Debian Patches

Status for rocsolver/7.2.4-1~exp1

Patch Description Author Forwarded Bugs Origin Last update
use-local-mathjax.patch use local mathjax
The sphinx.ext.mathjax extension defaults to loading mathjax from a
CDN, which results in the lintian warning 'privacy-breach-generic'.
Use a local copy of mathjax to prevent that problem.		 Cordell Bloor <cgmb@slerp.xyz> not-needed 2023-07-17
verbose-build-of-specialized-kernels.patch verbose build of specialized kernels
These files take so long to build on slower ppc64el and arm64 systems
that the build may time out due to inactivity. This is typically when
building for ten or more GPU architectures. Add the verbose flag so that
there is output printed as the compiler finishes building a translation
unit for each architecture.		 Cordell Bloor <cgmb@slerp.xyz> not-needed 2024-03-19
Extend-docs-conf.py-for-offline-build.patch Extend docs/conf.py for offline build
By setting these extra variables, we can suppress a remote call which
would cause the build to fail.		 Christian Kastner <ckk@debian.org> not-needed 2025-07-15
fix-getf2-synchronization-bug.patch [rocSOLVER] Fix GETF2 synchronization bug (#3743)
## Motivation
Fix a race condition within the kernel `getf2_small_kernel`.
## Technical Details
`common` is an LDS memory allocation first used to write the elements of
column k, which is later overwritten by the row that is found as the
pivot. This can cause work items of a wave to expect to read elements of
column k, but instead read elements of a the pivot row that another wave
has written to shared memory.		 John <121990449+john00003@users.noreply.github.com> yes upstream upstream, https://github.com/ROCm/rocm-libraries/commit/48896a2adf99c6018074a063d75725e43da34af5 2026-03-05
fix-stebz-synchronization-bug.patch [rocsolver] Fix stebz synchronization bug (#4735)
## Motivation
STEBZ and downstream functions show intermittent failing tests in some
architectures. A synchronization error in the splitting kernel is the
cause of this issue.
## Technical Details
The bug can be fixed by adding an extra thread synchronization to avoid
racing read-write conditions.		 Juan Zuniga-Anaya <50754207+jzuniga-amd@users.noreply.github.com> yes upstream upstream, https://github.com/ROCm/rocm-libraries/commit/95b16a977be7dba18b24ab232a5537b6f1310eda 2026-03-05
increase-sygvdx_inplace-test-tolerance.patch [rocSOLVER] Further increase sygvdx inplace test tolerance (#5403)

## Motivation
PR https://github.com/ROCm/rocm-libraries/pull/4436 didn't increase the
tolerance of SYGVDX_INPLACE tests enough and the test is still failing
on some architectures (gfx1201 and gfx1150).
## Technical Details

The initial patch from Angelo Gonzales (PR 4436 mentioned above)
altered the tolerance from 3 * n to 8 * n while this one raises it to
12 * n thus refreshing the patch.		 Angelo Gonzales <angonzal@amd.com> yes upstream upstream, https://github.com/ROCm/rocm-libraries/commit/a01ddb275e0b7a99d02236c9147f0749d766c74e 2026-03-23
skip-test-if-vram-is-insufficient.patch [rocSOLVER] Skip test if VRAM is insufficient (#3886)
## Motivation
Large tests will fail if the device does not have enough VRAM. This PR
introduces a method to skip tests if memory allocation fails due to
insufficient VRAM.
## Technical Details
This PR adds the following:
1. Propagate device memory allocation error from client container
initialization to `memcheck()` call. This helps to discern between
out-of-memory error and other hip errors so tests aren't unnecessarily
skipped.
2. Use macro to skip test if memory allocation failed due to
insufficient VRAM.
3. Use a similar macro to check if rocblas memory allocation fails.
4. Currently this is only applied to large tests. **Should this be
applied to all tests?**		 Angelo Gonzales <angonzal@amd.com> yes upstream upstream, https://github.com/ROCm/rocm-libraries/commit/d1c171fd516775f8d60f7edbb33903aaa4bcf5ce 2026-03-05
fix-buffer-overflow-causing-test-fails.patch Fix buffer overflow (#1019)
Fix buffer overflow in `dot` device function which is causing ASAN test
failures.		 Angelo Gonzales <angonzal@amd.com> yes upstream upstream, https://github.com/ROCm/rocm-libraries/commit/5ecfb5741a1f0584f1d9b249d4a952e183803c90 2026-03-05
fix-getri-in-rocsolver-failing.patch Fix GETRI in rocsolver failing for n=1023 (#1954)
## Motivation

<!-- This PR attempts to fix the problem that routine GETRI in rocsolver
would fail for n=1023. -->
This PR attempts to fix the problem that routine GETRI in rocsolver
would fail for n=1023
## Technical Details
<!-- The amount of scratch space for using rocblas TRSM is obtained by
calling rocblas_internal_trsm_workspace_max_size_64() -->
The amount of scratch space for using rocblas TRSM is obtained by
calling rocblas_internal_trsm_workspace_max_size_64()		 Ed D'Azevedo <98369794+EdDAzevedo@users.noreply.github.com> yes upstream upstream, https://github.com/ROCm/rocm-libraries/commit/a05aa9a323b3f0e29c8b7be63ed905aeba2c4e20 2026-03-08

All known versions for source package 'rocsolver'

Links