And that was the problem!!! Each time I recompiled (and finally got
everything to compile properly), I did a gmake install. What I didn't
realize was that, instead of overwriting the previous installation with the
new installation, it would simply skip it. Hence I was using a new
libpgtcl.so with a postgres not compiled for it.
Soooo...when I deleted the existing postgres installation and did a gmake
install, it worked fine.
Could use a gmake uninstall...
Tim
Zeugswetter Andreas SB <ZeugswetterA@wien.spardat.at> said:
> Looks like your postgres binary is not the correct one.
> That is: the make step for the target postgres.imp failed or went into the
> wrong directory, and you thus have a postgres binary that is not converted
> to export all symbols.
>
> This is one point that bugs me in the aix port. Maybe we should do it
> like other ports do it, first create a postgres.a then postgres.imp
> and finally postgres from postgres.a and postgres.imp.
>
> Comments ?
> Andreas
>
> PS: I do not think that you can successfully run all regression tests
> with your current installation. Can you confirm ?
>
>
>
> So when I try "load libpgtcl.so", I get the following error:
>
>
> Error: couldn't load file "./libpgtcl.so": 0509-130 Symbol resolution
> faile
> d for ./libpgtcl.so because:
> 0509-136 Symbol __start (number 0) is not exported from
> dependent module /home/postgres/postgresql/bin/postgres.
> 0509-192 Examine .loader section symbols with the
> 'dump -Tv' command.
>
>
> I did the dump command on the postgres executable and found:
>
> [177] 0x20022eb0 .data ENTpt DS SECdef [noIMid] __start
>
>
--
Tim Dunnington
Sr. Integration Engineer
Healthcare.com
804-239-7650