LVM manual partitioning segfault

[abucodonosor]

Describe the bug

LVM manual partitioning segfault.
Testing done in a VirtalBox 64bit VM.

( Kpmcore 3.3.0 , Calamares 3.2.{4,5,6} )

To Reproduce
Steps to reproduce the behavior:

1. Boot a VM with 2 HDDs
2. Prepare HDDs like this :
   .. Open an terminal

 for each HDD do:
 fdisk /dev/sdX
 type : n , p, enter , enter , enter , t ,8e , w

 pvcreate /dev/sda1 /dev/sdb1
 vgcreate calatest /dev/sda1 /dev/sdb1

3. Start calamares
4. On Partitioning site:
   Click Manual partitioning
   Click Next
   Click Deactivate Volume Group
   Click Back
   Click Erase ( to re-trigger partitions update )
   Click Manual partitioning and Next
   < segfault >

It seems to die from PartitionCoreModule::revertDevice().
At the moment I cannot provide good crash logs but I will if needed.

[adriaandegroot]

I did this on a machine with 3 disks; sda is the installed OS & Calamares dev disk, sdb and sdc are scratch disks. So I created the VG on the scratch disks, 5GB in size.

The crash can be reproduced, both straight from the command-line and when running Calamares in gdb. gdb and valgrind show that it's (simply?) a nullptr being passed to revertDevice.

[adriaandegroot]

Chasing this some more with debuggers and logging .. gets me back in the swamp of models being deleted unpredictably and scoped pointers being copied and the whole memory-management mess that is the partitioning module. At this point I think the right thing is to replace the whole thing.