On Tue, 2020-07-28 at 10:00 +0200, Pavel Hrdina wrote:
So based on all the comments we have these options for pushing this series:
1) Squash it into single commit.
2) Keep the patches as they are and running meson build & ninja will not fail.
3) Keep the patches as they are but error out in meson until the conversion is complete. The error can be used to detect if git bisect is withing the meson rewrite.
4) Rework the series to have patches adding meson bits without removing anything from autotools and drop the autotools files in a single commit once the meson rewrite is complete
I suggest going for option 1, with the caveat that before pushing you you should post a fully-reviewed and fixed vN to the mailing list and include a link to it in the message for the squashed commit. This will allow us to look back at the original reasoning behind a change in the same way the reviewer could, without affecting bisectability. Options 4 would be acceptable as well, but it requires more work on your side and still requires hitting the mailing list archives to get the full picture of the rewrite, so I don't think is worth it. Options 2 and 3 result in a partially-built libvirt over a pretty big range of commits, which I feel is actually worse than breaking bisectability. -- Andrea Bolognani / Red Hat / Virtualization