ABC - try, catch - Finally!
Alen Balja
privatefunction simpleTryCatchFinally():void
{
try
{
trace("trying");
}
catch(e:Error)
{
trace("caught");
}
finally
{
trace("finally")
}
}
It creates:
67 operations
2 exceptions, one of which is Error, one of which is 'null'
9 labels
No 'finally' block, but two 'try-catch', as they would be nested somehow
Bytecode:
operandStack: 7
scopeStack: 7
localCount: 4
2 exception(s):
(L0, L1) => L2, AbcQName('Error,AbcNamespace(22,')), AbcQName('e,AbcNamespace(22,'))
(L3, L4) => L5, AbcQName('null,AbcNamespace(0,'null)), AbcQName('null,AbcNamespace(0,'null))
67 operation(s):
+0|-0 DebugFile('/Users/alenbalja/Projects/BBC/code2/ActionScriptTestCases/src;;EmptyFunction.as)
+0|-0 DebugLine(38)
+1|-0 GetLocal(0)
+0|-1 PushScope()
+1|-0 NewActivation()
+2|-1 Dup()
+0|-1 SetLocal(1)
+0|-1 PushScope()
+0|-0 Debug(1,'simpleTryCatchFinally$0,0,38)
try {
L3: +0|-0 DebugLine(44)
try {
L0: +1|-0 FindPropStrict(AbcQName('trace,AbcNamespace(22,')))
+1|-0 PushString('trying)
+1|-2 CallProperty(AbcQName('trace,AbcNamespace(22,')),1)
+0|-1 Pop()
} catch {
(AbcQName('e,AbcNamespace(22,')), AbcQName('Error,AbcNamespace(22,'))) => L2
}
L1: +0|-0 DebugLine(47)
+0|-0 Jump(L6)
L2: +1|-0 GetLocal(0)
+0|-1 PushScope()
+1|-0 GetLocal(1)
+0|-1 PushScope()
+1|-0 NewCatch((L0, L1) => L2, AbcQName('Error,AbcNamespace(22,')), AbcQName('e,AbcNamespace(22,')))
+2|-1 Dup()
+0|-1 SetLocal(2)
+2|-1 Dup()
+0|-1 PushScope()
+2|-2 Swap()
+0|-2 SetSlot(1)
+0|-0 DebugLine(49)
+1|-0 FindPropStrict(AbcQName('trace,AbcNamespace(22,')))
+1|-0 PushString('caought)
+1|-2 CallProperty(AbcQName('trace,AbcNamespace(22,')),1)
+0|-1 Pop()
+0|-0 PopScope()
+0|-0 Kill(2)
L6: +1|-0 PushByte(-1)
} catch {
(AbcQName('null,AbcNamespace(0,'null)), AbcQName('null,AbcNamespace(0,'null))) => L5
}
L4: +0|-0 Jump(L7)
L5: +1|-0 GetLocal(0)
+0|-1 PushScope()
+1|-0 GetLocal(1)
+0|-1 PushScope()
+1|-0 NewCatch((L3, L4) => L5, AbcQName('null,AbcNamespace(0,'null)), AbcQName('null,AbcNamespace(0,'null)))
+2|-1 Dup()
+0|-1 SetLocal(2)
+0|-1 PushScope()
+0|-0 PopScope()
+0|-0 Kill(2)
+1|-1 CoerceAny()
+0|-1 SetLocal(3)
+1|-0 PushByte(0)
+0|-0 Jump(L7)
+0|-0 Label()
+0|-1 Pop()
L9: +0|-0 Label()
+1|-0 GetLocal(3)
+0|-0 Kill(3)
+0|-1 Throw()
+1|-0 PushByte(-1)
+0|-0 PopScope()
+0|-0 Kill(2)
L7: +0|-0 DebugLine(55)
+1|-0 FindPropStrict(AbcQName('trace,AbcNamespace(22,')))
+1|-0 PushString('<finaly<)
+1|-2 CallProperty(AbcQName('trace,AbcNamespace(22,')),1)
+0|-1 Pop()
+0|-1 LookupSwitch(L8,[L9])
L8: +0|-0 DebugLine(57)
+0|-0 ReturnVoid()
pleclech
Finally is a little more complicated as it has to been always
executed
Patrick
On Feb 24, 11:01 pm, Alen Balja <alen.ba...@gmail.com> wrote:
> So, I made try-catch work with my code and the next step is to finally
> implement - finally. By reverse engineering and observing ABC it seems
> pretty complex (at the end of the email). It looks like one try/catch is
> first created, and then another one is applied around it. Something like
> this algorithm, am I remotely correct?
