Thanks Simon and Jeremy!
--
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.
This bug was fixed in the package libsdl2 - 2.30.10+dfsg-1
Sponsored for Simon McVittie (smcv)
---
libsdl2 (2.30.10+dfsg-1) unstable; urgency=medium
* New upstream stable release
- Support multiple joystick buttons outputting the same gamepad button
- Improve portability of
> That means we can drop that delta (O2 and reproducible builds) on the
merge of the next version
Debian unstable now has 2.30.10, which I believe contains the relevant
upstream changes. Please resync if appropriate.
--
You received this bug notification because you are a member of Ubuntu
Bugs,
This fully migrated and works now.
Furthermore upstream was great and quick, they have fixes for the crash
and the LTO warning in the latest and the 2.30.x branch. Reproducible
builds is merged in both as well.
That means we can drop that delta (O2 and reproducible builds) on the
merge of the nex
This bug was fixed in the package libsdl2 - 2.30.9+dfsg-1ubuntu1
---
libsdl2 (2.30.9+dfsg-1ubuntu1) plucky; urgency=medium
* d/rules: reduce to -O2 to fix crashes that occur when combining O3
with LTO (LP: #2089779)
-- Christian Ehrhardt Thu, 28 Nov
2024 10:31:47 +0100
** C
Uploaded as 2.30.9+dfsg-1ubuntu1 and checking proposed migration in a
bit ...
--
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 notifi
** Merge proposal linked:
https://code.launchpad.net/~paelzer/ubuntu/+source/libsdl2/+git/libsdl2/+merge/477653
--
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 au
Tried O2 in variants via
- export DEB_CFLAGS_MAINT_APPEND = -O2
- export DEB_LDFLAGS_MAINT_APPEND = -Wl,-O2
Tested:
- ldflags + cflags = works
- cflags = works
- ldflags = still crashing
Ok, in that case cflags alone is the smallest (and more common) yet
working change to add.
--
You received t
Interesting build warning
libtool: link: ranlib build/.libs/libSDL2_test.a
libtool: link: ( cd "build/.libs" && rm -f "libSDL2_test.la" && ln -s
"../libSDL2_test.la" "libSDL2_test.la" )
In function 'SDL_memcpy_REAL',
inlined from 'SDL_SetCurrentDisplayMode' at
/home/ubuntu/SDL/src/video/SDL_
I used it more like a shotgun, but with local attribute((optimize(2)) in
the place of the warning and back up the stack of the crash I got still
a crash.
I'll use it a bit more widely and if working then refine that throughout
tomorrow as it is iterating "modify + run (for some time)".
If OTOH no
I checked if O3->O2 alone would also fix the issue - it does.
Julian had a great suggestion:
"""
I'd argue disable -O3; but it may also be fruitful to look where it crashes and
force -O2 for that function using `attribute((optimize(2))` (re @paelzer: Hi,
if LTO+O3 ...)
"""
Before the debugging
I could now, knowing that it is O3+LTO also reproduce it with upstream
git :-)
I checked the former 2.30.6 which would also behave the same way if LTO
and O3 are enabled, so at the end it wasn't the new version of libsdl2
at all.
I submitted it to upstream for their awareness.
Reproducible build
As a bonus of debugging this, the reproducible builds have fallen out as well.
The patch that I use works (that is what was tested above) and it is stable in
its binaries across 10 builds:
md5sum 0*/usr/lib/x86_64-linux-gnu/libSDL2-2.0.so.0*.objdump
cd47f413db025e99b4d888e84fac4a31
00/usr/lib/x
Second check good, the non-lto build makes it work.
buntu@autopkgtest:~$ /usr/games/wesnoth-1.18 -m --controller 1:ai --controller
2:ai --nogui
Battle for Wesnoth v1.18.3 x86_64
Started on Mon Dec 2 12:48:04 2024
Data directory: /usr/share/games/wesnoth/1.18
User configuration di
Just also adding -flto=auto to LDFLAGS (in addition to CFLAGS) did not
start to cause the issue on the git build.
But I think it is worth trying export
DEB_BUILD_MAINT_OPTIONS=optimize=-lto in the PPA builds.
And indeed the no-LTO build in:
https://launchpad.net/~paelzer/+archive/ubuntu/lp-208
I was re-verifying some of the major decision points to make sure this
goes the right paths (after all this crossed three digits of builds
overall) and as my gut feeling told me, something is even more odd.
I went back and ended up no more being able to reproduce the good case.
Not even with the P
New build, this time without debug-symbol and therefore no stripping on
build.
Results:
10x with "fixed reproducible build (or crap)" patch - 10x fail
10x rebuild "as-is" - 10x fail
So I have a single magic build that works in [1] and all others fail :-/
Even 10 of what should be the same.
?What
:-/ I accidentally built 20 times repro instead of 10/10 and one had a
failing builder (not build)
They ALL triggered the crash - at least it seems not flaky.
