[Monetdb-developers] suggestions on debugging/tracing

mobigital-nbl at yahoo.com mobigital-nbl at yahoo.com
Sun Mar 2 21:13:06 CET 2008


Hi,

I am trying to investigate this bug (https://sourceforge.net/tracker/?func=detail&atid=482468&aid=1905467&group_id=56967)
which i submitted earlier.

I was importing records with COPY TO a staging table in batches of 250K rows. Then copying them into final table.
After completing 4,750,000 rows and importing a new batch thebelow exception occurs in GDKreallocmax at the last line of the snippetbelow:

oldsize = GDK_MEM_BLKSIZE(s); 
with "s"=0x00000000100e0000"
GDK_MEM_BLKSIZE macro tries to read 8 bytes prior to "s", which is  0x00000000100dfff8 and turns out to be unallocated.

I am thinking the h->base of the heap is either free'ed at some point before being used or may be it is set to incorrect value?
in any case i wanted to add some tracing to Heap to maintain some history of base change (and who did the change), so that 
when I hit another access violation during debugging I can use that to pinpoint why Heap base points to unallocated space.
i wanted to see if anyone has already used something like that so that I can get a headstart quicker.

if anyone have any suggestions on fixing it would be appreciated.

Cheers.






More information about the developers-list mailing list