From 7293f484b22cbcc73310db872bc1bca2b6d6a1d8 Mon Sep 17 00:00:00 2001 From: Lars Wirzenius Date: Tue, 5 May 2020 08:57:46 +0300 Subject: Change: update NEWS for release --- NEWS | 83 +++++++++++++++++++++++++++++++++++++++++++++++++++----------------- 1 file changed, 62 insertions(+), 21 deletions(-) diff --git a/NEWS b/NEWS index 28e5708..cb8a0de 100644 --- a/NEWS +++ b/NEWS @@ -4,31 +4,72 @@ NEWS for vmdb2, the Debian disk image builder Version 0.14.1+git, not yet released ------------------------------------ -* Document and rename the `virtual-filesystems` plugin, and undo the - previous changes to have `/proc` automatically mounted in the +### New and changed features + +* Document and rename the `virtual-filesystems` plugin. This allows + installation of Debian packages that require `/proc` to be mounted + to install, such as `ca-certificates-java`. + + Undo previous changes to have `/proc` automatically mounted in the chroot, when running commands in the chroot. That wasn't enough - (e.g., it didn't cover Ansible), and `virtual-filesystems` - seems generally like the better solution. + (e.g., it didn't cover Ansible), and `virtual-filesystems` seems + generally like the more general and better solution. - The grub plugin handles the filesystems it needs itself, but doesn't - mind if they're already mounted. The `virtual-filesystems` step is - only needed when a Debian package is installed, possibly via - Ansible, or some other command is run that needs `/proc` or one of - the other virtual filesystems. + The `grub` step handles the filesystems it needs itself, but doesn't + mind if they're already mounted. This change should NOT require changes to existing, working .vmdb - files. - -* The `mkpart` step now waits for the new partition device to exist - before it's finished. This only applies when the target is a real - disk, not a disk image. - -* Bug fix: The `grub` step documentation claimed the step key for - specifying the image device is `device`. It has now been fixed to be - `image-dev`. Thanks: Sebastian Bachmann. - - -Version 0.14.1, not yet released + files. If it does, please report a bug. + +* The .vmdb file can now be passed in via the standard input, by + specifying `-` as the filename. Patch from voney. + +* The manual now documents that un-partitioned images are not + supported, at least not yet. + +* The "All went fine" message is now written at the very end of the + vmdb2 run, if at all. Previously all the messages from teardown + parts of steps were printed after it, hiding it effectively. + +* The standard output and error output of the commands vmdb2 runs are + no longer logged in real time, only after the command has finished. + This is a side effect of the `cliapp.runcmd` helper function no + longer being used. The Python standard library subprocess module is + used directly instead. This change is a step to drop the dependency + on the cliapp library. + +### Bug fixes + +* The `grub` step documentation now uses the correct key for + specifying the image device (`image-dev`). Thanks: Sebastian + Bachmann. + +* The `mkpart` step now waits for the new device node to be created. + Previously it sometimes took enough time to be created that the next + step failed. This only affects when the output disk is a real block + device, not a disk image file. + +### Changes mainly affecting vmdb2 and plugin developers + +* The `./check` script only runs the yarn integration testing tool if + it's new enough to have the yarnutils Python library included. Yarn + is being replaced with Subplot. + +* Step runners must now use a new way of specifying expected and + allowed keys for steps. The new way specifies types for mandatory + keys, and default values for optional keys. The old + `get_required_keys` method is gone. This change makes mistakes in + step runner code a little less likely, and opens up the possibility + of generating documentation about step keys automatically. + +* Failures when un-mounting filesystems are now ignored. Such failures + can happen for various reasons, but by far the most common reason is + because the mount had already been unmounted and it's simpler to + just ignore failures than keep meticulous track of what's been + unmounted already. + + +Version 0.14.1 ------------------------------------ This is a bug fix release to fix bugs in the 0.14 release that slipped -- cgit v1.2.1