Ok, this loves to stay weird :-/ The rebuild fails as well!
Which leaves very little difference left :-/

PPA build yesterday, works
https://launchpad.net/ubuntu/+source/libsdl2/2.30.9+dfsg-1build1/+build/29317183

Archive build yesterdday, fails
https://launchpad.net/ubuntu/+source/libsdl2/2.30.9+dfsg-1build1/+build/29317183

I'll attach the files to allow others to have a look ...

Meaningful diff in buildinfo:
- dbgsym being build (not default enabled in PPAs, but I could ...)

--- good.buildinfo      2024-11-29 07:50:07.884080493 +0100
+++ bad.buildinfo       2024-11-28 14:00:19.000000000 +0100
@@ -1,27 +1,33 @@
 Format: 1.0
 Source: libsdl2
-Binary: libsdl2-2.0-0 libsdl2-dev libsdl2-doc libsdl2-tests
+Binary: libsdl2-2.0-0 libsdl2-2.0-0-dbgsym libsdl2-dev libsdl2-doc 
libsdl2-tests libsdl2-tests-dbgsym
...checksums...
 Environment:
- DEB_BUILD_OPTIONS="noautodbgsym parallel=4"
+ DEB_BUILD_OPTIONS="parallel=4"
  DEB_BUILD_PROFILES="noudeb"
  LANG="C.UTF-8"
  LC_COLLATE="C.UTF-8"
  LC_CTYPE="C.UTF-8"
- SOURCE_DATE_EPOCH="1732797461"
+ SOURCE_DATE_EPOCH="1732796406"


Meaningful diff in buildlog (after unifying version and time [which was <1h 
minutes apart, and after ordering build commands)
- no dbgsym building
- some doc rendering and compression differs
- changelog compression and deduplication differs (no pkgstripfiles for PPA 
builds)
- test and build output sometimes order slightly different but seem to be the 
same
- debug-prefix-map is given in "" or not ???
- tests use seeds which seem to slightly change the test output

IMHO none of these should change the binary and it's behavior dramatically.
If anything then the pkgstrip?
It is not a reproducible build, and interestingly the notes I see in 
https://tests.reproducible-builds.org/debian/rb-pkg/trixie/amd64/libsdl2.html 
point to the very same SDL_DYNAPI_entry I found in my objdump analysis.


Note on repro:
As part of the work on version 3.x (still in preview) this was changed to 
python there 
https://github.com/libsdl-org/SDL/commit/8b4f5f09c1daf9f1d0ae689c31b676bea42eee11
But for now it it is perl as here 
https://github.com/libsdl-org/SDL/blob/release-2.30.9/src/dynapi/gendynapi.pl
Should (tm) just be a sort()?

experimental steps from here
0. ponder how to check if pkgstripfiles is what breaks this
1. try to fix this to be reproducible
2. build reproducible variant multiple times (ppa, amd64 only, proposed 
enabled, building dbgsyms)
3. build non-reproducible as-is multiple times (ppa, amd64 only, proposed 
enabled, building dbgsyms)

Check if:
- are ppa builds flaky with some working and some not?
- if so is the reproducible one reliably one or the other?

Maybe we spot something that way, and if not then if the reproducible
set includes good/bad it should be better comparable.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2089779

Title:
  Buffer overflow in autopkgtest of wesnoth

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/boost1.83/+bug/2089779/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to