On 7 September 2011 00:13, Peter Geoghegan <peter@2ndquadrant.com> wrote:
> * Within TUs, we unshadow a previously shadowed variable, so we link
> to a global variable rather than one local to the original/other new
> file. Unlikely to be a problem. Here's what I get when I compile
> xlog.c in the usual way with the addition of the -Wshadow flag:
I hit send too soon. Of course, this isn't going to matter in the case
I described because an extern declaration of int foo cannot appear in
the same TU as a static declaration of int foo - it won't compile. I
hastily gave that as an example of a general phenomenon that can occur
when performing this splitting process. An actually valid example of
same would be if someone refactored functions a bit as part of this
process to make things more modular, and now referenced a global
variable rather than a local one as part of that process. This is
quite possible, because namespace pollution is a big problem with
heavyweight C files - Just look at how much output that -Wshadow flag
gives when used on xlog.c.
--
Peter Geoghegan http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training and Services