My suggested change is as follows.
Besides the current output (including the unterminated "<info>"),
additionally output (on stdout):
<error id="E155007">'/tmp' is not a working copy.</error></info>
That provides some information in the xml rather than just failure.
-s
On 01/18/2012 05:02 PM, Daniel Shahaf wrote:
> Well, yes, but backwards compatibility means that we have to keep it
> returning an error ($? != 0 and stderr != ""), or we'll break everyone
> else's scripts.
>
> If you have a suggested change that is backwards compatible, we'd love
> to hear it.
>
> Steven R. Loomis wrote on Wed, Jan 18, 2012 at 10:27:45 -0800:
>> Agreed, errors can happen at any point- but this particular error
>> hardly seems like an unexpected internal fatal assertion, especially
>> given svnversion.
>>
>> Thanks, not urgent for me, just surprising.
>>
>> -s
>>
>> On 01/18/2012 09:25 AM, Daniel Shahaf wrote:
>>> An error may happen at any point during processing.
>>>
>>> Agreed that if an error pertains to one specific target it'd be sane to
>>> render the error as XML within that target's scope and render the
>>> remaining targets normally.
>>>
>>> But if there is a more global error (say: internal assertion triggered)
>>> I think it still makes sense to have it on stderr so that one doesn't
>>> have to parse screenfuls to get to it.
>>>
>>> Errors on stderr are not rendered in XML.
>>>
>>> Steven R. Loomis wrote on Wed, Jan 18, 2012 at 08:40:57 -0800:
>>>> Daniel,
>>>> I'm not asking to change the error condition.
>>>> I guess I would prefer to emit an error in XML format, or to emit
>>>> no XML at all. It doesn't seem like an "unexpected fatal error",
>>>> especially because svnversion doesn't return an error.
>>>>
>>>> Perhaps,
>>>> <info><error id="E155007">'/tmp' is not a working copy.</error></info>
>>>>
>>>> Thanks,
>>>> Steven
>>>>
>>>> On 01/17/2012 11:16 PM, Daniel Shahaf wrote:
>>>>> Or, are you asking that it doesn't consider "the target is not a working
>>>>> copy" an error condition in the first place?
>>>>>
>>>>> Daniel Shahaf wrote on Wed, Jan 18, 2012 at 08:53:21 +0200:
>>>>>> Why do you need well-formed XML when both stderr and $? indicate an
>>>>>> error condition?
>>>>>>
>>>>>> Steven R. Loomis wrote on Tue, Jan 17, 2012 at 15:21:10 -0800:
>>>>>>> svn info --xml gives malformed XML on error.. could it detect this
>>>>>>> situation and still cleanup the XML, or not output any XML at all?
>>>>>>> svnversion by comparison outputs 'exported'.
>>>>>>>
>>>>>>> -s
>>>>>>>
>>>>>>> $ svn info --xml /tmp
>>>>>>> <?xml version="1.0" encoding="UTF-8"?>
>>>>>>> <info>
>>>>>>> svn: E155007: '/tmp' is not a working copy
>>>>>>> $ svn --version
>>>>>>> svn, version 1.7.2 (r1207936)
>>>>>>> compiled Dec 2 2011, 15:21:13
>>>>>>>
>>>>>>>
>>>>>>>
Received on 2012-01-19 05:56:06 CET