Re: cdb self-tests, was Re: GCC 3.2 errors

On Mon, May 31, 2010 at 02:15:48PM +0100, Chris Pugh wrote:

> > Anyone has an update on how to fix the following error?
> > cdbmake: fatal: unable to read input: bad format
> 
> I recall seeing this as possibly being related to,
> 
>  - a machine where softlimit -d doesn't work properly, if at all.
> 
> -  a malloc()  issue wih Linux/glibc systems.
> 
> The original poster being 'Peff'?

I think you mean this thread:

  http://thread.gmane.org/gmane.comp.djb.cdb/123

but the situation has come up multiple times before:

  http://thread.gmane.org/gmane.comp.djb.cdb/77
  http://thread.gmane.org/gmane.comp.djb.cdb/88

The original poster in this thread doesn't say if he is trying to run
the self-tests, or encountering this error during regular use.

The softlimit issue affects only the tests. Basically a certain test
should cause cdbmake to abort due to a softlimit, but it doesn't. The
result is that cdbmake continues to run and finds a bogus point in the
input (which is not a bug; the input really is bogus, it was just never
supposed to get there in the test). If it's simply the test issue, then
it can be ignored (unless you really want to use softlimit, in which
case you might try "softlimit -m" as Dan suggests in the third thread
above).

If he is seeing this problem in actual use, then his input probably is
bad in some way. But I can't say how without seeing the input.

-Peff