[Toybox] FYI musl's support horizon.

[enh]

yeah, i think he (and i, when i have my bionic hat on) have an easier
problem than you ... supporting old kernels _usually_ (there are a few
exceptions) isn't too hard in the c library because it's generally
considered okay to just shrug and return -1 with errno == ENOSYS or
whatever.

whereas as an "app", your users expect you to just "do the thing". (which
at best tends to mean you have an untested/barely tested "success" case.
no-one worries that their return -1 has bitrotted, but the non-inotify path
or the non-O_TMPFILE or whatever ... that's a lot more likely!)

On Thu, Aug 26, 2021 at 1:49 PM Rob Landley <rob at landley.net> wrote:

> re: https://github.com/landley/toybox/pull/292#issuecomment-899895909
>
> I asked Rich Felker on IRC and he said he/musl plans to _never_ drop
> support for
> any old kernels, and 3.x or newer is plenty. (Didn't ask about compilers or
> build environments because you're shipping binaries...)
>
> Rob
>
> <landley> dalias: it means support horizon, ala
> https://landley.net/toybox/faq.html#support_horizon
> <dalias> i think i gathered it's about support for old kernels. musl has no
> moving target for that, or if it's moving it only moved further back
> <dalias> so yeah, there is no 'support horizon' for musl. kernel versions
> >=2.6.0 that work now will continue to work to the extent they do now
> <dalias> you need mid to late 2.6.x for full posix functionality; there's
> a slim
> possibility we may emulate more of that on older kernels in the future,
> but we
> won't drop any compatibility that's already there
> <dalias> making binaries to run on really old devices whose kernels can't
> be
> upgraded because nobody has the source or there's hacked up vendor
> spaghetti
> that's impossible to modernize is a major use case for musl
> <dalias> we actually have someone who just got musl arm-eabi binaries
> running on
> a pre-eabi kernel with a few minor hacks to the syscall entry point and
> similar
> ...
> <landley> So you're targeting "2.6 kernel or newer"?
> <landley> I dunno how to digest the above into a policy statement.
> <landley> Other than cut and paste the whole thing...
> <landley> Anyway, if you wanted to reply to the github thread directly,
> elliott
> would get the ping.
> <dalias> ok
> <dalias> it's documented as such. 2.6.0 is min version but subject to
> functionality limits of the kernel
> <dalias> >= 2.6.39 is recommended for full functionality (but actually .29
> or
> something is all you need
> _______________________________________________
> Toybox mailing list
> Toybox at lists.landley.net
> http://lists.landley.net/listinfo.cgi/toybox-landley.net
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.landley.net/pipermail/toybox-landley.net/attachments/20210826/8d8e382e/attachment.htm>