On Tue, Aug 07, 2018 at 09:17:34AM -0400, Tom Rini wrote:
On Tue, Aug 07, 2018 at 01:37:34PM +0200, Wolfgang Denk wrote:
Dear Tom,
In message 20180807111435.GD29229@bill-the-cat you wrote:
Its a "warn" and this should be visible to the user IMHO. One might discuss, if an "info" or "notice" text should printed, but warnings seem quite important (at least for my taste). Changing this to even
...
Please bump it per board/SoC, thanks!
So we stop printing warnings now, just to satisfy code size limitations on a few boards?
This is a totally wrong approach! If code size is a problem, these boards should disable unneeded or at least non-essential features, but you don;t want to ignore warnings, especially if these cause a command to not perform the expected operation.
Yes, by default "warnings", which means pr_warn, along with pr_notice and pr_info get optimized away. In no case should a warning be printed in a case where we aren't otherwise recovering and working correctly. I think there's a strong case here that the root problem here is that the prints in question ought to be pr_err and that should be corrected in the kernel too.
That said, I am not strictly against increasing the default, slightly, to 5 (which means 4 and lower included). The troublesome boards already set things to 3 in their configs and we can make SPL_LOGLEVEL be untied from LOGLEVEL. But I want someone to show me the before/after size deltas on that too.