Memory Leak on QNetworkAccessManager

[Christian Ehrlicher]

Apart from the really strange code construct - task manager is not a tool to measure the real memory footprint of an application and it really is not a tool to check for memory leaks...
valgrind on linux does not find anything which looks like a memleak here.

[Maickonn Richard]

Hi @aha_1980 .

I did it this way because the QNetworkAccessManager is asynchronous. I need to get it to wait for finalization or receive a timeout after a certain time, in the example case, 10 seconds.

Could guide me if there's a better way to do this?

[Maickonn Richard]

@Christian-Ehrlicher

Really the Windows task manager is not the best tool to check this problem. But I did not need another tool to identify that there is a leak, if there is one. It is very strange to see the program consuming RAM this way, this in the long run can be disastrous.

[aha_1980]

@Maickonn-Richard

I'd use a QTimer with 10 seconds timeout. When timer slot is fired, check if the request is finished, otherwise abort it.

Am I missing something?

[Maickonn Richard]

@aha_1980

True, I had forgotten that part. I'll do this and see if the problem continues.

[Christian Ehrlicher]

@Maickonn-Richard said in Memory Leak on QNetworkAccessManager:

Really the Windows task manager is not the best tool

It's not even a tool to catch such kind of errors in any way. It may be a hint but nothing more... and since valgrind (which is a tool to catch exactly such kind of problems) does not show anything... but do whatever you like.

[Maickonn Richard]

@Christian-Ehrlicher

I thank you very much for your help with Valgrind! unfortunately currently I can not access this tool.

[J.Hilk]

@Maickonn-Richard

If I may,

this is one way, how you can make use of the assynchon function:

QString getHTML(QString url)
{
    QNetworkRequest request;
    request.setUrl(QUrl(url));
    request.setRawHeader("User-Agent", "Mozilla/5.0 (Windows NT 6.1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/41.0.2228.0 Safari/537.36");
    request.setAttribute(QNetworkRequest::FollowRedirectsAttribute, true);
    
	QNetworkAccessManager *manager = new QNetworkAccessManager(this);
	connect(manager, &QNetworkAccessManager::finished, this, [=](QNetworkReply *reply){
		if(reply.error() == QNetworkReply::NoError)
			emit htmlRead(QString(reply->readAll());
		else
			emit htmlError(reply->error());
                manager->deleteLater();
                reply->deleteLater();
	});
	
	manager->get(request);
	
	QTimer *timeout = new QTimer(this);
        timeout->start(10000);
	connect(timeout &QTimer::timeout, manager, [=]{
		emit htmlTimeout();
		manager->deleteLater();
		timeout->deleteLater();
	});
	
	connect(manager, &QNetworkAccessManager::finished, timeout, QTimer::deleteLater);
}

You'll have to create slots that connect to the signals void htmlTimeout() , void htmlError(QNetworkReply::NetworkError) and void htmlRead(QString) and proceed from there.

This should work, but it's untested.

[aha_1980]

@J.Hilk very nice example!

And it perfectly shows how well signals works together with lambdas...

[J.Hilk]

@aha_1980
thanks, sadly I leaked the QNetworkReply *reply in the example x).
Fixed it.

[Maickonn Richard]

Well guys, after leaving the software running for a long period with the code I posted, I noticed that the consumption of RAM stabilizes. That is, leaks are not happening.

I thank all who cooperated, the help of you all were immense, you gave me a number of alternative ways in which I had not considered before.

A big hug to everyone!

[May May]

@aha_1980 Can you explane it. I have same problem with method PUT. Thank you so much!

[jsulm]

@May-May Hi!
Explain what?