Re: svn commit: r1658032 - /subversion/trunk/subversion/libsvn_fs_x/hotcopy.c

[Branko Čibej <br...@wandisco.com>]

On 07.02.2015 08:48, stefan2@apache.org wrote:
> Author: stefan2
> Date: Sat Feb  7 07:48:14 2015
> New Revision: 1658032
>
> URL: http://svn.apache.org/r1658032
> Log:
> * subversion/libsvn_fs_x/hotcopy.c
>   (entry_name_to_utf8): It is perfectly acceptable to use a RESULT_POOL
>                         for the construction of an error object.  Hence
>                         there is no need for a separate SCRATCH_POOL.

While the above may be true in this particular case, it's certainly not
a rule we can apply everywhere. You usually don't know where the result
pool came from and how long its lifetime is.

-- Brane

Re: svn commit: r1658032 - /subversion/trunk/subversion/libsvn_fs_x/hotcopy.c

[Stefan Fuhrmann <st...@wandisco.com>]

On Sat, Feb 7, 2015 at 10:56 AM, Branko Čibej <br...@wandisco.com> wrote:

> On 07.02.2015 08:48, stefan2@apache.org wrote:
> > Author: stefan2
> > Date: Sat Feb  7 07:48:14 2015
> > New Revision: 1658032
> >
> > URL: http://svn.apache.org/r1658032
> > Log:
> > * subversion/libsvn_fs_x/hotcopy.c
> >   (entry_name_to_utf8): It is perfectly acceptable to use a RESULT_POOL
> >                         for the construction of an error object.  Hence
> >                         there is no need for a separate SCRATCH_POOL.
>
> While the above may be true in this particular case, it's certainly not
> a rule we can apply everywhere. You usually don't know where the result
> pool came from and how long its lifetime is.
>

The error object is always allocated from its own private
root pool. My comment here is imprecise in that respect,
as only the temporaries used to construct the error message
will be allocated from RESULT_POOL. I corrected that.

-- Stefan^2.