Memory leaks

[John Darrington]

Over the last few weeks I've been hunting down leaks which "make 
check-valgrind" has reported.
There were over 200 tests which reported leaks.  Now the number is just under 
60.  I'd appreciate 
some help getting them fixed.

They seem to fall into 6 distinct categories, manifested in tests:


166, 168, 171: Seem to be caused by the DATASET CLONE command.  Some datasets 
aren't being destroyed.

185:  This seems to be a problem with the psql library.  The PQConnectdb 
routine is allocating memory,
      that the PQFinish is not freeing.  Not much we can do about that and it's 
probably no big deal.
      Maybe we should just file a bug report upstream if it's not fixed in a 
later version.

366:  This is a problem with the crosstabs command.


776,777,778,780,781: Huge amounts of memory leaked from the range-tower-test 
program.  I'm unsure if this
      is a leak in the test program itself, or if it is actually a problem with 
the range-tower implementation.

788, 789: A leak either in sparse-xarray or its test program.  Again, it is 
hard to be sure which.

All the rest: Are tests which use the cairo driver.  The majority of the leaks 
are not our fault, but a leak 
    in the pango library.  But I cannot be sure that they are not hiding other 
problems which are our responsibility.
    We should test with a recent pango library and report a bug to the 
maintainers if the leak  still exists.

-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://keys.gnupg.net or any PGP keyserver for public key.

signature.asc
Description: Digital signature