Calling TH1F function in a loop, segmentation violoation

fclopez · September 26, 2012, 7:10pm

Hi,
My question may not be Root related but I am not sure. I am faced with segmentation violation problems.
My goals are the following:

load a 2d histogram from a set of files
from the 2d histogram, extract 1d histograms in a for-loop
i use each 1d histogram to fit and retrieve a fit parameter (inside the for loop) then store in an array.
i start to fail in 2) and 3).

TH1F *readDataFile(char* fname, int nch) {
  int ch,counts;
  float fc;
  TH1F *h = new TH1F("h1","",nch,-0.5,nch-0.5);
  h->Sumw2();
  FILE* fp=fopen(fname,"r");
  if (fp==NULL) {
    printf("Could not open file %s\n",fname);
    return h;
  }
  for (int j=0; j<nch; j++) {  
      fscanf(fp,"%d %f",&ch,&fc);
      counts=fc;    
      h->SetBinContent(j+1,fc);
  }
  fclose(fp); 
  return h;
}

TH2F* createScan(char* fn, int mi, int ma, int step, int nch) {    
  char fname[1000];
  TH1F *hdum;
  TH2F *h2 = new TH2F("h2","",nch,-0.5,nch-0.5,((ma-mi)/step)+1,mi-(Double_t)step/2.,ma+(Double_t)step/2.);
  h2->SetStats(kFALSE);  
  for (int i=mi; i<ma+1; i=i+step) {
    sprintf(fname,fn,i); 
    hdum=readDataFile(fname, nch);
    for (int j=0; j<nch; j++)
      h2->Fill(j,i,hdum->GetBinContent(j+1));
    delete hdum;   
  }
  return h2; 
}

TH1F * getCh(TH2F* h2, int chan) {
  int nstep=h2->GetYaxis()->GetNbins();
  Double_t mi=h2->GetYaxis()->GetXmin();
  Double_t ma=h2->GetYaxis()->GetXmax();  
  char channel[]= "Channel";
  char chanNum[100]; 
  sprintf(chanNum, "%s %d", channel, chan);  
  TH1F *h1=new TH1F("h1",chanNum,nstep,mi,ma);
  for (int i=1; i<nstep+1; i++) { 
    h1->SetBinContent(i, h2->GetBinContent(chan+1,i));    
  }
  return h1;
}


Double_t FitERF4(TH1F *h1, float min,Double_t *mypar, Double_t *emypar, int plot) {
	float chi=100.;
	float flex;	
	float maxx = h1->GetMean()-15;
	mypar[3]=h1->GetBinContent(h1->GetXaxis()->FindBin(maxx-15));
		
	TF1* fitfunc;
	TF1 *func = new TF1("func",erfFunction3,min,maxx,5);
	
	func->SetParameters(mypar);
	func->SetParameter(0,0.);
	func->SetParLimits(0,0., 10.);	
	func->SetParameter(1,390.);
	func->SetParLimits(1,360, 430);	
	func->SetParameter(2,10.);
	func->SetParLimits(2,0.,100.);		
	func->SetParameter(4,0.1);
	func->SetParLimits(4,0., 1.);		
	func->SetRange(min,maxx);
	  
  	if (plot) {
  		h1->Fit("func","R"); 
  	}else{
  		h1->Fit("func","R0QV");
  	}  
  	fitfunc = h1->GetFunction("func");
	fitfunc->GetParameters(mypar);                               
  	emypar[1]=fitfunc->GetParError(1);                           
  	emypar[0]=fitfunc->GetParError(0);
  	emypar[2]=fitfunc->GetParError(2);
  	emypar[3]=fitfunc->GetParError(3);
  	emypar[4]=fitfunc->GetParError(4);
  	chi = fitfunc->GetChisquare()/(fitfunc->GetNDF());
  	flex = fitfunc->GetParameter(1);   	
  	delete func;
	if (plot)
           printf("Chi square =%e\n",chi);   
	return flex;
}

float *flexpoint(TH2F *h2) {
	float flex;
	float *flexArray =0;
	int nmod = 3;
	
	int nstep=h2->GetYaxis()->GetNbins();
	Double_t mi=h2->GetYaxis()->GetXmin();
	Double_t ma=h2->GetYaxis()->GetXmax(); 
	float min = 330;  
	
	Double_t mypar[5];
	Double_t emypar[5];
	
	for(int ichan=0; ichan<NCHAN*NCHIP*nmod; ichan++) {
		TH1F *hchan = new TH1F("hchan", "", nstep, mi, ma);		
		hchan = getCh(h2,ichan); //I start  failing here
		flex = FitERF4(hchan,min, mypar, emypar,1); // also here
		flexArray[ichan] = flex;
		delete hchan;		
	}
	return flexArray;
}

The culprit are the following:

Root > Function getCh() busy flag cleared
Function flexpoint() busy flag cleared

Details:


