Re: Variable-length FunctionCallInfoData

Hi,

On 2018-10-09 12:18:02 -0700, Andres Freund wrote:
> Here's an updated version of the patch. Besides a rebase the biggest
> change is that I've wrapped:
>
> On 2018-06-05 10:29:52 -0700, Andres Freund wrote:
> > There's some added uglyness, which I hope we can polish a bit
> > further. Right now we allocate a good number of FunctionCallInfoData
> > struct on the stack - which doesn't quite work afterwards anymore. So
> > the stack allocations, for the majoroity cases where the argument number
> > is known, currently looks like:
> >
> > union {
> > FunctionCallInfoData fcinfo;
> > char *fcinfo_data[SizeForFunctionCallInfoData(0)];
> > } fcinfodata;
> > FunctionCallInfo fcinfo = &fcinfodata.fcinfo;
> >
> > that's not pretty, but also not that bad.
>
> into a macro STACK_FCINFO_FOR_ARGS(varname, numargs). That makes the
> code look much nicer.
>
> I think we should go for this. If there's some agreement on that I'll
> perform a bit more polishing.
>
> I think it'd probably good to add accessors for value/nullness in
> arguments that hide the difference between <v12 and v12, for the sake of
> extension authors. Would probably mostly make sense if we backpatched
> those for compatibility.
>
>
> Also attached is a second patch that avoids all the duplication in
> fmgr.[ch]. While the savings are nice, I'm a bit doubtful that the
> amount of magic here is reasonable. Any opinions?

Any comments? Otherwise I plan to press forward with this soon-ish.

Greetings,

Andres Freund