Draw canvases produced in batch mode

ROOT
21

I can help if needed. Just let me know.

22

Too bad … the fix does not work on windows … looking further

23

Ok, actually we did the test on Windows with ROOT5.34… and I was testing on Mac with ROOT6 . They were different.
After applying all the changes on ROOT5 it fixes also the problem on windows. So it seems my fix works… which ROOT version are you using ? I am planning to apply the change on ROOT6 only.

24

Please fix ROOT 5 as you already have it.

25

Fixing in ROOT 5 requires to incorporate changes done for Cling… can be done… needs even more check…
Anyway I won’t commit the change tonight… I want to check more…

26

I used root 5.34.30 for all the results reported in my previous posts.

27

ok so it has to go in ROOT5 too …

28

:mrgreen:

29

:wink:

30

Sorry guys, for me it is a hard time even to make my colleagues update Root to a bug fixing release… the transition to Root 6 would require years of negotiates :mrgreen:

31

same here :-({|=

32

No problem Nicola … I am doing it for ROOT5 to… it takes a bit longer…

33

Now fixed in ROOT 5.34 and master (ROOT 6)

34

Thanks Olivier, I will test it as soon as possible (I’m out of office for a meeting).

35

With the patch in commit 859439690cb5ced590ca72a94da24ba4c505dcde, canvases created after putting back gROOT in interactive mode are drawn, so it works. Canvases created in batch mode are still not drawn because they are still in batch mode after gROOT has been set back to interactive. Setting the canvas in interactive mode and then drawing it produces a segfault in my system as reported before.
However, for what I need the patch is sufficient, so I consider this issue as fixed .Thanks again Olivier.

36

Hi Nicola,

Can you provide a small script I can easily run, reproducing the seg fault ?

Thanks,
Olivier

37

Actually there are two similar ways of triggering two different segfaults. The first:

And the second:

These errors are obtained with commit 859439690cb5ced590ca72a94da24ba4c505dcde (i.e. Olivier’s patch) on a 64 bit ArchLinux system with gcc 5.2.0.

38

Thanks Nicola. Looking at it.

39

I am sorry but I will be oblige to undo the “fix” for 5.34 … it has some side effect we saw today … That’s is not surprising because 5.34 and 6 are quite different now in that part. 5.34 need stability we cannot afford to make big changes like that… sorry … I keep the change in the master though…

40

OK, thanks for the heads-up.