Re: git: 8b197806dcdf - main - libsysdecode: Teach mktables to handle enums

From: Jessica Clarke <jrtc27_at_freebsd.org>
Date: Fri, 20 Nov 2026 02:40:54 UTC
On 19 Nov 2026, at 23:34, Gleb Smirnoff <glebius@freebsd.org> wrote:
> 
> On Thu, Nov 19, 2026 at 11:24:11PM +0000, Jessica Clarke wrote:
> J> Hi Gleb,
> J> Thanks for the reply as someone involved with this set of changes and
> J> able to comment on their importance; I’m not aware of the bigger
> J> picture here.
> J> If there are important changes that depend on it and people are happy
> J> to go with this kind of approach, I can look at doing some tinderbox
> J> builds to verify builds on FreeBSD at least aren’t modified (and that
> J> Linux gets the same result, since I have a big Linux box available).
> J> That should at least give re@ assurance about such a late-binding
> J> change and help me sleep at night; I generally favour reverting to a
> J> known state over fixing forwards, but probably we can avoid that here.
> J> 
> J> I’ll try and get some tinderbox builds kicked off later tonight.
> 
> Decoding is just a convenience feature for libsysdecode users, mostly
> the ktrace tool.  If mktables fails to build some table, the compilation
> still succeeds, just reduces user experience of later ktrace runs.  Numerics
> will be printed instead of enum literals.
> 
> So one option would be not parse enums on Apple.  This will make builds
> non-reproducible compared to native builds.
> 
> Other option is just to throw in your patch.  This is not a critical
> functionality, and I suppose re@ will not object for late change.

https://reviews.freebsd.org/D61783

I might as well do a macOS tinderbox overnight for good measure, but I
would be quite surprised if it differed with respect to its tables.h
files.

For those who might be wondering, yes, I did also check I had actually
applied my patch (by also making mktables emit a distinctive marker to
stderr that I later grep’ed for in _.*.buildworld).

Jessica