If it’s a use after delete, it may or may not result in visible errors depending on the exact position of objects in memory, which in turn depends on environment, compilation options, etc…
I am afraid that valgrind does not see memory access errors though:
~/R/build_v6-20-04 ~/Tools/valgrind-3.15.0/bin/valgrind --suppressions=etc/valgrind-root.supp root.exe
==25309== Memcheck, a memory error detector
==25309== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==25309== Using Valgrind-3.15.0 and LibVEX; rerun with -h for copyright info
==25309== Command: root.exe
==25309==
------------------------------------------------------------------
| Welcome to ROOT 6.20/04 https://root.cern |
| (c) 1995-2020, The ROOT Team; conception: R. Brun, F. Rademakers |
| Built for linuxx8664gcc on Jun 09 2020, 07:48:41 |
| From tags/v6-20-04@v6-20-04 |
| Try '.help', '.demo', '.license', '.credits', '.quit'/'.q' |
------------------------------------------------------------------
root [0] TCanvas a
(TCanvas &) Name: c1 Title: c1
root [1] TCanvas c
(TCanvas &) Name: c1_n2 Title: c1_n2
root [2] .q
==25309==
==25309== HEAP SUMMARY:
==25309== in use at exit: 26,214,113 bytes in 28,735 blocks
==25309== total heap usage: 107,041 allocs, 78,306 frees, 79,520,017 bytes allocated
==25309==
==25309== LEAK SUMMARY:
==25309== definitely lost: 16 bytes in 1 blocks
==25309== indirectly lost: 0 bytes in 0 blocks
==25309== possibly lost: 10,210,621 bytes in 7,679 blocks
==25309== still reachable: 582,618 bytes in 4,217 blocks
==25309== of which reachable via heuristic:
==25309== stdstring : 40 bytes in 1 blocks
==25309== multipleinheritance: 19,440 bytes in 11 blocks
==25309== suppressed: 15,420,858 bytes in 16,838 blocks
==25309== Rerun with --leak-check=full to see details of leaked memory
==25309==
==25309== For lists of detected and suppressed errors, rerun with: -s
==25309== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 213 from 6)
It’s also suspicious that valgrind segfaults and lldb dies on startup on your machine… at least in first approximation, it would point to some issue in your environment, but I’m afraid I can’t say what exactly.
Sorry I can’t be of more help! It seems impossible for us to reproduce the problem or even debug it from a distance with your help