Re: --with-tcl build on AIX (and others) fails - Mailing list pgsql-hackers

From Zeugswetter Andreas SB SD
Subject Re: --with-tcl build on AIX (and others) fails
Date
Msg-id 46C15C39FEB2C44BA555E356FBCD6FA41EB496@m0114.s-mxs.net
Whole thread Raw
In response to --with-tcl build on AIX (and others) fails  ("Zeugswetter Andreas SB SD" <ZeugswetterA@spardat.at>)
Responses Re: --with-tcl build on AIX (and others) fails  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
> I don't think we dare risk a wholesale change in the way 
> pltcl is built
> for 7.2.  We're already through the ports testing process for 7.2, and
> I don't want to start it over again.

Ok, that is what I asked for, that you look at it, thanks. 7.3 then.
(Of course there have been other portability problems introduced recently, 
so ...)

This leaves the question how many port testers actually built 
and tested --with-tcl and actually tried a load and pg_connect in tclsh. 
My guess is not too many.
Imho we should start to require port reports to contain certain --with 
features, like perl, tcl, ....

> As for the libpgtcl part of the patch, I don't understand why the Tcl
> library should be linked into libpgtcl.  libpgtcl is supposed to be
> loaded into a Tcl interpreter, not carry its own interpreter 
> along with
> it; but it sure looks like that's what will happen with this patch.

On AIX this is actually an imports file. Now the tricky part is 
the first line of this file. This line has to be correct.

Since the tcl Interpreter is linked with this same exports file
the library is only supposed to be loaded once.
Does anybody have an idea how to actually test this (without tk) ?

pgtclsh does work, but the shlib with above Makefile cannot be loaded 
by tclsh because it misses a -bnoentry flag :-(

With -bnoentry it does load and pg_connect and pg_exec and pg_result, 
is this a sufficient test ?
This again is probably a bug in Makefile.shlib on aix, since I guess 
all of our shlibs don't have a main function. Unfortunately a different 
flag has to be used on different aix Versions :-( Don't know if we still 
need to worry about AIX 4.2 and below, though.

Andreas


pgsql-hackers by date:

Previous
From: Daniel Kalchev
Date:
Subject: Re: again on index usage
Next
From: "Zeugswetter Andreas SB SD"
Date:
Subject: Re: Time as keyword