[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: allowed uses of non-baseline CPU extensions [and 1 more messages]

Simon McVittie writes ("Re: allowed uses of non-baseline CPU extensions"):
> I think this is perhaps a reasonable compromise: consider non-baseline
> CPU requirements to be a bug,

I think this is the right approach.  It neatly solves the problem of
asking the maintainer to do the porting and testing work.

Adam Borowski writes ("allowed uses of non-baseline CPU extensions"):
> The above cases, at the first glance, seem to provide clear rules.  Too bad,
> in all but the first case, there's no clear boundary.  Especially "not
> supported by upstream": how much work can be expected from the maintainer?

Applying Simon's answer to your example cases:

> So, let's list packages that want non-baseline:
> * multiple variants: package src:x provides x-unoptimized, x-sse3 and
>   x-avx1048576.  Clearly legitimate and a good idea.

Assuming the dependency arrangements work as intended, there is no bug
here because the package functions correctly on baseline systems.

> * useless on older CPUs: pcsx2 can't emulate its game console on hardware
>   older than said console; scientific number-crunching software is pointless
>   on a Pentium2.  Trying to provide baseline builds would be a pure waste
>   of time of the packager, and archive space.

If a user writes some kind of massive emulation patch that implements
the missing functionality with a JIT, then we should consider that
patch on its merits.  (We might well say "err, can you package that as
a separate package and we'll use it conditionally".  And of course
we'd expect to say "you did talk to upstream about this right?" if the
answer wasn't obviously.)  This is likely to be hypothetical...

> * not supported by upstream: chromium:i386 on !sse2, rust:armhf on !neon
>   (now fixed).  Not a good thing but a maintainer often lacks resources
>   to implement this h{im,er}self.

Again, if a patch is available and not too disruptive we should apply
it.  How hypothetical this is, and indeed the quality of the patch,
will vary from case to case.  Bad patches should not be applied,

> Then there is:
> * generic variants exist but someone decided that using fancy extensions
>   is "better" (it might even indeed be better on new CPUs, but...)

If a user submits a patch to turn off the fancy extensions it should
be applied.

The important point is that "makes it run rather slower, on the much
more common newer hardware" is not a good reason to reject the patch.


Reply to: