[Qt-interest] QTcpSocket Problem

[BRM]

And don't forget what the Qt docs say about the readRead() signal - it's only emitted once. So you probably want to do a readAll() on the socket each time you receive readyRead().
(Unless you have a protocol that lets you do otherwise!)

$0.02

Ben



----- Original Message ----
> From: Jason H <scorp1us at yahoo.com>
> To: Anthony Gabrielson <agabrielson1 at comcast.net>; Dan Mills <dmills at exponent.myzen.co.uk>
> Cc: qt-interest at trolltech.com
> Sent: Tue, January 19, 2010 10:50:48 PM
> Subject: Re: [Qt-interest] QTcpSocket Problem
> 
> I didn't read the source, but the 37k issue is very familiar to me.
> It sounds like you are expecting continual reads/blocking read from the device. 
> This is hardly the case. The IO buffers seem to be 32k, but there is the 
> interplay of nagel's algorithm I am pretty sure you should just collect data 
> when you get readyRead() signal whose slot is 
> 
> readyReadSlot()
> {
>        if (!outputFile.isOpen())
>        outputFile.open(QIODevice::readOnly);
>     outputFile.write(socket->readAll());
>     if buffer.length()==fileSize)
>     {
>         outputFile.close();
>     }
> 
> }
> 
> with (int) fileSize and (QFile) outputFile, (QTcpSocket) socket being previously 
> initialized.
> 
> This function will  accumipate bytes in the file and close the file when done. 
> The socket will stay open. Just connect it to the readyReadSignal
> 
> 
> 
> 
> ----- Original Message ----
> From: Anthony Gabrielson 
> To: Dan Mills 
> Cc: qt-interest at trolltech.com
> Sent: Tue, January 19, 2010 10:11:48 PM
> Subject: Re: [Qt-interest] QTcpSocket Problem
> 
> 
> On Tuesday, January 19, 2010 9:26 PM, Dan Mills wrote:
> 
> 
> > Your while loop is wrong as it assumes that a given packet will not be
> > fragmented. It is NOT an error for a read from a socket to come up
> > shorter then you were expecting, the rest of the data might be in the
> > next packet, what happens if data is available, but when you call read,
> > you only get one byte (which is perfectly valid behaviour)?. 
> 
> Where does QT's read fit in relative to the OSI model?  I know my code is
> simple - that's actually the point for the time being.  Hopefully I'm at
> least at the transport layer and I don't have to worry about packets
> assuming they are coming in a somewhat regular fashion.  Similar code
> utilizing UNIX read/write, what I'm quasi porting this from, has run across
> fairly large network segments without any issues in the past.  I'm really
> hoping the Qt implementation of read/write isn't that different from UNIX.
> 
> > You need to keep calling read until you have assembled a complete msg,
> > decreasing the size and incrementing the source pointer as you go. 
> 
> Read is populating a structure; if I keep calling it I will have no idea
> what is coming in based on how the code is currently structured.  I know the
> code is a bit simplistic and I would like to make it more robust, I just
> want simple to work first before I get overly bogged down and this
> convention has worked really well for me in the past.  Also depending on
> where Qt's read fits in on the OSI model I'm nowhere near the packet.  I'm
> also not currently having an issue getting to few bytes - I'm either getting
> the correct number or 0.
> 
> > Why is this reporting the file as phd.pdf, when getfile thinks it is
> > file.pdf?
> 
> I'm busted I edited the output for brevity and a bit of content :) and I
> missed one change...
> 
> > Plenty, but without seeing your revised code I am having a crystal ball
> > failure.
> 
> Here is the one updated function:
> 
> int getFile(QTcpSocket &inet_sock)
> {
>     NetMSG     msg;
>     QString filename;
>     qint64 bytesRead;
> 
>     //Get the filename
>     if( (bytesRead = inet_sock.read((char *) &msg, sizeof(msg))) !=
> sizeof(msg)){
>         qDebug() << "getFile: (name) bytes read " << bytesRead << " of " <<
> sizeof(msg);
>         return 1;
>     }
>     filename = QString(msg.mbody).trimmed();
> 
>     qDebug() << "getFile: type " << msg.msize << " size " << msg.mtype;
>     qDebug() << "getFile: filename " << filename;
> 
>     QFile file = filename.trimmed();
>     if(file.open(QFile::WriteOnly) != true){
>         qDebug() << "getFile: local_file open failed";
>         return 2;
>     }
> 
>     //loop over the file until END_DATA is recv
>     while(true){
>         inet_sock.waitForReadyRead(-1);
>         if( (bytesRead = inet_sock.read((char *) &msg, sizeof(msg))) == 0 ){
>             qDebug() << "getFile: only read " << bytesRead << " bytes " <<
> sizeof(msg);
>             file.close();
>             return 3;
>         }
> 
>         qDebug() << "getFile: recv " << msg.msize << " Bytes";
> 
>         switch(msg.mtype){
>             case DATA:
>                 if(file.write(&(msg.mbody[0]), msg.msize) == -1){
>                     qDebug() << "getFile: local write failed";
>                     return 4;
>                 }
>                 break;
>             case END_DATA:
>                 file.close();
>                 qDebug() << "getFile: transfer for file " << filename << "
> completed " << msg.msize;
>                 return 0;
>             default:
>                 qDebug() << "getFile: unknown message type " << msg.mtype <<
> " claims size of " << msg.msize;
>                 qDebug() << "getFile: this is an unrecoverable error,
> goodbye";
>                 file.close();
>                 return 5;
>         }     // end switch
>         memset(&msg,0,sizeof(msg)); //Catch - set mem to zero
>     }       // end while
> 
>     return 6;    //More of a place holder, shouldn't get here.
> }
> 
> Thanks,
> Anthony
> 
> _______________________________________________
> Qt-interest mailing list
> Qt-interest at trolltech.com
> http://lists.trolltech.com/mailman/listinfo/qt-interest
> 
> 
> 
>       
> 
> _______________________________________________
> Qt-interest mailing list
> Qt-interest at trolltech.com
> http://lists.trolltech.com/mailman/listinfo/qt-interest