[elephant-devel] secondary indices

[Robert L. Read]

OK; I will review your emails and try to find a specific piece of code
that we can run.

I maybe able to instrument the code sufficiently for you to run it and
get something valuable from it.





On Fri, 2006-01-13 at 18:49 -0500, Waldo Rubinstein wrote:

> Robert,
> 
> 
> I've tried to tackle this to no avail. Maybe my knowledge does not
> extend that far yet :(
> 
> 
> Little more help would be needed. Maybe someone else's participation
> could help.
> 
> 
> Thanks,
> Waldo
> 
> 
> On Jan 13, 2006, at 2:32 PM, Robert L. Read wrote:
> 
> 
> > It's not clear to me that it matters.
> > 
> > Going either direction will likely lead to a solution.  I would take
> > Andrew's advice, 
> > since his hunch may have vailidity, and also because it is the first
> > failure in the test suite.
> > 
> > But the course of action would be the same:
> > 
> > 1)  Take the test that fails, copy the code out of it into a
> > separate file where the RT 
> > code won't obscure the problem.
> > 2)  Instrument the code to find the earliest possible error; in this
> > case, that may be 
> > the failure of that instantiation code to let you get to that slot,
> > or maybe some other slot.
> > 3)  Whenever you want, I will take your test code and run it under
> > SBCL.
> > 
> > In this way we should be able to relatively quickly determine the
> > important difference
> > between SBCL and OpenMCL.  Depending on what you find, I will likely
> > need Andrew's
> > advice on how to fix it.
> > 
> > 
> > 
> > On Fri, 2006-01-13 at 14:10 -0500, Waldo Rubinstein wrote: 
> > 
> > > Thanks Andrew. Robert, how do you want to proceed. Where should I  
> > > look for potential CLOS problems?
> > > 
> > > - Waldo
> > > 
> > > On Jan 13, 2006, at 1:40 PM, Andrew Blumberg wrote:
> > > 
> > > > hi robert and waldo,
> > > >
> > > > 	before going too far in debugging the secondary indices part of  
> > > > the problem, i think it's worth trying to figure out the CLOS  
> > > > problem --- i suspect that the rest of the issue might be a symptom  
> > > > of that, and "fixing" the secondary indices stuff might just paper  
> > > > over the real problem.  if you're going to spend time instrumenting  
> > > > code, better to investigate the CLOS test failures first.
> > > >
> > > > 								- andrew
> 
> 
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.common-lisp.net/pipermail/elephant-devel/attachments/20060113/89043924/attachment.html>