Majority of PLists are now created as automatic/stack objects,
rather than on heap and initialized either with the known-capacity
reserved upfront or directly from an initializer_list (so there's no
wasted slack in the memory that gets allocated for lists containing
a fixed/known number of elements).
Added versions of the ConnectionEvent/QueueEvent methods that take
a val_list by value.
Added a move ctor/assign-operator to Plists to allow passing them
around without having to copy the underlying array of pointers.
The result of the |x| operator for interval and time types historically
returned a value of type double. This was changed as part of
3256ac7c49 to return interval/time, but
this now reverts to returning a double again to avoid introducing a
change that may break user code.
Fixes GH-219
Scripting errors/mistakes now consistently generate a runtime error
which have the behavior of unwinding the call stack all the way out of
the current event handler.
Before, such errors were not treated consistently and either aborted
the process entirely or emitted a message while continuing to execute
subsequent statements without well-defined behavior (possibly causing
a cascade of errors).
The previous behavior also would only unwind out of the current
function (if within a function body), not out the current event
handler, which is especially problematic for functions that return
a value: the caller is essentially left a mess with no way to deal
with it.
This also changes the behavior of the startup/initialization process
to abort if there's errors during bro_init() rather than continue one
to the main run loop. The `allow_init_errors` option may change this
new, default behavior.
When 'x' is an integral arithmetic expression, it's now coerced to
yield a signed integer before taking the absolute value of it to
prevent the common issue of unsigned integer overflow/wraparound for
values below zero.
Using a time or interval value/expression for 'x' now also yields a
time or interval, respective, from the |x| operation instead of
a double.
* origin/topic/jsiwek/val_mgr:
Pre-allocate and re-use Vals for bool, int, count, enum and empty string
Preallocate booleans and small counts
I added a tiny change to CompHash to make sure that nothing messes this
up in the future.
While expiring a table, DoExpire checks at the end to see if NextEntry
returned nothing to determine if it should sleep for the short
table_expire_delay or the long table_expire_interval.
However, the check to see if the expire_func deleted the entry
re-assigns the same variable. This means that:
If you have a large table that is behind on expiring values
& The table defines an expire_func
& That expire_func deletes the item
& It so happens that the last item checked in the batch of
table_incremental_step size had expired
then DoExpire will reset the cookie and sleep for table_expire_interval
* TypeType values are now printable and yield the type name/alias
* Fix record_fields BIF to return correct type name for fields
* Allow TypeType values that point to a RecordType to be used with
record_fields BIF
* origin/topic/vern/set-ops2:
documentation, test suite update
implemented set relationals
bug fix for set intersection
set intersection implemented
mirroring previous topic/vern/set-ops to get branch up to date, since I'm a n00b
Fixed a couple memory leaks and added a leak test
This commit marks (hopefully) ever one-parameter constructor as explicit.
It also uses override in (hopefully) all circumstances where a virtual
method is overridden.
There are a very few other minor changes - most of them were necessary
to get everything to compile (like one additional constructor). In one
case I changed an implicit operation to an explicit string conversion -
I think the automatically chosen conversion was much more convoluted.
This took longer than I want to admit but not as long as I feared :)
The expiration attribute expression is now evaluated for every use. Thus
later adjustments of the value (e.g. by redefining a const) will now
take effect. Values less than 0 will disable expiration.
I was hoping to report this right at startup through a static check
but turns out we don't have the right machinery in place for that.
That would need to be done after the AST has been finalized, but our
AST traversal code can't iterate over types. So instead I've changed
this so that it's still being reported at runtime but at least
doesn't crash Bro anymore.
Closes BIT-1597.
* 'topic/johanna/filter_subnet_table' of https://github.com/J-Gras/bro:
Fixed &read_expire for subnet-indexed tables
Added &read_expire testcase for subnet tables
Includes a bit of refactoring of commit code & code related to the
feature.
This bif works similar to the matching_subnet bif. The difference is
that, instead of returning a vector of the subnets that match, we return
a filtered view of the original set/table only containing the changed
subnets.
This commit also fixes a small bug in TableVal::UpdateTimestamp
(ReadOperation only has to be called when LoggingAccess() is true).
Specifically observed when redef'ing the same index of a table that uses
subnets as indices, though the bug seems like it applies more generally
to anytime TableVal::Assign is provided with just the HashKey parameter
and not the index Val.
Addresses BIT-1202.
SerializationFormat::EndWrite now transfers ownership of the buffer
to the caller instead of doing a memcpy.
ChunkedIO::Chunk is no longer a POD type, hopefully the ctor/dtor
make it easier to manage its associated memory. It also now
tracks how to deallocate its buffer (i.e. delete vs. free).
This changes the internal type that is used to signal that a vector
is unspecified from any to void.
I tried to verify that the behavior of Bro is still the same. After
a lot of playing around, I think everything still should worl as before.
However, it might be good for someone to take a look at this.
addresses BIT-1144
