Add armv7h builds #723
Add armv7h builds #723
Conversation
Combines: fixup! Correct armv7 arch naming fixup! lopsided/archlinux arch name for armv7h fixup! Dockerfile tmpfs workaround arm32 on amd64 Adds a workaround mounting the .cargo directory as a tmpfs to avoid https://gitlab.com/qemu-project/qemu/-/issues/263
|
I don't really get this as, you're adding armv7 to the matrix but then also adding armv7 specific steps to the workflow. You also tweak the dockerfile a bit and am not quite sure why. |
|
Wait scratch that. It's been a while and I didn't realise how much of the CI was duplicated already. Still not sure what the Dockerfile change is for though. |
|
And scratch that again, Didn't realise explanation was in the commit message just not the PR message. This looks like it could be really good then. |
|
Works. Thanks for finding the fix. |
|
Build seems to be broken' any idea what's up? |
|
The armv7 build just ran fine on master. The aarch64 hung in the middle of
compiling one of the dependencies, though, so I think that's where the
issue's coming up.
…On Sat, Jun 25, 2022, 07:24 Lulu ***@***.***> wrote:
Build seems to be broken' any idea what's up?
—
Reply to this email directly, view it on GitHub
<#723 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AEDNSOKXJPPSWORSEDSSR7DVQ3T7XANCNFSM5TFMER3A>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
|
I actually figured it out to be cargo downloading the index filling memory and getting killed by oom. Not sure what caused this change as it's the same error rerunning the .10 release. Fixed with a nightly flag for now. |
|
Could be an update in cargo. Maybe try using the arch rust package instead
of rustup?
…On Mon, Jun 27, 2022, 13:34 Lulu ***@***.***> wrote:
I actually figured it out to be cargo downloading the index filling memory
and getting killed by oom. Not sure what caused this change as it's the
same error rerunning the .10 release.
Fixed with a nightly flag for now.
—
Reply to this email directly, view it on GitHub
<#723 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AEDNSOJT37UMUFCYNNTIIKTVRHQZFANCNFSM5TFMER3A>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
|
It may be due to some change in cargo but I don't really want to bisect it. If i do bisect it, it just means Paru will have to build on an old rust version in CI. A version that paru may no longer support. The nightly flag will do for now. |
|
Ok. It might be partially my fault, too, since this pull put the cache in
tmpfs as well.
…On Mon, Jun 27, 2022, 13:47 Lulu ***@***.***> wrote:
It may be due to some change in cargo but I don't really want to bisect
it. If i do bisect it, it just means Paru will have to build on an old rust
version in CI. A version that paru may no longer support. The nightly flag
will do for now.
—
Reply to this email directly, view it on GitHub
<#723 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AEDNSOIYQWKTBR5DFTVPAGDVRHSK3ANCNFSM5TFMER3A>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
|
You'd think but it still broke when I reverted that. |
|
It looks like this is a known issue with cargo: rust-lang/cargo#10583, and possibly the fault of libgit2/libgit2. Conveniently, libgit2 is also the culprit for the size mismatch on 32bit platforms. We can bypass it by exporting |
Adds arm7 builds to the workflows, closing #83
Since ALARM no longer supports
armorarmv6h, this should be the only arm binary still necessary.