John pointed out that patch 2 of my original bug fix did too much in one patch. Now that we are in freeze for 5.1, I've proposed two variants of the same fix: patch 1 is the bare minimum to fix the bug and nothing else, while patches 3-7 are more in line with my v1 patch at doing other refactoring work along the way (but now split into multiple logical steps) that will prove useful for my other pending snapshot improvements (https://www.redhat.com/archives/libvir-list/2019-February/msg01350.html, but now 5.2 material).
I'm proposing both variants for comparison, although I already suspect the answer will be 'use patch 1 for 5.1, then rebase what remains of patches 3-7 into the other snapshot cleanups for 5.2'. variant2 is cleaner than my v1 patch 2/2 in that I no longer have to use a default: label to hack around gcc's enum sanity checking within a switch statement, but it required introducing a large mechanical rename of all use of snapshot state values. Eric Blake (7): qemu: Fix snapshot redefine vs. domain state bug Revert "qemu: Fix snapshot redefine vs. domain state bug" qemu: Refactor check for _LIVE vs. _REDEFINE qemu: Factor out qemuDomainSnapshotValidate() helper snapshot: Rework virDomainSnapshotState enum qemu: Use virDomainSnapshotState for switch statements qemu: Fix snapshot redefine vs. domain state bug src/conf/snapshot_conf.h | 21 ++++- src/conf/snapshot_conf.c | 28 +++---- src/qemu/qemu_driver.c | 160 +++++++++++++++++++++++---------------- src/test/test_driver.c | 20 ++--- src/vbox/vbox_common.c | 4 +- 5 files changed, 137 insertions(+), 96 deletions(-) -- 2.20.1 -- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list
