Memoy leak using QtCharts

Christian Ehrlicher

@aawawa said in Memoy leak using QtCharts:

But it is supposed to be deleted 100000 times in my case no when I use RemoveAllSeries no?

You're right. I think I found the culprit:

qlineseries.cpp:112

QLineSeries::QLineSeries(QObject *parent)
    : QXYSeries(*new QLineSeriesPrivate(this), parent)

this is never deleted... :/

JonB

@Christian-Ehrlicher
I think this is the second time I've seen that *new cause this. When will people decide that C++ is getting too clever for its boots? ;-)

SGaist

Hi,

@Christian-Ehrlicher said in Memoy leak using QtCharts:

QLineSeriesPrivate

~~It's QObject based, so when the QXYSeries object is deleted it should also get deleted.~~

It's indeed QObject based but stored in a QScopedPointer so it will automatically get delete on object's destruction.

aawawa

@Christian-Ehrlicher Thank you. It's the first time I see *new in c++ what does it mean?

mrjj

@aawawa
Hi
Its a Dereference
http://www.cplusplus.com/doc/tutorial/pointers/
See section Dereference operator (*)

aawawa

@mrjj yes... I understand... But I don't understand the location of the asterix (*). I am used for example to write

int *a = new int;

what does

*new int

means?

mrjj

@aawawa
Means
value pointed to by ptr
The actual value.

int copy = *SomeIntPtr;
that will copy the value to the copy var.
if you did
int copy = SomeIntPtr;
you get the address of pointer into copy

Christian Ehrlicher

Ok, no memleak at all. You testcase is wrong / inaccurate.
The problem with your testcase is that your eventloop is not (yet) running and that you also call it in an loop without a chance for the eventloop to do it's work. Since the internals of QLineSeries is cleaned up with a deferred delete it is not cleaned up until the eventloop is called (and after app.exec()). That's also the reason I thought the dtor is not called at all.

JosuGZ

@Christian-Ehrlicher said in Memoy leak using QtCharts:

Ok, no memleak at all. You testcase is wrong / inaccurate.
The problem with your testcase is that your eventloop is not (yet) running and that you also call it in an loop without a chance for the eventloop to do it's work. Since the internals of QLineSeries is cleaned up with a deferred delete it is not cleaned up until the eventloop is called (and after app.exec()). That's also the reason I thought the dtor is not called at all.

It seems to me that there are some memory leaks happening here.

This code increases memory usage continuously:

aSeries->remove(aSeries->points().size() - 1);
aSeries->append(timestamp, mean);

Bit this one does not:

aSeries->replace(aSeries->points().size() - 1, timestamp, mean);

I'm using Qt 5.11.2.

JonB

@JosuGZ
In the last post 4 months ago made by @Christian-Ehrlicher, he explained that memory would only be recovered (and hence [hopefully] not permanently grow) when the OP allowed his code to reach the main event loop. I don't know how replace works, but remove surely will do a deferred delete. So in your example you need to be clear/show us how your code hits the event loop each time after do the remove/append?

J.Hilk

hi @JosuGZ and welcome

if you only monitor the memory your OS gives your application, then that is inaccurate.

You have no influence over when the os decides that your freed memory will no longer be reserved for your application.

Therefore
remove -> append is memory is "freed" and a new allocation happens
replace -> new data overwrites old data

JosuGZ

@J.Hilk said in Memoy leak using QtCharts:

hi @JosuGZ and welcome

if you only monitor the memory your OS gives your application, then that is inaccurate.

You have no influence over when the os decides that your freed memory will no longer be reserved for your application.

Therefore
remove -> append is memory is "freed" and a new allocation happens
replace -> new data overwrites old data

An order of magnitude more RAM needed for a simple app is not an OS problem.

Two tests more replacing the offending line, one with small allocations:

char *test = (char *)malloc(10);
test[5] = 't'; // Touching memory
aSeries->replace(aSeries->points().size() - 1, ms, mean);
free(test);

Another with bigger allocations:

char *test = (char *)malloc(10000);
test[5000] = 't'; // Touching memory
aSeries->replace(aSeries->points().size() - 1, ms, mean);
free(test);

Both keep memory usage low as expected.

Even this keeps the memory constant:

QWidget *test = new QWidget;
aSeries->replace(aSeries->points().size() - 1, ms, mean);
test->deleteLater();

Something is happening with this library and I'm not the only one with issues.

thiagohd

Is this fixed in any way? I seem to have the same problem.

mrjj

@thiagohd
Hi
I dont think I bug report was ever created or it was further looked at.