apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Justin Erenkrantz <jerenkra...@ebuilt.com>
Subject Re: K&R 6.1.4 [2nd ed] Still Applicable?
Date Thu, 12 Apr 2001 01:59:10 GMT
I've heard reports that HPUX's function pointers *may* cause some problems 
when dealing with C function pointers.  HPUX seems to use indirection in the 
function pointers (they call it PLT or plabel).

This is a recent thread with some HP people which seems to go over the 
problem:

http://devresource.hp.com/CXX/Forums/hpux-devtools/hpux-devtools.0102/0019.html

The only implementation I can find that deals with PLT function pointers
in HPUX is in NACHOS:

http://http.cs.berkeley.edu/~tea/nachos/

See code/threads/thread.cc for more information.

These links seem to be dealing with lower-level implementations than
APR might have to deal with.  This might not be an issue for dlopen...

Others on this list may have more information as to how it would pertain
specifically to APR.  I think Jeff has access to an HPUX box (our HPUX
box at eBuilt was returned).

HTH.  -- justin

On Wed, Apr 11, 2001 at 12:36:55AM -0500, William A. Rowe, Jr. wrote:
> Before I dig into this any further...
> 
> we make a number of assumptions in APR that sizeof(void(*)(void)) == sizeof(void*).
> 
> By K&R "C: A Reference Manual" 2nd ed 6.1.4 Pointer Sizes...
> 
> "Although function pointers are usually no larger than data pointers, there are 
> a few computers on which this is no true ..."
> 
> Is this philiosophy irrelevant for the platforms we support with APR?
> 
> It's mostly an issue in DSOs (possibly not even implemented on such platforms),
> and in the optional_fn code.
> 
> If anyone has a _yes_ or something definative, please speak up.
> 
> Bill


Mime
View raw message