On May 9, 2009, at 9:31 PM, Arfrever Frehtes Taifersar Arahesis wrote:
> Dnia 2009-05-09 o 22:31:13 Hyrum K. Wright napisał(a):
>> On May 9, 2009, at 2:16 PM, C. Michael Pilato wrote:
>>
>>> Author: cmpilato
>>> Date: Sat May 9 12:16:40 2009
>>> New Revision: 37662
>>>
>>> Log:
>>> Add a new public API for trimming out the trace-only error chain
>>> links,
>>> and use it in mod_dav_svn to avoid sending useless error messages
>>> across
>>> the wire. This makes still more tests pass over WebDAV.
>>>
>>> * subversion/include/svn_error.h
>>> (svn_err_purge_tracing): New function prototype.
>>>
>>> * subversion/libsvn_subr/error.c
>>> (is_tracing_link): New function, taking the place of ...
>>> (is_trace_message): ... this now-removed one. Callers updated to
>>> call
>>> is_tracing_link() instead.
>>> (svn_err_purge_tracing): New function.
>>>
>>> * subversion/mod_dav_svn/version.c
>>> (merge): Use svn_err_purge_tracing() before squirreling away the
>>> post-commit hook error message.
>>>
>>> * subversion/mod_dav_svn/util.c
>>> (build_error_chain): Revert changes made in r37649.
>>> (dav_svn__convert_err): Use svn_err_purge_tracing() before operating
>>> on the Subversion error.
>>>
>>> Modified:
>>> trunk/subversion/include/svn_error.h
>>> trunk/subversion/libsvn_subr/error.c
>>> trunk/subversion/mod_dav_svn/util.c
>>> trunk/subversion/mod_dav_svn/version.c
>> ...
>>> Modified: trunk/subversion/libsvn_subr/error.c
>>> URL: http://svn.collab.net/viewvc/svn/trunk/subversion/libsvn_subr/error.c?pathrev=37662&r1=37661&r2=37662
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> =
>>> ====================================================================
>>> --- trunk/subversion/libsvn_subr/error.c Sat May 9 12:06:18 2009
>>> (r37661)
>>> +++ trunk/subversion/libsvn_subr/error.c Sat May 9 12:16:40 2009
>>> (r37662)
>>> @@ -311,16 +311,63 @@ svn_error_clear(svn_error_t *err)
>>> }
>>> }
>>>
>>> -/* Is MESSAGE the message used for the "fake" errors used in
>>> tracing? */
>>> -static svn_boolean_t is_trace_message(const char *message)
>>> +/* Is ERR a tracing-only error chain link? */
>>> +static svn_boolean_t is_tracing_link(svn_error_t *err)
>>> {
>>> #ifdef SVN_ERR__TRACING
>>> - return (message != NULL && !strcmp(message, SVN_ERR__TRACED));
>>> + /* ### A strcmp()? Really? I think it's the best we can do
>>> unless
>>> + ### we add a boolean field to svn_error_t that's set only for
>>> + ### these "placeholder error chain" items. Not such a bad
>>> idea,
>>> + ### really... */
>>> + return (err && err->message && !strcmp(err->message,
>>> SVN_ERR__TRACED));
>>> #else
>>> return FALSE;
>>> #endif
>>> }
>>>
>>> +svn_error_t *
>>> +svn_err_purge_tracing(svn_error_t *err)
>>> +{
>>> +#ifdef SVN_ERR__TRACING
>>> + svn_error_t *tmp_err = err;
>>> + svn_error_t *new_err = NULL, *new_err_leaf = NULL;
>>> +
>>> + if (! err)
>>> + return SVN_NO_ERROR;
>>> +
>>> + while (tmp_err)
>>> + {
>>> + /* Skip over any trace-only links. */
>>> + while (tmp_err && is_tracing_link(tmp_err))
>>> + tmp_err = tmp_err->child;
>>> +
>>> + /* Add a new link to the new chain (creating the chain if
>>> necessary). */
>>> + if (! new_err)
>>> + {
>>> + new_err = tmp_err;
>>> + new_err_leaf = new_err;
>>> + }
>>> + else
>>> + {
>>> + new_err_leaf->child = tmp_err;
>>> + new_err_leaf = new_err_leaf->child;
>>> + }
>>> +
>>> + /* Advance to the next link in the original chain. */
>>> + tmp_err = tmp_err->child;
>>> + }
>>> +
>>> + /* If we get here, there had better be a real link in this error
>>> chain. */
>>> + assert(new_err_leaf);
>>
>> Perhaps use SVN_ERR_ASSERT()?
>
> FYI, this call to assert() even breaks compilation on systems with
> non-ancient compiler:
> subversion/libsvn_subr/error.c: In function
> ‘svn_err_purge_tracing’:
> subversion/libsvn_subr/error.c:361: error: implicit declaration of
> function ‘assert’
> subversion/libsvn_subr/error.c:361: warning: nested extern
> declaration of ‘assert’
> subversion/libsvn_subr/error.c: In function
> ‘svn_error_abort_on_malfunction’:
> subversion/libsvn_subr/error.c:595: warning: will never be executed
>
> http://crest.ics.uci.edu/buildbot/builders/x64-ubuntu%20gcc/builds/553/steps/Build/logs/stdio/text
Mike,
Ordinarily, I like to let people fix their own bugs, but just to get
the build working again, I committed this in r37666. Feel free to
tweak as needed.
-Hyrum
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2158479
Received on 2009-05-10 04:55:14 CEST