[committed] fix IA-64 as infinite loop with bad input

[James E Wilson]

On Tue, 2005-11-22 at 23:47, Jan Beulich wrote:
> Wouldn't it be more appropriate/consistent to issue an error only when
> manual_bundling, and otherwise fill the entire bundle with nop-s and
> issue only a warning?

I don't see any practical use for this.  No one has ever apparently
written code like this, as we never got any bug reports on user assembly
code.  The bug report we did get was for code that was the output of
testcase reduction software.  It just tries deleting code until a
testcase is as small as it can get it without user help.  Emitting an
error for invalid assembly will not confuse this tool, it knows how to
deal with things like syntax errors introduced by incorrect edits. 
Looking at the existing code, I don't see anyplace where we emit a
warning/error for a bad template (other than the hint in B slot issue
which I think it not relevant here), so I don't think there is any clear
existing policy on whether we should emit a warning or error here.

That said, I'm not opposed to a warning instead of an error, if you want
to write the patch that does this.  Or if you can provide a good enough
example to show why we should support this for end users, in which case
I would be willing to write the patch myself.
-- 
Jim Wilson, GNU Tools Support, http://www.specifix.com