Menu

Src (SVN)-old Commit Log


Commit Date  
[r4986] by dwarren

Change way terms are copied in creating incr lists of changed terms,
so that the heap is in a valid state if (when) glstack_realloc and
gc_heap might be called. Avoids segfault in gc_heap.

2006-10-11 21:06:22 Tree
[r4985] by tswift

Changes to error_xsb.[ch] to turn current xsb_throw to xsb_throw_internal, and then re-introduce xsb_throw(). This allows xsb_throw to be used more easily
by foreign C functions and will hopefully solve Roberto Bagnara's problem.

Other changes fixed some typos in error messages.

2006-10-08 23:40:39 Tree
[r4984] by kifer

Saikat eliminated many leaks

2006-10-08 17:55:50 Tree
[r4983] by dwarren

Modified consult so that it will consult files with extensions other than
the standard ones that XSB recognizes. (Stimulated by aleph, an ILP program
that wants to consult a file with a .b extension.)
Also fixed to allow these files to be consulted on windows, which doesn't
have a file link operation that is used. It now uses filecopy for windows.
Not great, but better than crashing...

2006-10-06 21:24:34 Tree
[r4982] by dwarren

Updated memory management in hashtable (which manages hashing for incremental
table maintenance.) It now uses XSB mem_alloc and friends, and there is a new
space type of INCR_TABLE_SPACE. It is included in SLG Table space in statistics
and is also broken out (if nonzero).
And doing that showed a huge amount of unused space due to many large hashtables
when there's only one edge. So I changed the minimum size of initial hashtables
(they are automatically expanded if necessary.) [It reduced incr space usage in
my DSS example from about 340M to 8M... much!! better.]

2006-10-06 20:04:28 Tree
[r4981] by beatas

A bug fix in propagation rule's compilation

2006-10-05 20:24:43 Tree
[r4980] by dwarren

Take out earlier attempt to ensure that string gc doesn't access uninitialized
portions of findall chunks. The mem_calloc makes this code redundant (and it didn't
work anyway.)

2006-10-02 14:46:57 Tree
[r4979] by dwarren

Initialized chunks by using mem_calloc (just to see if it shuts valgrind up...)

2006-09-28 15:50:14 Tree
[r4978] by tswift

Changes to get rid of uninitialized memory as reported by valgrind.

1) Initialized IGRhead and deadlock_brk_leader
2) Ensured that tries always use private structure manager
3) initialized num_heap_term_vars to be 0 at each pass through answer_return

The last fixes an obscure memory error in the sequential engine, as well.

2006-09-27 22:03:43 Tree
[r4977] by tswift

Modified test to get rid of a race condition.

2006-09-27 18:49:36 Tree
Older >
MongoDB Logo MongoDB