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)

To: debian-go@lists.debian.org
Subject: 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)
From: Jérémy Lal <kapouer@melix.org>
Date: Tue, 18 Feb 2025 10:37:28 +0100
Message-id: <[🔎] CAJxTCxyCfU0=9gD-z1PMF_v_ExjV+zRxgRi2pweXPwXYVH=yyg@mail.gmail.com>
In-reply-to: <[🔎] 87jz9n4nz9.fsf@josefsson.org>
References: <[🔎] 87jz9n4nz9.fsf@josefsson.org>

Le mar. 18 févr. 2025 à 10:11, Simon Josefsson <simon@josefsson.org> a écrit :

Hi

Does anyone here understand this error message and backtrace?

https://github.com/notaryproject/notary/issues/1708

It happens repeatedly only on arm64 and it looks serious, so I wouldn't
want to disable a self-check like this unless we have no other option.

Last known good build for comparison:

https://buildd.debian.org/status/fetch.php?pkg=notary&arch=arm64&ver=0.7.0%2Bgit20240416.9d2b3b3%2Bds1-2%2Bb4&stamp=1735520313&raw=0

It is using the same git commit as the now failing test, so I suspect
some change in our toolchain is triggering this.

Probably golang-1.24 has changed something that no longer applies to debian's arm64 supported features:

https://github.com/golang/go/commits/master/src/runtime/asm_arm64.s

Reply to:

Follow-Ups:
- 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)
  - From: Jérémy Lal <kapouer@melix.org>

References:
- 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)
  - From: Simon Josefsson <simon@josefsson.org>