On 2013-02-04 17:27:39 -0500, Tom Lane wrote:
> Alvaro Herrera <alvherre@2ndquadrant.com> writes:
> > pg_xlogdump needs access to the *_desc functions for each rmgr. We
> > already moved forward quite a bit by splitting those functions out of
> > their containing files; so now they are compilable separately. Good.
> > The remaining task is enabling the code to find those functions in the
> > first place; currently, the function pointers live in rmgr.c which is
> > not compilable by frontend code because it contains pointers to other
> > functions. Hence the attached patch splits RmgrData into two; the names
> > and rm_desc functions go into a new file which can be compiled easily by
> > frontend.
>
> > Proposed patch attached.
>
> This seems pretty ugly to me.
>
> Couldn't we do something similar to the design for SQL keyword constants,
> wherein the actual data is in macros in a header file (providing exactly
> one source of truth for each RM) and then various .c files can #include
> that after #defining the macro as they need? See
> src/include/parser/kwlist.h and the files that include that.
My original patch just surrounded the backend-only functions by a macro
which was defined to NULL in #ifdef FRONTEND. I still think thats the
best way to go. People objected to that though...
Other than that I find duplicating the whole table by far the best
approach. Splitting parts off seems to be more complex than warranted
without actually getting rid of duplication.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services