can be a I/O reference but we don't care that?
> OK
>
>> + $in = '' if not defined $in;
>>
>> run($cmd, '<', \undef) seems to work, maybe assuming "<
>> /dev/null", which might be better?
> Is /dev/null a thing on Windows?

Not as such, although there is NUL (see src/include/port.h).

However, I don't think we should be faking anything here. I think it
would be better to  avoid setting $in if not supplied and then have this:

if (defined($in))

{

    IPC::Run::run($cmd, '<', \$in, '>', \$stdout, '2>', \$stderr);

}

else

{

    IPC::Run::run($cmd, >', \$stdout, '2>', \$stderr);   

}

cheers

andrew

--
Andrew Dunstan https://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services