On Tue, Jun 25, 2013 at 5:35 PM, Bert Huijben <bert_at_qqmail.nl> wrote:
>> -----Original Message-----
>> From: ivan_at_apache.org [mailto:ivan_at_apache.org]
>> Sent: dinsdag 25 juni 2013 14:01
>> To: commits_at_subversion.apache.org
>> Subject: svn commit: r1496434 - /subversion/trunk/subversion/svn/svn.c
>>
>> Author: ivan
>> Date: Tue Jun 25 12:00:42 2013
>> New Revision: 1496434
>>
>> URL: http://svn.apache.org/r1496434
>> Log:
>> * subversion/svn/svn.c
>> (svn_cl__check_cancel): Validate passed cancel baton for better test
>> coverage.
>>
>> Modified:
>> subversion/trunk/subversion/svn/svn.c
>>
>> Modified: subversion/trunk/subversion/svn/svn.c
>> URL:
>> http://svn.apache.org/viewvc/subversion/trunk/subversion/svn/svn.c?rev=
>> 1496434&r1=1496433&r2=1496434&view=diff
>> ==========================================================
>> ====================
>> --- subversion/trunk/subversion/svn/svn.c (original)
>> +++ subversion/trunk/subversion/svn/svn.c Tue Jun 25 12:00:42 2013
>> @@ -1651,6 +1651,8 @@ signal_handler(int signum)
>> svn_error_t *
>> svn_cl__check_cancel(void *baton)
>> {
>> + /* Cancel baton should be always NULL in command line client. */
>> + SVN_ERR_ASSERT(baton == NULL);
>> if (cancelled)
>> return svn_error_create(SVN_ERR_CANCELLED, NULL, _("Caught signal"));
>> else
>
> Nice fix.
> I was thinking about adding this last weekend when I noticed the problem on dev_at_s.a.o.
>
> Personally I would have used assert() to optimize the check away in
> release mode, but I don't think this difference is measurable.
>
I've also considered using assert() but decided to stay with
SVN_ERR_ASSERT() to keep code consistent.
--
Ivan Zhakov
CTO | VisualSVN | http://www.visualsvn.com
Received on 2013-06-25 15:39:58 CEST