TEventList problems

ROOT
1

Here are two pieces of code. I would expect that both of them would give the same result, but that is not true.

...
vector< vector< TEventList > > listTree;
...
map< double, vector< TEventList > >::iterator it = selectionLists[j].find( curCutValue[j] );
...
for (unsigned k = 0; k < samples.size(); ++k) {
	listTree[j][k] = listTree[j + 1][k];
	listTree[j][k].Intersect( &(it->second[k]) );
}
for (unsigned k = 0; k < samples.size(); ++k) {
	listTree[j][k] = listTree[j + 1][k];
	TEventList * nn = new TEventList( listTree[j][k] );
	nn->Intersect( &(it->second[k]) );
	listTree[j][k] = *nn;
	delete nn;
}

First piece gives following error, but not first time the loop is executed, but somewhere later

*** glibc detected *** ./HZZ2l2nu: double free or corruption (!prev): 0x0a53ece8 ***
======= Backtrace: =========
/lib/tls/i686/cmov/libc.so.6(+0x6b591)[0x5293591]
/lib/tls/i686/cmov/libc.so.6(+0x6cde8)[0x5294de8]
/lib/tls/i686/cmov/libc.so.6(cfree+0x6d)[0x5297ecd]
/usr/lib/libstdc++.so.6(_ZdlPv+0x21)[0x1596741]
/usr/lib/libstdc++.so.6(_ZdaPv+0x1d)[0x159679d]
/home/zablocki/CMS/root/lib/libTree.so(_ZN10TEventList9IntersectEPKS_+0xf2)[0x557a732]
./HZZ2l2nu(_Z11scanAllCutsSt6vectorI6sampleSaIS0_EERKS_IPK3CutSaIS5_EERKS_IPS3_SaISA_EERKS_IP4optCSaISG_EE+0x120a)[0x806161b]
./HZZ2l2nu(main+0x5562)[0x80554af]
/lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6)[0x523ebd6]
./HZZ2l2nu[0x804fde1]
======= Memory map: ========
...

but second one works without any problems. I really don’t understand what’s going on. Can anyone explain that? Is this bug or I’m missing something?

2

Hi,

CINT has problems tracking the temporaries of the line

Please either compile your code or spell the arguments out by creating variables that store intermediary results (it->second, [k], &).

Cheers, Axel.

3

Hello,

the problem is that this code was compiled (with g++ as executable).

I don’t fully understand what you mean by saying [quote]spell the arguments out by creating variables that store intermediary results (it->second, [k], &)[/quote]
Isn’t that what I’m doing in the piece of code that is working?

4

Hi,

I take back all I said :slight_smile:

Can you provide running code that shows the problem? I.e. please fill in the “…” in between the code fragments that you posted, such that I can call the code and see the error.

Cheers, Axel.

5

The code is pretty long and this just one function of many. Here is gzipped version. To run the code though you would need also specific root-tuples. The problem is in scanAllCuts() function in CreatePlot.cpp.

6

code.tgz (15.2 KB)

7

There might be issue with the TEventList operator=. What does valgrind says if you use it to run your failling example?

Cheers,
Philippe.

8

For this code valgrind gives like 40k errors, where probably all of them come from root libraries. I reported this in other post. This makes almost impossible to find something useful with valgrind.

9

Hi,

Did you try the ROOT suppression file $ROOTSYS/etc/valgrind-root.supp ?

Philippe.

10

Yes, the number of errors I quoted was after using suppression file.

11

[quote]For this code valgrind gives like 40k errors, where probably all of them come from root libraries
Yes, the number of errors I quoted was after using suppression file.[/quote]humm … odd. Can you send us the valgrind log file?

Cheers,
Philippe.