[elephant-devel] stack overflow during deserialization

Andrew Philpot philpot at ISI.EDU
Mon Jan 23 18:41:51 UTC 2006


Hi, All.

Allegro 6.2 on Linux.  FWIW, the feature combination (and allegro ics)
returns true in this session.

After a very long (10+ hours) load of data into elephant (1.2 GB of
data loaded), I have encountered a stack overflow error in elephant in
(CL:LABELS ELE::DESERIALIZE ELE::%DESERIALIZE), i.e., the LABELS
function named ELE::%DESERIALIZE inside ELE::DESERIALIZE.  

The original call was to asking for (the relatively benign):

  (ELE:GET-FROM-ROOT '(:WNVFRM :CORE "Frame:08"))

whose return value should be a hash table mapping Lisp keywords to
lists, or may not yet exist.

I couldn't do much hacking here in the error handler, but one thing
that I did see was that when ELE::DESERIALIZE was run interpreted, I
did not get any stack overflow, but when compiled, I sometimes did.  I
notice that the definition has highly aggressive optimization
declarations.  I'm going to take them out, recompile and retry my data
load.

Any suggestions or insights would be appreciated.

Andrew P.




More information about the elephant-devel mailing list