On Wed, Aug 17, 2022 at 10:23:23AM -0700, Nathan Chancellor wrote:
On Wed, Aug 17, 2022 at 05:21:23PM +0100, Mark Brown wrote:
It's probably worth talking to the 0day people about prioritising what they're reporting against, especially given that the reports have evolved into a bit of an eye chart - this was reported against a Hexagon randconfig with an unreleased compiler which is underselling it rather.
At the same time, I would expect developers and maintainers to focus on the warning text first and foremost, not what architecture, configuration, or compiler is being used. This issue is very clearly not architecture or configuration specific, there is no #ifdef in this
That's the eye chart bit of it - part of the problem with 0day specificially is that a lot of their reports have become quite hard to read, they've been putting in something that looks a lot like git annotate output which makes things very wide which causes wrapping issues (this one is actually a bit better than most now I go look at it again since it doesn't have that indentation). Picking obscure configurations makes it more likely that people won't get round to figuring out what the issue being reported is since it seems less urgent and therefore gets pushed further to the back of the queue. For something that's cropping up on a wide range of configurations it'd be good to priorirtize the more prominent ones to mitigate against this.
function that changes the nature of the warning. While it is compiler specific (because possible uninitialized variable warnings are disabled with GCC), it is not dependent on the version (although I guess that isn't apparent). I suppose I can just comment on future randconfig reports to point out how they will affect allmodconfig and such.
That'd probably help.