On Fri, 24 Sep 2010, Greg Stein wrote:
> On Fri, Sep 24, 2010 at 15:04, Dani Church <dchurch_at_cheri.shyou.org> wrote:
>> I wanted to suggest using a preprocessor directive to target a
>> codebase to a specific API version, something like the following:
>>
>> #define USE_SVN_API 105
>> #include <svn_client.h>
>>
>> I figure that could have two effects: first, it wouldn't set SVN_DEPRECATED
>> for any functions current as of that API version, and second, it could use
>> #pragma GCC poison [1] (or an equivalent for other compilers) to mark any
>> functions introduced in a later version as compile errors.
>
> This is an interesting thought. I will note, however, that this would
> imply a large set of SVN_DEPRECATED macros for us to use. For example,
> SVN_DEPRECATED_1, ... SVN_DEPRECATED_7 to indicate a specific function
> was deprecated in 1.1 through 1.7. Those macros would evaluate
> differently, based on the USE_SVN_API value. This also implies going
> back over all the functions and using the correct DEPRECATED macro on
> it (most of the time, the version is in the comments, but not
> rigorously).
>
>> Does that make sense? I could go through and write the patch, but I wanted
>> to see if there was any buy-in from the community.
>
> That would be a pretty big patch. I see almost 400 deprecated
> functions in our headers. Each of those would need to be touched.
>
> I'm not against the concept in general, but would like to see how all
> those DEPRECATED macros would turn out, concretely.
Well, I have to say I'm interested to see how it would turn out, too.
I'll do a little bit of work on it when I get the chance and send a patch
or possibly just some code, depending on how involved it is.
-Dani Church
Received on 2010-09-24 23:39:07 CEST