This release I have been seeing a lot of bouncing around in games spin sizes, even when there aren't changes that would seem to explain things. Other than one image being over the target size eariler in the process, this hasn't broken things, but it still seems odd. For example for RC6 the i686 image was roughly 130 MB bigger than the x86_64 image, but for RC9 things flip flopped and the x86_64 image was roughly 130 MB bigger than the i686 image. I wouldn't have expected any significant size change for the games spin between RC6 and RC9. So I think something weird is going on. I don't have a lot of time right now to figure out the root cause of this, but I'll keep a lookout for it in F24.
On Thu, 2015-10-29 at 14:23 -0500, Bruno Wolff III wrote:
This release I have been seeing a lot of bouncing around in games spin sizes, even when there aren't changes that would seem to explain things. Other than one image being over the target size eariler in the process, this hasn't broken things, but it still seems odd. For example for RC6 the i686 image was roughly 130 MB bigger than the x86_64 image, but for RC9 things flip flopped and the x86_64 image was roughly 130 MB bigger than the i686 image. I wouldn't have expected any significant size change for the games spin between RC6 and RC9. So I think something weird is going on. I don't have a lot of time right now to figure out the root cause of this, but I'll keep a lookout for it in F24.
We've seen this on a smaller scale when looking at blocker images very close to the size limit, they seem to flop around a bit. It's probably something to do with the nitty gritty details of squashfs'es and ISO file construction...