Re: New hipcc causes FTBFS (Re: Simplifying ROCm compiler upgrades)
On 2025-10-19 22:37, Christian Kastner wrote:
> I understand from this thread [2] that a /usr/bin/hipcxx was rejected,
> for sound reasons. However, see the workaround I suggested in #1118435
> of shipping this link in some private directory.
I needed a workaround for ggml now, so I went with dynamically querying
the version in d/rules [3].
This could also be a workaround for #1118435 for the other packages.
It's less pretty, but it works.
Best,
Christian
[3]: https://salsa.debian.org/deeplearning-team/ggml/-/commit/9a11ba492973759fdc2ae7d5e039e29bdf1c93bd
Reply to: