On 4/17/19 10:56 AM, Cole Robinson wrote:
So broadly I think the options are
- Flip the switch. Double error reporting until we remove now redundant calls. Worse error reporting in some cases like tristate and virstoragetype without special consideration. No or less issues with having half converted codebase. IMO Easier to patch out the redundant calls and easier to review the removals because we can do it per file rather than per enum usage which might be spread across multiple files.
- Do it incrementally: will force us to consider each case individually resulting in better overall error reporting. Until codebase is converted, possible dev confusion and risk of new code neglecting to raise an error. IMO the total dev and reviewer time is likely to be significantly higher
I definitely favor 'flip the switch' mostly because I think it will get this done the quickest, and once it's in git it distributes the load of working out the kinks to the whole dev team. Depending on uptake the incremental approach might never get finished, it's not clear. But beyond that I'm not tied to any specific naming or method so I'm open to ideas.
If consensus is to go for the incremental approach then I will support that
I can live with 'flip the switch'. I know my incremental backup patches will have to rebase to the new style, but that's true for either style (and more a question of whether we can detect unconverted code via compilation failure or syntax-check once the bulk of the code base is converted). -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org