Use of QTreeWidget causing heap corruption

kshegunov

@eos-pengwern said in Use of QTreeWidget causing heap corruption:

However, I've run the above code through Windows Application Verifier, and Intel Inspector, and no errors get flagged up while it is executing.

Enable all stack smashing dummies and heap-boundary checks your compiler supports (usually compiling in debug mode is enough) and try to get a precise location for the crash. I can't see anything plainly wrong, but in release mode you can write over foreign memory without much of a problem. Put asserts where appropriate to facilitate debug checks.

eos pengwern

@kshegunov Thank you. That is what I'm doing, and the problem is that each time an error occurs it is different and apparently random; usually an access violation thrown from deep within a system DLL, usually relating to a memory location which has no relation to any variable that I've declared as far as I can see, and yet I often notice that others of my own data structures are corrupted (even though they reside at addresses far from the access violation). The only pattern I can spot is that they never happen if I comment out 'redrawTable()'.

VRonin

what is fitItemList ?

eos pengwern

@VRonin I'm sorry; that should be 'theItemList' (I've edited the original question now). A typical mistake when one changes the variable names in the one's original code in the hope of making things clearer!

VRonin

This post is deleted!

VRonin

can you replace theData.getCurrentColumnZeroText(index), theData.getCurrentColumnOneText(index), etc. with some constant text and see if you still get the heap corruption?

eos pengwern

@VRonin I've just tried that and I'm afraid it makes no difference.

VRonin

one more try:
turn QVector<QTreeWidgetItem *> theItemList; into QVector<QPointer< QTreeWidgetItem > > theItemList;

then after theTable -> clear(); call

QCoreApplication::processEvents();
Q_ASSERT(std::all_of(theItemList.constBegin(),theItemList.constEnd(),[](const QPointer< QTreeWidgetItem >& val)->bool {return !val;}));

eos pengwern

@VRonin Thank you; I won't be able to pursue this until later this evening, but I'll report back when I've done so.

eos pengwern

@VRonin OK, I'm suitably embarrassed about this... I've found the problem, and it wasn't in the way that I've been reconstructing the QTreeWidget. Tucked in with the redrawing code (and removed from the example code I gave above for the sake of simplicity) was a reference to an innocent-looking QVector in which I stored the values of 'index' in an order which, as it turns out, was wrong. I've corrected that error, which was leading to the QVector being addressed out-of-bounds later on. It was that out-of-bounds address later on which was precipitating the crash; I still don't understand why the debugger couldn't just tell me that, rather than giving baroque and outlandish heap errors, but there we go. Thank you for your help along the way, which made me look at the code a bit more critically and so led me to notice the error.