[Piuparts-devel] Bug#873733: piuparts: please ${COLOR}-list isa-support

[Adam Borowski]

Package: piuparts
Version: 0.78
Severity: normal


Hi!
A package of mine, sse4.2-support (src:isa-support) is blocked from
migrating to testing because of a piuparts failure.  Except, this failure
is not only ok but proves the package does its job (on a machine with no
SSE4.2).

It might be even good to check that it _does_ fail (on the machine in
question sse3-support should install but sse4.2-support refuse to),
but as your test hardware comes and goes, it's probably too much hassle
for a single snowflake.

Holger Levsen said:
> I'm not convinced it's ok to fail in this case. One might want to prepare a
> filesystem on a machine not captable for a captable machine…

That's why there's an override: you can set ISA_IGNORE=anything.
On the other hand, 99.9% of installations are on the same or identical
machine, and in most of that 0.1% it's a more capable machine rescuing the
worse one.

> > Thus, what should I do?
> > * do you have a ${COLOR}list I can request to be placed on?
> 
> we have, see piuparts.git/custom-scripts/scripts/*_exceptions
> 
> > * if not, may I detect piuparts testing and ignore ISA inadequacy?
> 
> I dont think you should do that.

Note that a package that depends on *-support will obviously also have this
problem.  On the other hand, all currently known packages with a strict ISA
extension dependency want SSE2 or SSE3 which require some serious dumpster
diving if you wanted to obtain a machine which lack that (alternatively, you
can use qemu without kvm and request an ancient CPU).  As I don't suspect
you of raiding dumpsters for hardware to test, it's probably
counterproductive to add an exception for sse2-support and sse3-support.


On the other hand, it'd be nice if you could add sse4.2-support to the list
of exceptions; it is currently blocked from migrating.


Meow!