Reading from file TGraphErrors that have been fitted gives seg fault when compiling twice

[marini]

Dear users,

I just updated ROOT from 6.04 to 6.14 and ubuntu from 15 (LTS) to 18.04.
TGraphErrors were fitted with a function from an other code and then saved in a file. The fitting function is saved together with the graph (by default if I well understand).
I try to read these graphs from the file and plot them. The code worked on the previous ROOT (and gcc) version. Now I can compile and run the code, but I get this error:

Input_line_71:2:10: error: redefinition of 'TFormula____id1627201304017190959'
Double_t TFormula____id1627201304017190959(Double_t *x,Double_t *p){ return (p[0]*1.+p[1]*x[0]+p[2]*0.5*(3.*TMath::Power(x[0],2)-1.)) ; }
         ^
input_line_69:2:10: note: previous definition is here
Double_t TFormula____id1627201304017190959(Double_t *x,Double_t *p){ return (p[0]*1.+p[1]*x[0]+p[2]*0.5*(3.*TMath::Power(x[0],2)-1.)) ; }

That TFormula is indeed the TF1 function that I used for the fit. Still the code gets to the end and plots correctly. But if I compile again I get Segmentation Fault. That was not the case with the previous ROOT version.

I’m attaching an example of the code.
What am I doing wrong?

Thank you very much for your help
Cheers
Paola

---

_ROOT Version:ROOT 6.14/04
_Platform: Ubuntu 18.04
_Compiler: gcc 7.3.0
___test.C (1.7 KB)
test.root (12.3 KB)

[Axel]

Hi Paola,

Looks like the good old TFormula persistency error - @moneta could you have a look, please?

Cheers, Axel.

[marini]

Thank you Axel!
Cheers
Paola

[moneta]

Hi,

This problem should be fixed. I cannot reproduced it with the latest master and ROOT 6.14

Lorenzo

[marini]

Hi Lorenzo,

ehm, what is the “latest master”?
Do you mean that when you run my code you don’t have any error?
Cheers
Paola

[moneta]

Hi,

It is the latest version of the code available in GitHub. You can also find it a version built every night at CERN in cvmfs, if you want to try.

Yes, I am not getting any error when running your code

Lorenzo

[marini]

Hi Lorenzo,

I got the ROOT source from here:
https://root.cern.ch/content/release-61404
last monday. Should I download and install ROOT again? Is there a difference with the latest version available in GitHub (which, I admit it, I have no clue on what it is )
Thank you very much for your help
Cheers, Paola

[moneta]

Hi,

The problem was reported and fixed a bit more than one month ago.
See https://sft.its.cern.ch/jira/browse/ROOT-9467

I think the build version of 6.14 that you can download has still this problem. You should build from source as explained here https://root.cern.ch/building-root or, if you can, use the version built every night at CERN cvmfs

Lorenzo

[marini]

Hi,
that’s actually what I did: build it from source (downloaded last monday).
I’m a bit lost then…
Thank you
Paola

[moneta]

Hi

You have it probably download from a

Tar file. You should get the head of the e master or 6.14 patches from GitHub.

See Direct git repository access at

https://root.cern.ch/get-root-sources

Lorenzo

[marini]

Hi,
yes, I got the .tar root_v6.14.04.source.tar.gz
Ok, I’ll try from the git repository and let you know.
Thank you very much for your help
Cheers
Paola

[marini]

Hi,

I recompiled the root source from the GIT repository, download as:

git clone http://github.com/root-project/root.git

I didn’t get any error during compilation, and

here is the screen when I start root:

---

| Welcome to ROOT 6.15/01 https://root.cern |
| © 1995-2018, The ROOT Team |
| Built for linuxx8664gcc on Nov 02 2018, 12:03:00 |
| From heads/master@v6-13-04-1996-gcd6290ba1c |
| Try ‘.help’, ‘.demo’, ‘.license’, ‘.credits’, ‘.quit’/’.q’ |

---

so I guess I really got the master from the GIT.
Also the ROOT version was 6.14 and now it is 6.15.

I can compile and run the test.C code, and don’t have anymore the warning

Input_line_71:2:10: error: redefinition of 'TFormula____id1627201304017190959' Double_t TFormula____id1627201304017190959(Double_t *x,Double_t *p){ return (p[0]*1.+p[1]*x[0]+p[2]*0.5*(3.*TMath::Power(x[0],2)-1.)) ; }

but when I try to compile the second time I still get a segmentation fault.
I attach the stack trace in the uploaded file.
Do you have an idea of the problem?

Thank you very much for your help

Cheers
Paola

stack.txt (7.4 KB)

[moneta]

Hi

Good that with the master (6.15) it works now.
For your second problem, I cannot reproduce it. But, first of all in ROOT 6 you really don’t need to compile the macros. Using AClic it is not in general either faster or better, since Cling is a real compiler.
Second we don’t have full support for reloading macros a second time, after being executed. This because the unloading of the symbols does not work 100%. It is better exiting ROOT and re-execute the macro

Cheers

Lorenzo

[pamputt]

Does it mean that Paola should do

.x test.C

instead of

.L test.C

Second we don’t have full support for reloading macros a second time, after being executed. This because the unloading of the symbols does not work 100%. It is better exiting ROOT and re-execute the macro

Is there any JIRA ticket to track this issue?

[moneta]

Hi,

What I have said that it is not recommended to re-load the macro, i.e. doing

.L test.C
test()
.L test.C
test()

Lorenzo

[Axel]

Plenty: https://sft.its.cern.ch/jira/browse/ROOT-7939?jql=project%20%3D%20ROOT%20AND%20resolution%20%3D%20Unresolved%20AND%20text%20~%20"unloading"%20ORDER%20BY%20priority%20DESC%2C%20updated%20DESC

It’s a known major annoyance.

Axel.

[Axel]

…but instead:

.L text.C
test()
test()

[marini]

Hello,
thank you to everybody!
Sure I can run twice without recompiling. The reason for recompiling is that I change the code meanwhile And I do need to compile because the macro that I sent you is just a small part of a much bigger code.
Lorenzo, which version of gcc do you use? Because you say you cannot reproduce the seg fault, so as we use the same version, a difference could be the compiler.
Thank you very much

Cheers
Paola