Debian Bug report logs - #1003922
recastnavigation: reproducible-builds: BuildId differences triggered by RPATH

Reply or subscribe to this bug.

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Vagrant Cascadian <vagrant@reproducible-builds.org>

To: submit@bugs.debian.org

Subject: recastnavigation: reproducible-builds: BuildId differences triggered by RPATH

Date: Mon, 17 Jan 2022 22:00:25 -0800

[Message part 1 (text/plain, inline)]

Source: recastnavigation
Severity: normal
Tags: patch
User: reproducible-builds@lists.alioth.debian.org
Usertags: buildpath
X-Debbugs-Cc: reproducible-bugs@lists.alioth.debian.org

The RPATH contains the build path resulting in different buildid:

  https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/recastnavigation.html

The attached patch to debian/rules passes
-DCMAKE_BUILD_RPATH_USE_ORIGIN=ON via a dh_auto_configure override,
which should use a relative path for RPATH.

Alternately, updating the packaging to debhelper compat level 14 should
fix this, although it is currently an experimental compat level.

With this patch applied, recastnavigation should build reproducibly on
tests.reproducible-builds.org!

Thanks for maintaining recastnavigation!

live well,
  vagrant

[0001-debian-rules-Pass-DCMAKE_BUILD_RPATH_USE_ORIGIN-ON-v.patch (text/x-diff, inline)]

From 4b8a78935baaffeb65671f36fe1a1695a34ada4b Mon Sep 17 00:00:00 2001
From: Vagrant Cascadian <vagrant@reproducible-builds.org>
Date: Tue, 18 Jan 2022 05:58:20 +0000
Subject: [PATCH] debian/rules: Pass -DCMAKE_BUILD_RPATH_USE_ORIGIN=ON via
 dh_auto_configure override.

This avoids embedding the full path in RPATH, which triggers BuildId
differences.

https://tests.reproducible-builds.org/debian/issues/unstable/cmake_rpath_contains_build_path_issue.html
---
 debian/rules | 1 +
 1 file changed, 1 insertion(+)

diff --git a/debian/rules b/debian/rules
index 1bc7077..afb14dc 100755
--- a/debian/rules
+++ b/debian/rules
@@ -25,6 +25,7 @@ DEB_CMAKE_OPTIONS ?= -DCMAKE_VERBOSE_MAKEFILE=ON \
                 -DRECASTNAVIGATION_DEMO=OFF \
 		-DRECASTNAVIGATION_TESTS=OFF \
 		-DRECASTNAVIGATION_EXAMPLES=OFF \
+		-DCMAKE_BUILD_RPATH_USE_ORIGIN=ON \
                 ..
 
 override_dh_auto_configure:
-- 
2.34.1

[signature.asc (application/pgp-signature, inline)]

Message #10 received at 1003922@bugs.debian.org (full text, mbox, reply):

From: bret curtis <psi29a@gmail.com>

To: vagrant@reproducible-builds.org, 1003922@bugs.debian.org

Subject: Re: Bug#1003922: recastnavigation: reproducible-builds: BuildId differences triggered by RPATH

Date: Tue, 18 Jan 2022 00:41:00 -0800

[Message part 1 (text/plain, inline)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Thank you for the patch, it's applied along with a changelog update.

It's awaiting review and upload from a DD. :)

Cheers,
Bret

On 2022-01-18 at 06:00, vagrant@reproducible-builds.org wrote:
> Source: recastnavigation
> Severity: normal
> Tags: patch
> User: reproducible-builds@lists.alioth.debian.org
> Usertags: buildpath
> X-Debbugs-Cc: reproducible-bugs@lists.alioth.debian.org
>
> The RPATH contains the build path resulting in different buildid:
>
>
https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/recastnavigation.html

