- Package:
- nvidia-driver
- Source:
- nvidia-driver
- Submitter:
- Paul R. Tagliamonte
- Date:
- 2026-10-06 21:07:02 UTC
- Severity:
- normal
- Tags:
After an update to gcc-16 nvidia-current (and nvidia-tesla-550) FTBFS with:
nvidia/os-interface.c:24:10: fatal error: string.h: No such file or directory
24 | #include <string.h>
| ^~~~~~~~~~
Curiously, adding an import doesn't work, I suspect due to the
nostdinc flag being passed. There's likely some sort of real fix.
conftest.sh looks like it could be one such place but I didn't want to
muck around too much.
Paul
retitle 1148494 nvidia-graphics-drivers-tesla: dkms ftbfs due to kernel dropping strncpy for strscpy retitle 1148493 nvidia-driver: dkms ftbfs due to kernel dropping strncpy for strscpy thanks https://www.phoronix.com/news/Linux-7.2-Drops-strncpy I'll give this a whirl when I get some free cycles; but fixing the first failing call to strncpy to strscpy (along with some minor cleanup due to the api difference) moved the build along. Shouldn't be too bad. Not a header issue. Paul
tags 1148493 + patch tags 1148494 + patch tags 1148495 + patch kthxbyte Attached are two absolute messes of a debdiff, one for nvidia and one for nvidia-tesla. I don't think these should be applied without a grownup looking at what's going on here -- both are a mess. Hopefully its a net workload reduction for you. I started with the tesla patches (I didn't know about the dependency to nvidia's current dkms module), so the diffs are a bit of a mess and not exactly the same. There's also two patches in the tesla debdiff that should be one -- and I didn't make any DEP3 headers. There's a lot of YOLO in here and not my finest work. ``` $ uname -a Linux theseus 7.2.6+deb14-amd64 #1 SMP PREEMPT_DYNAMIC Debian 7.2.6-1 (2026-09-16) x86_64 GNU/Linux $ nvidia-smi Mon Sep 21 12:55:38 2026 +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 550.163.01 Driver Version: 550.163.01 CUDA Version: 12.4 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 Tesla PG500-216 Off | 00000000:06:00.0 Off | 0 | | N/A 36C P0 36W / 250W | 1MiB / 32768MiB | 1% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | No running processes found | +-----------------------------------------------------------------------------------------+ ``` Running a test program, it appears to be working still. Thank you for maintaining this! paultag
(Actually attached, apologies)
Hi, please for us, less experienced with DKMS, please can you attach diff to apply on DKMS tree? Or describe how to apply it? thanks
[Slavko 2026-09-23] I ran into this problem on a llama.cpp test machine. I tested the nvidia-graphics-drivers.debdiff patch, and can confirm that it work. I used 'apt-get source nvidia-kernel-dkms', then cd into the extracted directory and applied the patch using 'patch -p1 < nvidia-graphics-drivers.debdiff' before doing 'debuild' to generate a new dkms deb.
I was wondering why my dkms yolo-patch didn't match what i submitted to the tesla src package, and I realized that the 550 version was split into its own src package. attached is a debdiff for that src package too. same caviats apply, and same patches for the most part. paultag
[Petter Reinholdtsen] Is there some reason I fail to grasp that block this fix from being uploaded into unstable, or should someone just NMU it?