RFR [9] 8059101: unshuffle_patch.sh should be able to deal with stdin/stdout

Ivan Gerasimov ivan.gerasimov at oracle.com
Thu Sep 25 10:26:13 UTC 2014


On 25.09.2014 14:18, Daniel Fuchs wrote:
> Hi Ivan,
>
> Should 'usage' also be redirected?
>
This would be inconsistent with other command line utilities.
They usually print help/usage to stdout.

Sincerely yours,
Ivan

> best regards,
>
> -- daniel
>
> On 25/09/14 12:12, Ivan Gerasimov wrote:
>> Thank you Daniel for the comments!
>>
>> On 25.09.2014 13:49, Daniel Fuchs wrote:
>>> Hi Ivan,
>>>
>>> When setting output & input, I wonder if it would be simpler
>>> to use an 'if else if' construct in order to avoid the '-a'
>>> in the 'if' that follows.
>>>
>> We would still need that -a in if, as the user could pass /dev/stdin and
>> /dev/stdout as the arguments to the script.
>> Dashes are meant to be the shortcut for these long names, but we should
>> allow the long names to be passed too.
>>
>>> something like this (pseudo code) might be easier to read:
>>>
>>> if "x$output" is "x-" then
>>>    substitute - with /dev/stdout
>>> else if $output file exists
>>>    complain
>>> endif
>>>
>>> Also the script uses echo to print warnings & error.
>>> It should probably be changed to print those on stderr ( >&2 )
>>> so that they don't mix with the patch when the output is '-'.
>>>
>> Agreed!
>> Redirected all the error messages to stderr.
>>
>>> Similarly - I feel that either verbose should redirect to
>>> stderr, or the script should complain and exit if output
>>> is '-' and verbose is on.
>>>
>> Done.
>>
>> Please see the updated webrev:
>> http://cr.openjdk.java.net/~igerasim/8059101/1/webrev/
>>
>> Sincerely yours,
>> Ivan
>>
>
>
>



More information about the jdk9-dev mailing list