On 06/29/2013 02:27 PM, Cédric Villemain wrote:
>>
>> I haven't seen a response to this. One thing we are missing is
>> documentation. Given that I'm inclined to commit all of this (i.e.
>> cedric's patches 1,2,3, and 4 plus my addition).
> I did so I sent the mail again. I believe your addition need some changes to
> allow the PGXS+VPATH case as it currently depends on the source-tree tool. So
> it needs to be in postgresql-server-devel packages, thus installed by
> postgresql "make install".
> I'm going to update the doc. It's probably worth to have some examples.
I think we can survive for now without an additional tool. What I did
was a proof of concept, it was not intended as a suggestion that we
should install prep_buildtree. I am only suggesting that, in addition to
your changes, if USE_VPATH is set then that path is used instead of a
path inferred from the name of the Makefile.
A tool to set up a full vpath build tree for extensions or external
modules seems to be beyond the scope of this.
>
>> I'm also inclined to backpatch it, since without that it seems to me
>> unlikely packagers will be able to make practical use of it for several
>> years, and the risk is very low.
> Yes, it is valuable to help packagers here, thanks.
cheers
andrew