[JUJU-3600] Remove 32-bit architecture support #15534
Conversation
The following removes 32-bit architecture support. Although we've dropped support for 32-bit architectures, we didn't actively remove the code to drop support. As we no longer want to support 32-bit word sizes with dqlite, as the number of jujud permutations explodes and so it's best to completely remove the support. The brings in the the juju/utils arch support, with the addition of a UnsupportedArches field, which will ensure that we can at least say what we use to support, but no longer do. Additionally, I've dropped support for ppc and we explicitly callout ppc64(le|el).
SimonRichardson
force-pushed
the
remove-32-bit-arches
branch
from
0196dff
to
45101b9
Compare
SimonRichardson
force-pushed
the
remove-32-bit-arches
branch
from
d5ebb3d
to
2c9d402
Compare
| {"ppc64le", "ppc64el"}, | ||
| {"s390x", "s390x"}, | ||
| {"riscv64", "riscv64"}, | ||
| {"risc", "riscv64"}, |
There was a problem hiding this comment.
I'm unsure about this one, as we do move it to riscv64, I'm not a fan of dropping the word size, I think we should be a lot stricter.
core/arch/arches_test.go
Outdated
| raw string | ||
| arch string | ||
| }{ | ||
| {"windows", "windows"}, |
There was a problem hiding this comment.
I've just noticed, how is "windows" an arch?
There was a problem hiding this comment.
Remove it - we no longer support windows charms anyway
|
/build |
SimonRichardson
force-pushed
the
remove-32-bit-arches
branch
2 times, most recently
from
88a4885
to
08a7e6f
Compare
SimonRichardson
force-pushed
the
remove-32-bit-arches
branch
from
08a7e6f
to
baec899
Compare
There was a problem hiding this comment.
Should follow up and remove arch from utils, make it utils v4 so we don't accidentally start using it again. Especially with merges from 3.1 etc.
What are the odds there's a juju config running focal on i386 somewhere that might try to upgrade to juju 3.2? What happens if so?
core/arch/arches_test.go
Outdated
| raw string | ||
| arch string | ||
| }{ | ||
| {"windows", "windows"}, |
There was a problem hiding this comment.
Remove it - we no longer support windows charms anyway
|
/merge |
|
@hmlanigan RE:
I can't land it because there is a ssh test race which I'm unable to fix atm juju/utils#330 |
#15561 Merges the following patches: - #15558 - #15554 - #15555 - #15553 - #15552 - #15549 - #15538 - #15541 - #15534 - #15537 - #15535 - #15533 - #15529 - #15530 - #15522 - #15527 - #15523 - #15514 - #15524 - #15515 - #15520 - #15518 - #15510 - #15506 - #15500 - #15508 - #15497 - #15499 - #15492 - #15490 - #15487 ### Conflicts None.
#15565 Merges the following patches: - #15561 - #15558 - #15554 - #15555 - #15553 - #15552 - #15549 - #15538 - #15541 - #15534 - #15537 - #15535 - #15533 - #15529 - #15530 - #15522 - #15527 - #15523 - #15514 - #15524 - #15515 - #15520 - #15518 - #15510 - #15506 - #15500 - #15508 - #15497 - #15499 - #15492 - #15490 - #15487 ### Conflicts None.
This is the first step in fixing the problematic arch supported list from LXD. Whilst investigating that bug, it seems that we still support 32-bit word sizes. Except that, we no longer want to support 32-bit word sizes with dqlite, as the number of jujud permutations explodes and so it's best to completely remove the support. We officially dropped support for 32-bit architectures over 3 years ago, although we didn't actively remove the code to drop support.
This brings in what the juju/utils arch support had, with the addition of a UnsupportedArches field, which will ensure that we can at least say what we use to support, but no longer do. This doesn't bring in the changes from juju/utils PR as technically we don't need a new version of juju/utils sans arch package to do this.
Checklist
QA steps
$ juju bootstrap lxd test --build-agentBug reference
https://bugs.launchpad.net/juju/+bug/2013869