Re: defined(IRIX)

[Thomas Dickey]

On Sat, Sep 28, 2002 at 01:09:57PM +0200, Peter Eisentraut wrote:
> Thomas Dickey writes:
> 
> > There is form, but no content.  This can be done using AC_TRY_RUN (or other
> > runtime check), is otherwise no better than a random guess.  Alignment 
> > problems
> > don't show up in compiling, but at runtime.
> 
> I'm afraid they do.  The check answers the question "If I put this type
> into a struct, where will the compiler align it?", which is ultimately
> what our code is interested in, and it also happens to correspond to GCC's
> __alignof__ operator, hence the name.
> 
> If you want to answer a question of the form "If I put this type at this
> address, will my program crash?", you can write a different test, but it
> would be a lot more cumbersome and won't get you substantially better
> results.

that's assuming the compiler is correct (which, in the case of gcc, has
not always been true).  But your perspective on the matter probably reflects
the platforms where you've run gcc...

-- 
Thomas E. Dickey <address@hidden>
http://invisible-island.net
ftp://invisible-island.net