[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Crash in python bindings on windows (in apache)

From: Branko ─îibej <brane_at_xbc.nu>
Date: 2005-10-19 22:16:06 CEST

Jonathan Gilbert wrote:

>At 03:25 PM 19/10/2005 -0400, David James wrote:
>
>
>>On 10/19/05, Branko ??bej <brane@xbc.nu> wrote:
>>
>>
>>>David James wrote:
>>>
>>>
>>>>On 10/19/05, Marc Haesen <Marc.Haesen@telindus.be> wrote:
>>>>
>>>>
>[snip]
>
>
>>>>This is fantastic news, Marc. I have followed your suggested approach,
>>>>using pool = pool, to avoid a warning in r16817. Thanks so much for
>>>>your help and your patience!
>>>>
>>>>
>>>I'd really like to know what sort of Windows compile problem the "(void)
>>>pool" thing caused. This is an idiom we've been using throughout the
>>>code to silence warnings about unused variables. Changing this single
>>>isntance is nonsense, especially since the "compile problem" report
>>>wasn't exactly enlightening.
>>>
>>>
>>It does seem strange that "(void) pool;" would not work on Windows,
>>especially considering how often we use the same idiom elsewhere in
>>the code:
>>
>>subversion/libsvn_subr/config.c: (void)(baton); /*
>>Unused parameter. */
>>subversion/libsvn_subr/config.c: (void)(section); /*
>>Unused parameter. */
>>subversion/libsvn_wc/adm_files.c: (void)pool; /* Silence compiler
>>warnings about unused parameter. */
>>subversion/libsvn_wc/adm_files.c: (void)pool; /* Silence compiler
>>warnings about unused parameter. */
>>
>>
>
>It seems odd to me to use this directly in the code. It is essentially
>relying on an undocumented feature of the compiler to work around an
>unspecified (as far as I know) portion of the language (specifically
>whether a warning should be emitted in this case, and what should be done
>to prevent the warning). I wouldn't be at all surprised if there are
>compilers out there that would take one look at "(void)pool" and say "...
>that doesn't do anything!" In any event, the presence of this hack
>certainly doesn't do anything positive for the appearance & readability of
>the code.
>
>
Sure. Of course, I've yet to see a compiler where this hack doesn't
help, but you're right.

>If I were writing Subversion from scratch, I would use a macro for this. It
>permits the behaviour to be turned on and off or changed altogether
>depending on the compiler, which is as it should be since the warning is
>compiler-specific as well.
>
>
There's no need to wait for writing Subversion from scratch -- a patch
will be quite acceptable. :)

>Supposing "(void)pool" really was the source of the problem in Windows (I
>don't see how it could be, but...), perhaps something like this would make
>sense somewhere in a core header:
>
>#if NEED_TO_SUPPRESS_UNUSED_PARAMETER_WARNING
># if defined(WIN32) || defined(WIN64)
>
>
Note that this had nothing to do with the platform (e.g., windows in
this case), but the compiler that's being used ...

># define UNUSED_PARAMETER(x) ((x) = (x))
>
>
... and MSVC does the right thing with (void)foo;

># else
># define UNUSED_PARAMETER(x) ((void)(x))
># endif /* WIN32 || WIN64 */
>#else
># define UNUSED_PARAMETER(x)
>#endif /* NEED_TO_SUPPRESS_UNUSED_PARAMETER_WARNING */
>
>Then, the code that had unused parameters would look like:
>
>void fuu(void *baton, int bar, double qux, ...)
>{
> UNUSED_PARAMETER(baton);
> ...
>}
>
>
+1. Please post a real patch, I'd like to see this in the code, too.
Of course, the name of the macro has to be namespace protected.

>That's just my take on it :-) It makes the behaviour centralized,
>standardized and easy to change, and its usage self-descriptive, obviating
>the need for an explanatory comment of the type shown above by David James.
>
>

Heh. And note, if we'd written Subversion in C++, there'd be no need for
such a macro at all; C++ has a much more elegant way of telling the
compiler that a particular parameter is not being used.

-- Brane

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Oct 19 22:17:15 2005

This is an archived mail posted to the Subversion Dev mailing list.