Tom Lane wrote:
> Andrew Dunstan <andrew@dunslane.net> writes:
> > On 04/20/2011 01:16 PM, Bruce Momjian wrote:
> >>> This implies to me that we changed something about how we handle this
> >>> since we did the 9.0 runs, but I don't know what it was. Should I?
>
> >> I think Andrew also supplied the typedef list for the 9.0 run.
>
> > Yes. But in November, the server where all my animals were running died.
> > The rebuilt machines all used newer versions of the OS, new compilers
> > and newer tools such as objdump. As I pointed out at the time I
> > committed the new typedefs list, that accounts for a lot of the changes.
>
> It wouldn't surprise me in the least to find out that newer gcc's
> stopped emitting symbol table entries for unreferenced typedefs.
>
> In fact, using HEAD, I get this on my old HPUX box:
>
> (gdb) p sizeof(BulkInsertStateData)
> $65 = 8
>
> and this on my Fedora 13 box:
>
> (gdb) p sizeof(BulkInsertStateData)
> No symbol "BulkInsertStateData" in current context.
>
> (gcc 2.95.3 and 4.4.5 respectively) So the tools definitely changed
> sometime in the last N years.
So the list of possible additions Andrew supplied are cases where we
never reference those typedefs --- seems like a cleanup opportunity.
-- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB
http://enterprisedb.com
+ It's impossible for everything to be true. +