[Openmcl-devel] Weird, non-reproducible bug in 0.14.2-050219

[David Steuber]

On Mar 1, 2005, at 12:10 PM, Gary Byers wrote:
>
> On Sun, 20 Feb 2005, David Steuber wrote:
>
>> I sometimes get this condition signaled when building my application:
>>
>> Duplicate key: #<Package "CL-OPENGLU"> in #<HASH-TABLE :TEST EQ size
>> 252/338 #x60612CE> ((T . "Package for the glu* functions and such"))
>> ((T . "Package for the glu* functions and such")) 117 201
>>
>> Trying again seems to make it go away.  I haven't been able to get it
>> to happen just loading the file with the relevant package definition.
>> My build is run from a shell script, so presumably threads are not
>> involved.
>>
>> Any ideas on how I might narrow this down into a useful test case?
>>
>
> Sorry that I missed this last week.

No problem.

> Is the string "Package for the "glu* functions and such" the
> documentation string for the CL-OPENGLU package?

Yes it is.  I've had this error crop up for a different package also.  
In both cases, simply trying to build again (a process launched from a 
shell script) works.  I haven't seen this error the past few times 
though and I've done a couple CVS updates since on the ccl-dev branch.  
The only thing in common with the two packages is that they have a lot 
of symbols exported.

> (If so, that wouldn't explain what's triggering a duplicate key
> error, but might help to narrow things down a bit.)

If I was doing something stupid like trying to export the same symbol 
twice, would this trigger the error?  Also, wouldn't the error show up 
every single time?

It's the intermittent nature of the error that is so odd.