Re: svn add and inconsistent line endings

On 24.02.2013 18:56, Bert Huijben wrote:
>
>
>> -----Original Message-----
>> From: Stefan Küng [mailto:tortoisesvn_at_gmail.com]
>> Sent: zondag 24 februari 2013 17:01
>> To: Bert Huijben
>> Cc: 'Subversion Development'
>> Subject: Re: svn add and inconsistent line endings
>>
>> On 24.02.2013 16:47, Bert Huijben wrote:
>>>
>>>
>>>> -----Original Message-----
>>>> From: Stefan Küng [mailto:tortoisesvn_at_gmail.com]
>>>> Sent: zondag 24 februari 2013 15:39
>>>> To: Subversion Development
>>>> Subject: svn add and inconsistent line endings
>>>>
>>>> Hi,
>>>>
>>>> When auto-props are set up for e.g., cpp files that set the
>>>> svn:eol-style property, adding those files is not possible if the file
>>>> has inconsistent line endings.
>>>> The --force flag won't help either, the only way to add the file is to
>>>> first fix the line endings in an editor.
>>>> Or add the file with --no-auto-props and then add the properties later.
>>>>
>>>> I think this situation is not good.
>>>> Maybe another param added to svn_client_add5 that ignores the EOL
>> check?
>>>> Or just pass the 'force' flag on to svn_wc__canonicalize_props()
> instead
>>>> of passing 'false' unconditionally in libsvn_wc\adm_ops.c,
>>>> svn_wc_add_from_disk2() ?
>>>
>>> I don't think simply ignoring this and allowing the add to continue is a
>>> good enough solution.
>>>
>>> Just passing the skip checks flags will allow adding many wrong
> properties
>>> which will cause a lot of grief later, and I would guess adding the file
>>> would just delay the error until we try to commit this file.
>>>
>>>
>>> But looking at this from the AnkhSVN side, I would like to have some api
> to
>>> make the eols on files consistent... Would this work for you?
>>
>> The API already exists, since when passing the --force flag to the
>> propset command (adding the svn:eol-style property), then the file is
>> automatically adjusted.
>
> Are you sure the file is adjusted?
>
> Looking at the code it appears we just skip the check (and many others for
> other properties).

You're right. And I just tested: even if the property is set (indicating
inconsistent eols should be either ignored or corrected), a commit for
such a file fails as well.

Maybe someone can explain why that check is needed? Is there really a
situation where the svn:eol-style property is set on a file and users
don't want the inconsistencies in eols either ignored or fixed
automatically? I can't think of one such situation.
Because if there isn't, then why not remove that check?

Stefan

-- 
        ___
   oo  // \\      "De Chelonian Mobile"
  (_,\/ \_/ \     TortoiseSVN
    \ \_/_\_/>    The coolest interface to (Sub)version control
    /_/   \_\     http://tortoisesvn.net