Change in behavior of CloneTree w/ SetMakeClass(1) in ROOT 6.10

[kpedro88]

I have noticed a change in behavior of some of my ROOT code when moving from version 6.06.00 to 6.10.08.

Historically, I have had no problem processing a tree via a MakeClass and using CloneTree to make a copy of the tree (possibly with branches added, removed, or modified; or some events skipped). However, in 6.10 this does not work for vector branches: they are empty in the copied tree.

A minimal example demonstrating the problem is attached:
testclone.C (1.0 KB)

It can be run simply:

root -b -l -q testclone.C+

And the expected output is (e.g. in 6.06.00):

Processing testclone.C+...
30
30

But the output in 6.10.08 is:

Processing testclone.C+...
30
0

Perhaps this is expected for some reason, but if so I would like to understand why.

---

ROOT Version: 6.10.08
Platform: SL6
Compiler: gcc630

---

[Wile_E_Coyote]

CloneTree in TSelector (or CopyTree)

Hi Julien, The skeleton created by MakeSelector puts the TTree in mode where the reading of the TTree content is decomposed in ints and floats (‘MakeClass’ mode). This mode is currently incompatible with cloning the TTree (mismatch in what the writing and reading code expect). The result of MakeProxy on the other hand is compatible with CloneTree and should be such that the change in your code should be minimal. Cheers, Philippe.

[kpedro88]

And yet it works fine in 6.06, which was released after that message. So clearly, something changed after that release. The question is what changed, and if that change is a bug or a feature.

[kpedro88]

I would still like to know what changed in ROOT 6.10 to break this functionality (which worked in ROOT 6.06).