Re: Patch: various Syscall changes

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Yao" == Yao Qi <qiyaoltc@cn.ibm.com> writes:

Yao> if (num >= 0 && num < syscallList.length)
Yao>      return syscallList[num];
Yao> else if (num < 0)
Yao>      throw new RuntimeException ("Negative Syscall Number:" +
Yao> Integer.toString(num));

Yeah, I considered this.  I kept it the way it was because I did not
know whether the code was intentionally written to allow negative
syscall numbers.

Yao> IMO, the name of syscall is the ID of it.  We could say they are the
Yao> same syscalls only if their names are the same.

What I was going for here is having Syscall work more or less like an
'enum'.  Of course we can't use enum.  And Syscall has an unusual
need: unrecognized syscalls can be created.

Still, the idea is that each system call is represented by a single
Syscall object.  That's why I added the cache for unrecognized
syscalls, and why equals was written the way it was.

>> * Adds syscallByName.  This is needed for 'ftrace -e' (or "--trace" as
>> I spell it in my impl...)

Yao> What do you want "--trace" to do in ftrace? Only trace a certain
Yao> syscall?

Yes.  I'll send my updated ftrace shortly so you can see.

Yao> I do not think syscallByName is necessary here.  We could do something
Yao> like this in ftrace.java,

Yeah, it isn't strictly necessary.  It just seemed like a reasonably
clean way to implement this functionality, and it was the first thing
that came to mind.

Yao> We will seldom want to know which syscall is named "read" when
Yao> the program does not meet an syscall.  Comments?

I've thought about this a bit more.  syscallByName has the nice
property that we can tell the user if he misspelled a syscalll name
during option processing.  I think that is enough reason to keep it.

Tom