 *** Break *** segmentation violation
Attaching to program: /proc/7589/exe, process 7589
warning: .dynamic section for "/lib/libdl.so.2" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/usr/lib/libstdc++.so.6" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/lib/libm.so.6" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/lib/libgcc_s.so.1" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/lib/libpthread.so.0" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/usr/lib/libXrender.so.1" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/usr/lib/libXinerama.so.1" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/usr/lib/libXext.so.6" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/usr/lib/libXau.so.6" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/usr/lib/libXdmcp.so.6" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/usr/lib/libXcursor.so.1" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
warning: .dynamic section for "/usr/lib/libXfixes.so.3" is not at the expected address
warning: difference appears to be caused by prelink, adjusting expectations
[Thread debugging using libthread_db enabled]
0x00bdf402 in __kernel_vsyscall ()
#1  0x069dccb3 in __waitpid_nocancel () from /lib/libc.so.6
#2  0x0698162b in do_system () from /lib/libc.so.6
#3  0x00386f3d in system () from /lib/libpthread.so.0
#4  0x006d0477 in TUnixSystem::Exec(char const*) () from /users/root_v5.20/lib/libCore.so
#5  0x006d5f01 in TUnixSystem::StackTrace() () from /users/root_v5.20/lib/libCore.so
#6  0x006d2b46 in TUnixSystem::DispatchSignals(ESignals) () from /users/root_v5.20/lib/libCore.so
#7  0x006d2bd4 in SigHandler(ESignals) () from /users/root_v5.20/lib/libCore.so
#8  0x006d1e19 in sighandler(int) () from /users/root_v5.20/lib/libCore.so
#9  <signal handler called>
#10 0x027f35eb in G__G__Hist_107_0_79(G__value*, char const*, G__param*, int) () from /users/root_v5.20/lib/libHist.so
#11 0x00c1c2e3 in Cint::G__ExceptionWrapper(int (*)(G__value*, char const*, G__param*, int), G__value*, char*, G__param*, int) () from /users/root_v5.20/lib/libCint.so
#12 0x00c2e7a2 in G__exec_asm () from /users/root_v5.20/lib/libCint.so
#13 0x00c31f26 in G__exec_bytecode () from /users/root_v5.20/lib/libCint.so
#14 0x00c92228 in G__interpret_func () from /users/root_v5.20/lib/libCint.so
#15 0x00c80966 in G__getfunction () from /users/root_v5.20/lib/libCint.so
#16 0x00c64e70 in G__getitem () from /users/root_v5.20/lib/libCint.so
#17 0x00c67abb in G__getexpr () from /users/root_v5.20/lib/libCint.so
#18 0x00c6d55d in G__getexpr () from /users/root_v5.20/lib/libCint.so
#19 0x00cdab4f in G__exec_statement () from /users/root_v5.20/lib/libCint.so
#20 0x00ce612a in G__exec_loop () from /users/root_v5.20/lib/libCint.so
#21 0x00ce5a3b in G__exec_statement () from /users/root_v5.20/lib/libCint.so
#22 0x00c93a8b in G__interpret_func () from /users/root_v5.20/lib/libCint.so
#23 0x00c80966 in G__getfunction () from /users/root_v5.20/lib/libCint.so
#24 0x00c64e70 in G__getitem () from /users/root_v5.20/lib/libCint.so
#25 0x00c67abb in G__getexpr () from /users/root_v5.20/lib/libCint.so
#26 0x00c572e4 in G__define_var () from /users/root_v5.20/lib/libCint.so
#27 0x00ce2a67 in G__exec_statement () from /users/root_v5.20/lib/libCint.so
#28 0x00c52c68 in G__exec_tempfile_core () from /users/root_v5.20/lib/libCint.so
#29 0x00c53fa3 in G__exec_tempfile_fp () from /users/root_v5.20/lib/libCint.so
#30 0x00cebb2c in G__process_cmd () from /users/root_v5.20/lib/libCint.so
#31 0x006bf8db in TCint::ProcessLine(char const*, TInterpreter::EErrorCode*) () from /users/root_v5.20/lib/libCore.so
#32 0x00604a16 in TApplication::ProcessLine(char const*, bool, int*) () from /users/root_v5.20/lib/libCore.so
#33 0x00238e9a in TRint::HandleTermInput() () from /users/root_v5.20/lib/libRint.so
#34 0x002374f0 in TTermInputHandler::Notify() () from /users/root_v5.20/lib/libRint.so
#35 0x0023978e in TTermInputHandler::ReadNotify() () from /users/root_v5.20/lib/libRint.so
#36 0x006cee02 in TUnixSystem::CheckDescriptors() () from /users/root_v5.20/lib/libCore.so
#37 0x006d3024 in TUnixSystem::DispatchOneEvent(bool) () from /users/root_v5.20/lib/libCore.so
#38 0x0065c8bc in TSystem::InnerLoop() () from /users/root_v5.20/lib/libCore.so
#39 0x0065c683 in TSystem::Run() () from /users/root_v5.20/lib/libCore.so
#40 0x00604b02 in TApplication::Run(bool) () from /users/root_v5.20/lib/libCore.so
#41 0x00237c6e in TRint::Run(bool) () from /users/root_v5.20/lib/libRint.so
#42 0x08048d83 in main ()
A debugging session is active.

