Re: CVS commit: src/sys/arch/sh3/include

[Valery Ushakov <uwe%stderr.spb.ru@localhost>]

On Mon, Sep 29, 2025 at 20:08:54 +0000, Roland Illig wrote:

> Module Name:	src
> Committed By:	rillig
> Date:		Mon Sep 29 20:08:54 UTC 2025
> 
> Modified Files:
> 	src/sys/arch/sh3/include: dmacreg.h
> 
> Log Message:
> sh3/dmacreg.h: fix snprintb definitions
> 
> In the previous definitions, the 'f' conversion was combined with the
> ':' conversion, which snprintb rejects at runtime, starting with
> snprintb.c 1.47 from 2024-04-07.

I really would have appreciated a courtesy heads up...

I completely forgot what was the issue with F/f.  The claim in the
commit message that changed snprintb was:

  These combinations would lead to garbled output.

The old format prints (libutil on 10) for SH3_DMAC_CMT_CMCSR_BITS e.g

  0x80<CMF,CKS=0(1/4)>

and the new one prints

  0x80<CMF,CKS=0=1/4>

What exactly about the first (old) one, that combines f and : is
garbled?  I detest the latter quite intensly ("0=1/4" is both bogus
and the whole thing looks ugly with those two equal signs).

I wrote those formats some 20 years ago when I was working on some
related drivers and I tuned them to look the way that made sense to
me.  I do not appreciate it when I'm suddenly being given an errand to
go and figure this out post-factum while the trail is hot instead of
being able to have at least some planning so that I can do it at a
convenient time.

-uwe