So maybe, it is more the debugsymbol stripping as assumed above.
Because in the try to be
Let me do a few without, at least it
Oddly bu
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+
** Attachment added: "good.buildinfo"
https://bugs.launchpad.net/ubuntu/+source/libsdl2/+bug/2089779/+attachment/5841032/+files/good.buildinfo
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2089779
** Attachment added: "bad.buildinfo"
https://bugs.launchpad.net/ubuntu/+source/libsdl2/+bug/2089779/+attachment/5841033/+files/bad.buildinfo
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2089779
T
** Attachment added: "good.buildlog"
https://bugs.launchpad.net/ubuntu/+source/libsdl2/+bug/2089779/+attachment/5841030/+files/good.buildlog
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2089779
T
** Attachment added: "bad.buildlog"
https://bugs.launchpad.net/ubuntu/+source/libsdl2/+bug/2089779/+attachment/5841031/+files/bad.buildlog
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2089779
Tit
Uploading libsdl2_2.30.9+dfsg-1build1.dsc
Uploading libsdl2_2.30.9+dfsg-1build1.debian.tar.xz
Uploading libsdl2_2.30.9+dfsg-1build1_source.buildinfo
Uploading libsdl2_2.30.9+dfsg-1build1_source.changes
Waiting for build and re-tests ...
--
You received this bug notification because you are a mem
Uh wait, the bad build in proposed was against ALL in -proposed (as of 13 days
ago).
But the PPA build was against plucky as-is.
Let me change the PPA dependencies to include proposed and rebuild once more ...
(just to make sure nothing still in -proposed is what breaks this and would
break the
The only reasonable thing in there with a function is the lib itself.
$ find good
good
good/usr
good/usr/share
good/usr/share/doc
good/usr/share/doc/libsdl2-2.0-0
good/usr/share/doc/libsdl2-2.0-0/README.md
good/usr/share/doc/libsdl2-2.0-0/CREDITS.txt
good/usr/share/doc/libsdl2-2.0-0/changelog.gz
Very interesting - both the 2.30.6 but even the
2.30.9+dfsg-1ubuntu1~pluckyppa1 no change rebuild work.
$ sudo apt install libsdl2-2.0-0=2.30.9+dfsg-1ubuntu1~pluckyppa1
..
works
So something was wrong when built as
https://launchpad.net/ubuntu/+source/libsdl2/2.30.9+dfsg-1
Which was better now t
Adding the cflgas from our build to the mix:
$ git clean -xfd
$ git clean checkout .
$ export CFLAGS="-Wdate-time -D_FORTIFY_SOURCE=3 -g -O2
-Werror=implicit-function-declaration -fno-omit-frame-pointer
-mno-omit-leaf-frame-pointer -flto=auto -ffat-lto-objects
-fstack-protector-strong -fstack-c
The test in infra against:
wesnoth-1.18/1:1.18.3-1 postgresql-common/267 python3-defaults/3.12.7-1
openmpi/5.0.5-6 boost1.83/1.83.0-3.2ubuntu3
worked.
Thereby all others are un-entangled
** Changed in: boost1.83 (Ubuntu)
Status: New => Fix Released
** Changed in: python3-defaults (Ubunt
Between versions could most likely be packaging, upstream or something in the
build env changing.
Most changes are in the upstream version, so let me try how
https://github.com/libsdl-org/SDL.git behaves.
This is only what is in
libsdl2-2.0-0:amd64 2.30.9+dfsg-1
Not the other s
The crash itself is, despite debuginfod, not more informative than the
following:
(gdb) bt
#0 __pthread_kill_implementation (threadid=, signo=6, no_tid=0)
at ./nptl/pthread_kill.c:44
#1 __pthread_kill_internal (threadid=, signo=6) at
./nptl/pthread_kill.c:78
#2 __GI___pthread_kill (threadid=,
I can recreate this locally, and really the new libsdl2 seems to be the
culprint.
Fine:
$ sudo ~/work/autopkgtest/autopkgtest/runner/autopkgtest --no-built-binaries
--apt-upgrade --shell wesnoth-1.18_1.18.3-1.dsc -- qemu --ram-size=8192 --cpus
2 ~/work/autopkgtest-plucky-amd64.img
Triggers the
Always passed in oracular
https://autopkgtest.ubuntu.com/packages/w/wesnoth-1.18/oracular/amd64
In Plucky it seems to fail mostly with:
libsdl2 | 2.30.9+dfsg-1
(oracular had 2.30.6+dfsg-1)
And at the same time still working otherwise
https://autopkgtest.ubuntu.com/packages/w/wesnoth-1.18/plucky
** Also affects: wesnoth-1.18 (Ubuntu)
Importance: Undecided
Status: New
--
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 n
34 matches
Mail list logo
