Integral/floating-point division/modulo by zero in C++ is undefined
behavior, so to prevent such cases in a script from crashing Bro,
they're now reported as an error (with script location information) and
the event handler in which it occurred returns immediately.
Closes#983.
* origin/topic/jsiwek/983:
Add named constructor examples to docs.
Allow named vector constructors. Addresses #983.
Allow named table constructors. Addresses #983.
Improve set constructor argument coercion.
Allow named set constructors. Addresses #983.
Allow named record constructors. Addresses #983.
And changed the endianness parameter of bytestring_to_count() BIF to
default to false (big endian), mostly just to prove that the BIF parser
doesn't choke on default parameters.
Closes#946.
* origin/topic/jsiwek/ticket946:
Fix memory leaks resulting from 'when' and 'return when' statements.
Fix three bugs with 'when' and 'return when' statements. Addresses #946
- 'when' statements were problematic when used in a function/event/hook
that had local variables with an assigned function value. This was
because 'when' blocks operate on a clone of the frame and the cloning
process serializes locals and the serialization of functions had an
infinite cycle in it (ID -> BroFunc -> ID -> BroFunc ...). The ID
was only used for the function name and type information, so
refactoring Func and subclasses to depend on those two things instead
fixes the issue.
- 'return when' blocks, specifically, didn't work whenever execution
of the containing function's body does another function call before
reaching the 'return when' block, because of an assertion. This was
was due to logic in CallExpr::Eval always clearing the CallExpr
associated with the Frame after doing the call, instead of restoring
any previous CallExpr, which the code in Trigger::Eval expected to
have available.
- An assert could be reached when the condition of a 'when' statement
depended on checking the value of global state variables. The assert
in Trigger::QueueTrigger that checks that the Trigger isn't disabled
would get hit because Trigger::Eval/Timeout disable themselves after
running, but don't unregister themselves from the NotifierRegistry,
which keeps calling QueueTrigger for every state access of the global.
The RecordVal ctor refs the type arg via the MutableVal -> Val ctors,
so this line was double incrementing the type's ref count, but could
only decrement it once upon the Val's destruction.
These cases should be avoidable by fixing scripts where they occur and
they can also help catch typos that would lead to unintentional runtime
behavior.
Adding this already revealed several scripts where a field in an inlined
record was never removed after a code refactor.
* origin/topic/jsiwek/string-indexing:
Change substring index notation to use a colon (addresses #422).
Tweaked slightly to make it more generic, we may index other types
with slices eventually too.
Closes#422.
The index expression can take up to two indices for the start and end
index of the substring to return (e.g. "mystring[1,3]"). Negative
indices are allowed, with -1 representing the last character in the
string. The indexing is not cyclic -- if the starting index is >= the
length of the string an empty string is returned, and if the ending
index is >= the length of the string then it's interpreted as the last
index of the string. Assigning to substrings accessed like this isn't
allowed.
Both local and global variables declared with "const" could be modified,
but now expressions that would modify them should generate an error
message at parse-time.
For an index expression list, ListExpr::InitVal() passed the TypeList
to Expr::InitVal() for each expression element in the list instead of
the type for that element. This made RecordConstructorExpr::InitVal()
complain since it expects a RecordType and not a TypeList with a
RecordType element as an argument. In most other cases, Expr::InitVal()
worked because check_and_promote() "flattens" the list to a single type.
* origin/topic/jsiwek/table-init-container-ctors:
Add test of record() constructor to table initializer unit test.
Fix table(), set(), vector() constructors in table initializer lists.
Closes#5.
- Identifiers that are initialized with set()/table() constructor
expressions now inherit attributes from the expression. Before,
statements like
const i: set[string] = set() &redef;
associated the attribute with the set() constructor, but not the
"i" identifier, preventing redefinition. Addresses #866.
- Allow &default attribute to apply to tables initialized as empty
(via either "{ }" or "table()") or if the expression supplied to it
can evaluate to a type that's promotable to the same yield type as
the table.
The return value of the call is an implicit boolean value of T if all
hook handlers ran, or F if one hook handler exited as a result of a
break statement and potentially prevented other handlers from running.
Scripts don't need to declare hooks with an explicit return type of bool
(internally, that's assumed), and any values given to (optional) return
statements in handler definitions are just ignored.
Addresses #918.
* origin/fastpath:
Fix the "-=" operator for intervals
Fix "!=" operator for subnets
Add sleeps to configuration_update test for better reliability.
Fix a segfault when iterating over a set
Also removed RefExpr::Eval(Val*) method since it was never called
(Clang emitted warning about this hiding overloaded virtual function
UnaryExpr::Eval(Frame*)) and doesn't appear to be necessary even if it
was called to avoid the default vector handling of UnaryExpr::Eval
(as the comment suggests as the intention).
* master-merge-helper:
possible use after free forbidden
Suppression of unused code
Fix of some memory leaks
removing dead code
A destructor must free the memory allocated by the constructor
Good overridance with the good qualifier
Better use of operators priorities
protection from bad frees on unallocated strings
