Range check error with TSysLogServer

[djhfwk 0]

The UdpDataAvailable procedure should add len check before setlength,

sometimes FUdpSocket.RcvdCount will return 0, and this will cause range error in delphi 11.1, although this can be turn off by uncheck the Range Check option in Project Options.

 

procedure TSysLogServer.UdpDataAvailable(Sender: TObject; ErrCode: Word);
var
    RawMessage : AnsiString;
    Len        : Integer;
    Src        : TSockAddrIn;
    SrcLen     : Integer;
    SrcIP      : AnsiString;
    SrcPort    : AnsiString;
begin
    SrcLen := SizeOf(Src);
    Len    := FUdpSocket.RcvdCount;

    if Len = 0 then  //this should be added, but will cause the UdpDataAvailable event hang randomly
        Exit;

    SetLength(RawMessage, Len);
    Len    := FUdpSocket.ReceiveFrom(@RawMessage[1], Length(RawMessage),
                                     Src, SrcLen);
    if Len < 0 then
        Exit;
    SetLength(RawMessage, Len);
    SrcIP   := IcsStrPas(inet_ntoa(Src.sin_addr));
    SrcPort := AnsiString(IntToStr(ntohs(Src.sin_port)));
    if Assigned(FOnDataAvailable) then
        FOnDataAvailable(Self, SrcIP, SrcPort, RawMessage);
end;

 

 

Edited March 31, 2022 by djhfwk

[David Heffernan 2288]

The correct solution, in my view, is to pass Pointer(RawMessage) rather than @RawMessage [1]

[djhfwk 0]

9 minutes ago, David Heffernan said:

The correct solution, in my view, is to pass Pointer(RawMessage) rather than @RawMessage [1]

🤣

seems the problem is in FUdpSocket.RcvdCount , when it return 0,  i changed it to 65535 manually, FUdpSocket.ReceiveFrom still can successfuly got data

[Angus Robertson 519]

That function should not be using RcvdCount, it may not be accurate when it is called.  Most components receive data into a fixed size buffer, not a string, I'll change it next week. 

 

Angus

 

[David Heffernan 2288]

2 hours ago, djhfwk said:

🤣

seems the problem is in FUdpSocket.RcvdCount , when it return 0,  i changed it to 65535 manually, FUdpSocket.ReceiveFrom still can successfuly got data

Maybe so. But my point about how to deal with getting a pointer from a string that may be zero length stands. So long as the use of the pointer respects the zero length and doesn't try to de-reference it. 

[Angus Robertson 519]

I agree Pointer() would be better here, and probably in many other places in ICS...

 

Angus

 

[Fr0sT.Brutal 897]

23 hours ago, David Heffernan said:

The correct solution, in my view, is to pass Pointer(RawMessage) rather than @RawMessage [1]

I too came to this solution and changed almost all "[1]" occurrences in my projects