        Inferior 1 [process 7589] will be detached.

Quit anyway? (y or n) [answered Y; input not from terminal]
Detaching from program: /proc/7589/exe, process 7589
Root > Function getCh() busy flag cleared
Function flexpoint() busy flag cleared

Compiling with filename.C++, no complains to my buggy program was noted. Any suggestions how to fix this?

F.

pcanal · October 3, 2012, 3:19pm

Hi,

Does the crash exist also when you compile (i.e. I suspect a CINT bug)?

Philippe.

fclopez · December 7, 2012, 1:15pm

No, it does not crash

Wile_E_Coyote · December 7, 2012, 4:42pm

This looks really weird … float *flexpoint(TH2F *h2) { // ... float *flexArray =0; // ... for(int ichan=0; ichan<NCHAN*NCHIP*nmod; ichan++) { // ... flexArray[ichan] = flex; // ... } return flexArray; } I would say you need something like: float *flexArray = new float[(NCHAN*NCHIP*nmod)]; // ... and afterwards, when you no longer need it ... delete[] some_flexArray_pointer_returned_once_by_flexpoint;
Also, in the “FitERF4” function you “delete func” … I’m not sure if you are allowed to do this, as the “h1” histogram uses it (as its “fitfunc”).

fclopez · December 10, 2012, 2:46pm

Thanks, but in the loop

for(int ichan=0; ichan<NCHAN*NCHIP*nmod; ichan++) {
      TH1F *hchan = new TH1F("hchan", "", nstep, mi, ma);      
      hchan = getCh(h2,ichan); //I start  failing here    
      ......//forget about the other lines
      delete hchan;      
   }

while if I do

	M1:
	if(ichan>=chMin && ichan<chMax) { 
                        TH1F *hchan = new TH1F("hchan", "", nstep, mi, ma); 
			hchan = getCh(h2, ichan); //this seems to work
			c1->Modified();				
			c1->Update();
			usleep(70000);
			delete hchan;			
			
	}
        ichan++;
	if (ichan  <= chMa+1)
			goto M1;

which looks funny

Wile_E_Coyote · December 10, 2012, 2:59pm

Well, this looks weird … TH1F *hchan = new TH1F("hchan", "", nstep, mi, ma); // you create a new "hchan" histogram hchan = getCh(h2,ichan); // you OVERWRITE THE POINTER with a pointer to "h1" (returned by "getCh") // the pointer to the "hchan" histogram, created in the first line above, is "lost" (a memory leak) delete hchan; // you delete "h1" (that was created in "getCh") Try with: TH1F *hchan = getCh(h2,ichan); // the hchan pointer will point to the "h1" histogram (created in "getCh") // ... delete hchan; // delete "h1" (that was created in "getCh")

fclopez · December 11, 2012, 9:29am

It is a good point, Wile. Thanks. But it does not address my main problem

Root > Function getCh() busy flag cleared
Function flexpoint() busy flag cleared

see above…

Wile_E_Coyote · December 11, 2012, 10:22am

How about you pack your source code and your data so that I can try it myself?

ksmith · December 11, 2012, 5:17pm

What are the values of NCHAN and NCHIP? Is ichan exceeding h2->GetNbinsX()?

ksmith · December 11, 2012, 5:26pm

A simple if statement in getCh() whould indicate where the issue is and avoid the seg fault.

for (int i=1; i<nstep+1; i++) { 
   if (chan+1 <= h2->GetNbinsX() && i <= h2->GetNbinsY())
      h1->SetBinContent(i, h2->GetBinContent(chan+1,i));    
   else
      fprintf(stderr,"ERROR: Unknown bin %d %d",chan+1,i);
}

Wile_E_Coyote · December 11, 2012, 9:39pm

Both methods, TH1F::SetBinContent and TH2F::GetBinContent, are internally protected against “incorrect” bin numbers (there should be no case in which they cause any segmentation fault).

ksmith · December 11, 2012, 10:43pm

This is true. Although, one might waste time though creating histograms that don’t have a corresponding bin in the TH2.

I do see a different issue. You define a pointer flexArray and then use it like an array flexArray[ichan] = flex; This will most definitely lead to a segmentation fault.

Also, why don’t you use GetParErrors() in FitERF4(), instead of getting each value one at a time? fitfunc->GetParErrors(emypar);