Re: [Boost-users] [assert] Control Flow Warnings with BOOST_ASSERT(false)

From: Michiel Helvensteijn (m.helvensteijn_at_[hidden])
Date: 2008-07-09 10:08:08

Nat Goodspeed wrote:

> In general, I think the major issue with changing macro contents from a
> simple statement to a compound statement is that it can have peculiar
> effects on code surrounding macro invocations. For instance (contrived
> example):
> 
>      if (! data.valid())
>          BOOST_ASSERT(false);
>      else
>      {
>          // ...process 'data'...
>      }

Yes, I see. The semicolon would be a second empty statement. It would cause
a compiler error.

> The classic workaround is to wrap the compound statement in a no-op 'do'
> statement:
> 
> #define MACRO(stuff) do /* intended macro contents */ while (0)

Of course. the do-loop expects a semicolon as part of the statement. I've
made the change. It works fine.

> If there are special constraints unique to BOOST_ASSERT, I don't know
> them.

I see two possible reasons not to use this construction.

1. The programmer may not always want to abort (or throw) at an assertion
failure. That's why I propose to make this behavior optional.

2. It may somehow become unsafe in multi-threaded programs. But I don't
think this is a big issue.

I would like to hear from the developer on this.

-- 
Michiel Helvensteijn