Re: notary arm64-specific failure: runtime: marked free object in span 0xffff5ec83ac0, elemsize=208 freeindex=3 (bad use of unsafe.Pointer or having race conditions? try -d=checkptr or -race)
On Fri, 21 Feb 2025 at 21:42, Nilesh Patra <nilesh@riseup.net> wrote:
> On 22 February 2025 6:19:44 am IST, Tianon Gravi <tianon@debian.org> wrote:
> >I managed to reproduce in a qemu-system-aarch64 VM so that I could
> >mess with this and try to figure out more information (it's not super
> >fast, but it *does* reproduce the problem, so that's a bonus - someone
> >with real hardware could probably iterate on this a lot faster 🙈).
>
> FYI, all DDs have access to porter boxes (which has every release architecture) which is "real" hardware.
>
> See: https://wiki.debian.org/PorterBoxHowToUse
Fair, thanks for the reminder! I usually shy away from the
porterboxes because in my experience they're slow and they're
restrictive against the (admittedly odd) ways I tend to debug/test
things, but it's a good callout!
On Tue, 18 Feb 2025 at 01:11, Simon Josefsson <simon@josefsson.org> wrote:
> Hmm, but those are related to illegal instructions, where the panic in
> our notary package is somehow related to use-after-free. 😭
On the plus side though, I managed to successfully git bisect in
between Go 1.23 and 1.24.0:
https://github.com/notaryproject/notary/issues/1708#issuecomment-2676121207
(perhaps someone can escalate to a person with more arm64 assembly
knowledge and/or the Go project? not sure if there's an easy way to
create a smaller reproducer than the "go test -count 100 -run
TestPublishDelegations ./client" I used there though, which would be
helpful for reporting it)
♥,
- Tianon
4096R / B42F 6819 007F 00F8 8E36 4FD4 036A 9C25 BF35 7DD4
Reply to: