Changing private symbols to public (was: Re: CVS Update: xc (branch: trunk))

Mike A. Harris mharris at redhat.com
Tue Feb 1 10:20:32 PST 2005


On Tue, 1 Feb 2005, Keith Packard wrote:

>Date: Tue, 01 Feb 2005 09:16:49 -0800
>From: Keith Packard <keithp at keithp.com>
>To: Mike A. Harris <mharris at redhat.com>
>Cc: Daniel Stone <daniel at fooishbar.org>, xorg at lists.freedesktop.org,
>     Keith Packard <keithp at keithp.com>
>Content-Type: multipart/signed; boundary="==_Exmh_563624921P";
>	 micalg=pgp-sha1; protocol="application/pgp-signature"
>Subject: Re: Changing private symbols to public (was: Re: CVS Update: xc
>    (branch: trunk)) 
>
>
>Around 12 o'clock on Feb 1, "Mike A. Harris" wrote:
>
>> Using this logic, every single symbol that isn't explicitly 
>> hidden, is public, and should now be moved to be a public symbol.  
>> That would pretty much expose most if not all internal symbols of 
>> all of the X libraries, as they're not hidden with 
>> attribute((hidden)).
>
>No.  This symbol is used in another library which we publish (libSM).  
>
>That makes all the difference.

Ok, so there is some concensus behind this one change then.  
That was one of the things I was curious about to begin with,
however no public discussion seemingly ever occured, so it wasn't
possible to find that out until I posted my message to the list
asking about why the change went in, etc.

Shouldn't it make sense to investigate all issues of this nature 
right now, and decide wether other libraries should:

- have certain symbols officially exported and documented that 
  are used similarly

- have usage of internal symbols removed if it can be done 
  reasonably

- some other solution I'm not thinking of


It'd also be a good idea to explore the hide-private-symbols
stuff Jakub was working on before, which also has the benefit of
lowering the number of startup relocations that need to be done,
thus cutting down runtime startup costs.  This would have the
benefit of making the internal symbols permanently private as
well, so such symbol abuses don't occur again down the line.

TTYL

-- 
Mike A. Harris, Systems Engineer - X11 Development team, Red Hat Canada, Ltd.

IT executives rate Red Hat #1 for value:  http://www.redhat.com/promo/vendor



More information about the xorg mailing list