Hi Rouca, (re-including debian-lts ML, hope that is ok).
On So 15 Nov 2026 16:27:43 CET, Bastien Roucaries wrote:
Le dimanche 15 novembre 2026, 14:37:36 heure normale d’Europe centrale Mike Gabriel a écrit :Hi Rouca, On So 15 Nov 2026 12:21:21 CET, Bastien Roucaries wrote: > Hi, > > I have fixed every cve except CVE-2026-104041 (but i have asked help > from upstream) > > Could you review > > I use for generating build file > dpkg-buildpackage -S -nc -sa -k003A1A2DAA41085F > > But at top of trixie it fail with something weird > can't find file to patch at input line 23 > Perhaps you used the wrong -p or --strip option? > The text leading up to this was: > -------------------------- > |From: Dan Lavu <[email protected]> > |Date: Fri, 24 Jul 2026 14:13:49 -0400 > |Subject: tests: widen KCM TGT renewal poll window > | > |Use a 10s lifetime and 30s renewable life so renew_till stays past > |endtime, and poll for half-life plus slack after the UAF fix. > | > |Co-authored-by: Cursor <[email protected]> > |Model used: Cursor Sonnet 5 > | > |Reviewed-by: Alexey Tikhonov <[email protected]> > |Reviewed-by: Justin Stephenson <[email protected]> > |(cherry picked from commit 3efe248071fb37c57455dc5ac45b05bdf7c4e275) > |origin: backport,> https://github.com/SSSD/sssd/pull/9017/changes/ba17b16c41afadaad774f3c4c62563cda812630f> |--- > | src/tests/system/tests/test_kcm.py | 32 ++++++++++++++++++++++++++------ > | 1 file changed, 26 insertions(+), 6 deletions(-) > | > |Index: sssd/src/tests/system/tests/test_kcm.py > |=================================================================== > |--- sssd.orig/src/tests/system/tests/test_kcm.py 2026-11-15 > 11:28:56.955870597 +0100 > |+++ sssd/src/tests/system/tests/test_kcm.py 2026-11-15 > 11:28:56.953480561 +0100 > -------------------------- > No file to patch. Skipping patch. > 1 out of 1 hunk ignored > > > Quilt work. Can you help here > > rouca The problem with sssd in trixie (and before? Haven't checked so far) is that the orig tarball is different from the upstream-pulled-in source tree in the Git repository on Salsa. In src/tests/ there is a subfolder src/tests/tests/ that duplicates the test sources (and it is the complete test folder). As I have just picked up maintenance of sssd, I haven't found time to clean-up this mess, yet. The src/tests/tests/ code duplication in the orig tarball is not critical IMHO, but we need to make sure which part of the code is actually used during build for unit testing. The fix for not-applying CVE-2026-104034_3.diff is this: ``` diff --git a/debian/patches/CVE-2026-104034_3.diff b/debian/patches/CVE-2026-104034_3.diff index 7b2516f8e6..52a490861c 100644 --- a/debian/patches/CVE-2026-104034_3.diff +++ b/debian/patches/CVE-2026-104034_3.diff @@ -18,8 +18,8 @@ origin: backport, https://github.com/SSSD/sssd/pull/9017/changes/ba17b16c41afada Index: sssd/src/tests/system/tests/test_kcm.py =================================================================== ---- sssd.orig/src/tests/system/tests/test_kcm.py 2026-11-15 11:28:56.955870597 +0100 -+++ sssd/src/tests/system/tests/test_kcm.py 2026-11-15 11:28:56.953480561 +0100 +--- sssd.orig/src/tests/tests/system/tests/test_kcm.py 2026-11-15 11:28:56.955870597 +0100 ++++ sssd/src/tests/tests/system/tests/test_kcm.py 2026-11-15 11:28:56.953480561 +0100 @@ -383,12 +383,32 @@ with client.ssh("tuser", "Secret123") as ssh: ``` For a proper trixie-security upload, it needs to be investigate which of those files+folders under src/tests/(tests/) gets actually used at build time. My recommendation is to ignore src/tests/ and use src/tests/tests.Could you do please ?
I took a closer look. The flawed tarball originates from before https://github.com/SSSD/sssd/commit/f260623f95497cd7f629b4a3915ec2724889602c, where the tarball generation has been fixed.
Another, but bigger surgery would be repacking the orig tarball and clearing up that code-duplication mess in src/tests/ (and while at it, find out where it originated from). The sssd orig tarball of 2.13.1 (testing/unstable) does not have this problem, btw.Another solution is to ask upstream to made a LTS release or take last git at https://github.com/SSSD/sssd/tree/sssd-2-10
That would be nice, but this will take too long imho, to get security fixes landed in Debian.
For now, I'd suggest: + apply my above change to CVE-2026-104034_3.diff+ ignore all the upstream code that is on salsa (it is not helpful anyway, because debian/watch pulls in a pre-packed upstream tarball from Github, so no point in
having the upstream code from upstream Git synced into Salsa imho)
(need to discuss with the other maintainers)
+ just take the debian/ folder from Salsa
+ Combine it with the 2.10.1 orig tarball that is in Debian trixie already
+ and continue with your work on security fixes from there.
The src/tests/system/tests/test_kcm.py (and nothing else in
src/tests/system/, similar to src/tests/multihost/) will get used
during the Debian package build.
Cheers, Mike -- mike gabriel aka sunweaver (Debian Developer) mobile: +49 (1520) 1976 148 landline: +49 (4351) 486 14 27 GnuPG Fingerprint: 9BFB AEE8 6C0A A5FF BF22 0782 9AF4 6B30 2577 1B31 mail: [email protected], http://sunweavers.net
pgpS6PJX00Nmw.pgp
Description: Digitale PGP-Signatur