>
>
> CallProp('trace)
>
> *2. Apply first try/catch*
>
> L0
> CallProp('trace)
> *AbcQName('Error) => L2*
> Jump(L3)
> L2
> // some code in catch block
> L3
> ReturnVoid()
>
>
> *L4*
> L0
> CallProp('trace)
> *AbcQName('Error) => L2*
> Jump(L3)
> L2
> // some code in catch block
> L3
> ReturnVoid()
> *L5*
> *Jump(L7)*
> *L6*
> *//some code in finally block*
> *L7*
> *ReturnVoid()*
> *
> *
> *
> *
> *
> *
> *
> *
> *
> *
> *Reverse Engineered Example *
> *
> *
Alen Balja
pleclech
catches block because you can have exception that have not been trap
by your try/catch and finally block must be executed even in this
case.
2. in your catch you can have a return , so you have to jump to the
code to be executed by the finally block and then do your return
3. you can have also a throw into your catch etc...
All these flows are represent by the compiler by a state machine
(everywhere you see a PushByte(xx)) and then dispatch using the
lookupswitch.
I suggest you to study how this is resolved by the compiler outputing
the dot graph from the apparat dump command and looking at the graph
produced using graphviz or another DOT graph viewer.
Best,
Patrick
Alen Balja
Alen Balja
and you have to use -bc cfg option too. So do I have to copy paste the
graph out of that output? There seem to be several of them nested.
cat graph.dot | grep graph
digraph G {
graph[
digraph G {
graph[
digraph G {
graph[
digraph G {
graph[
digraph G {
graph[
digraph G {
graph[
digraph G {
graph[
digraph G {
graph[
digraph G {
graph[
digraph G {
graph[
Joa Ebert
If you want only a single graph for a specific method you have to do
that manually using Scala (maybe 5 lines of code).
Best,
Joa
Alen Balja
dump -i myswf.swf -abc -bc cfg
is unreadable by GraphWiz. It's structure also doesn't seem to make sense in that form. However, there are structures inside that might resemble some dot graphs. I don't know how to easily extract them apart from manual copy paste, which could be extraordinary tedious for huge files.
Alen Balja
pleclech
look at the file within an editor. Each graph start with digraph G
{ and end with }
On Feb 26, 1:49 am, Alen Balja <alen.ba...@gmail.com> wrote:
> Even a single class with a constructor and single function seem to produce
> what seems to be 8 nested graphs.
>
> cat EmptyFunction.swf.abc0.txt | grep graph
> digraph G {
> graph[
> digraph G {
> graph[
> digraph G {
> graph[
> digraph G {
> graph[
>
>
>
>
>
>
>
> On Fri, Feb 25, 2011 at 7:06 PM, Alen Balja <alen.ba...@gmail.com> wrote:
> > The file as it gets produced by
>
> > is unreadable by GraphWiz. It's structure also doesn't seem to make sense
> > in that form. However, there are structures inside that might resemble some
> > dot graphs. I don't know how to easily extract them apart from manual copy
> > paste, which could be extraordinary tedious for huge files.
>
> > >>> Which dump parameter produces dot graph? -swf, -abc, -uml,.... ?
>
>
> read more »