>
> The attached patch to debian/rules passes
> -DCMAKE_BUILD_RPATH_USE_ORIGIN=ON via a dh_auto_configure override,
> which should use a relative path for RPATH.
>
> Alternately, updating the packaging to debhelper compat level 14 should
> fix this, although it is currently an experimental compat level.
>
> With this patch applied, recastnavigation should build reproducibly on
> tests.reproducible-builds.org!
>
> Thanks for maintaining recastnavigation!
>
> live well,
>   vagrant
-----BEGIN PGP SIGNATURE-----
Version: FlowCrypt Email Encryption 8.2.1
Comment: Seamlessly send and receive encrypted email

wsFzBAEBCAAGBQJh5n0bACEJEMI3e31YTLkwFiEE6+UVEsAp/iYIDx+Wwjd7
fVhMuTAJ8hAAjUd+DUyV26EckH60Fn5qp/93j+iuO7FOIJzCvyjFg6DYXzZy
rCQtRmZiMrYumgVYlPvw9xD9PtdxGZ7/YZzp7hfml14xubrK5QDpc5FZ8VMc
PIrHE9yfttylYQV4yztb1lZOINheyg95cHlSFdZ9KaOvzqeY2xndPLco0tXt
jVMfAIoNojmcAPVPssHSofuFqzx5QhffPRO0CsLushjv1uBagh1eayCRnJV6
doGvIw5P/JuigA9AjUD3s1V+tJYROecsQE1eqiYYHKFdn4CmVj4kOYtcEcEO
OXwxjO7FJEj1sc/UtQiuS5MK7Jk414fG99pFdrq+EkYoW1iKNJ6cizAuzfMJ
1MmaLmQWDmRtxztY7ySoQPBoWRftXvabupTEQvvJQjU3sbzE+0lyZ9z1aeFj
YFnWO0nNp+aLVvt23WEmWGPO6RlOV9CiCDKdBLhGxZT8763o1EoPaPoQvLKU
DBqgq2eElGOCPDppyV9fMSELUwR3WyByBLgXoXOOO5ij4myFwE6FlZjIYYax
5d6bMgMAqRMNa5xmaksCJM8gKzzzS/luMb9TiG7VmgmhDPNxGFMe5xxpGO9f
uaIMSio4eeMsFPEvaAZ1qMgwdLoflkDy5nH+YnEu2Z7gabjceLx5QSRQU+dy
iGxQup8LHS2PXP6lIBQk9jGPRxkHGTL9mCg=
=+Byt
-----END PGP SIGNATURE-----

[0xC2377B7D584CB930.asc (application/pgp-keys, attachment)]

Message #15 received at 1003922@bugs.debian.org (full text, mbox, reply):

From: Vagrant Cascadian <vagrant@reproducible-builds.org>

To: bret curtis <psi29a@gmail.com>

Cc: 1003922@bugs.debian.org

Subject: Re: Bug#1003922: recastnavigation: reproducible-builds: BuildId differences triggered by RPATH

Date: Tue, 18 Jan 2022 10:46:24 -0800

[Message part 1 (text/plain, inline)]

On 2022-01-18, bret curtis wrote:
> It seems like the patch didn't fix this?
>
> https://salsa.debian.org/games-team/recastnavigation/-/jobs/2377314

Hmmm. That only shows that they were different, not what was different.

Is there a way you can download the artifacts and compare them with
diffoscope locally? Can you configured the pipeline to call reprotest
using diffoscope instead of just comparing which files differ (the
default for reprotest is to use diffoscope if available)?


I just re-ran my test with reprotest, and it still builds reproducibly
with the patch, and produces differences without the patch.


It also looks to be reproducible in the bookworm suite:

  https://tests.reproducible-builds.org/debian/rb-pkg/bookworm/amd64/recastnavigation.html

The only real difference between bookworm and unstable there is bookworm
does not vary the build path, but varies most of the things reprotest
does.


There is some chance that the locale differences reprotest uses are
causing differences, as reprotest tests some very obscure locales by
default, though it sometimes picks different locales...


That said, I'm pretty confident that the patch submitted fixes the build
path issues; there may be other issues that may be non-deterministic, or
only triggered in certain environments...


live well,
  vagrant

[signature.asc (application/pgp-signature, inline)]

Send a report that this bug log contains spam.