Another integer overflow issue #138
Comments
lkeegan
added a commit
to lkeegan/pykdtree
that referenced
this issue
Jan 29, 2025
- correct the condition for using 32bit int indices - previous condition: `n_points < UINT_MAX` - however the data array in the C code has size `n_points * n_dim` and is indexed using a 32bit int type (see e.g. the PA macro) - this means results were incorrect for the case where `n_points < UINT_MAX` and `n_points * n_dim >= UINT_MAX` - new condition: `n_points * n_dim < UINT_MAX` - resolves storpipfugl#138 - render mako template - the generated _kdtree_core.c did not contain latest changes from _kdtree_core.c.mako (sorry my mistake) - this meant results were incorrect for the case `n_points < UINT_MAX` and `n_query * k >= UINT_MAX` - resolves storpipfugl#137 - use signed 64bit ints for the OpenMP loop on all platforms - to support OpenMP versions < 3 which only allow signed integers as loop counters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
When using a 32-bit integer for indexing, the data array in the C code has size
n_points * n_dimand is indexed using a 32bit int type (see e.g. the PA macro).Currently the condition to use a 32-bit int is that
n_points < UINT_MAX, but then the above can overflow.The correct condition should instead be
n_points * n_dim < UINT_MAXThe text was updated successfully, but these errors were encountered: