Debian Patches
Status for simgrid/4.1-6
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| java-MANIFEST.patch | no | |||||
| ns3-47.diff | Adapt to the recent ns3 API change. Picked from upstream. diff --git a/examples/cpp/network-ns3/s4u-network-ns3-timed.tesh b/examples/cpp/network-ns3/s4u-network-ns3-timed.tesh index 23ec6c55de..dab74fe29d 100644 |
no | ||||
| ns3-48.diff | Fix FTBFS with libns3 | yes | debian | 2026-08-10 | ||
| pybind-302.diff | Try to adapt our python bindings to pybind11 v3.0.2 See https://github.com/pybind/pybind11/pull/5533 diff --git a/src/bindings/python/simgrid_python.cpp b/src/bindings/python/simgrid_python.cpp index 97aabb175a..faa6bac1a8 100644 |
Martin Quinson <martin.quinson@ens-rennes.fr> | no | 2026-02-21 | ||
| eigen5.diff | Adapt to Eigen 5 Eigen 5.0 switched its versioning scheme, so Eigen3ConfigVersion is messed up by a version name such as "3.3". We thus have to manually check the version rather than trusting Eigen3ConfigVersion.cmake. . Recent Eigen3Config.cmake (Eigen >= 5.0, and some backports) also stopped setting the legacy EIGEN3_VERSION_STRING/EIGEN3_INCLUDE_DIR variables. Use the standard CMake-provided Eigen3_VERSION and the Eigen3::Eigen imported target instead, which are always set by find_package() regardless of what Eigen's own config script does. . See https://gitlab.com/libeigen/eigen/-/issues/2970 See https://gitlab.com/libeigen/eigen/-/issues/2486 |
no | ||||
| stack_growth_fallback.diff | For some reason, the try_run() seems to fail sometimes on official build daemons, as https://buildd.debian.org/status/fetch.php?pkg=simgrid&arch=arm64&ver=4.1-4 reports that PTH_STACKGROWTH is 1, meaning that the stack grows up while it goes down on that arch, which is utterly wrong. My best guess is that some virtualization fools cmake to belive that it's a cross-compilation, bypassing try_run(). We could extend the fallbacks a bit, but actually the list of architectures on which SimGrid will ever be run is not very large, so hardcode the expected value instead in a way that should cover the Debian architectures. The fools trying to build SimGrid on another architecture will have to type 6 to 10 chars to make it work. |
no |