Visitar URL original
Aborted (core dumped) with `filter1d` · Issue #4472 · GenericMappingTools/pygmt · GitHub
Skip to content

Aborted (core dumped) with filter1d #4472

New issue
New issue
Open
Open
Aborted (core dumped) with filter1d#4472
Labels
bugSomething isn't working

Description

@mdtanker
mdtanker
opened on Mar 26, 2026

Description of the problem

I'm experiencing a free(): invalid next size (fast) error when I run pygmt.filter1d. I've tested lots of environment variations with the below Minimal Complete Verifiable Examples. I can reliably reproduce the issues with an environment made with:

mamba env create -n pygmt_mkl python==3.14.3 numpy==2.4.3 pandas==3.0.1 pygmt==0.18.0 liblapack=*=*mkl

while an environment made with the below works as expected:

mamba env create -n pygmt_openblas python==3.14.3 numpy==2.4.3 pandas==3.0.1 pygmt==0.18.0 liblapack=*=*openblas

These seems to indicate it's a similar issue to #4038.

Outputs from working environment:

PyGMT information:
  version: v0.18.0
System information:
  python: 3.14.3 | packaged by conda-forge | (main, Feb  9 2026, 21:56:02) [GCC 14.3.0]
  executable: /home/sungw937/miniforge3/envs/pygmt_openblas/bin/python
  machine: Linux-5.15.167.4-microsoft-standard-WSL2-x86_64-with-glibc2.39
Dependency information:
  numpy: 2.4.3
  pandas: 3.0.1
  xarray: 2026.2.0
  packaging: 26.0
  contextily: None
  geopandas: None
  IPython: None
  pyarrow: None
  rioxarray: None
  gdal: 3.12.3
  ghostscript: 10.06.0
GMT library information:
  version: 6.6.0
  padding: 2
  share dir: /home/sungw937/miniforge3/envs/pygmt_openblas/share/gmt
  plugin dir: /home/sungw937/miniforge3/envs/pygmt_openblas/lib/gmt/plugins
  library path: /home/sungw937/miniforge3/envs/pygmt_openblas/lib/libgmt.so
  cores: 8
  grid layout: rows
  image layout: 
  binary version: 6.6.0

pygmt-session [DEBUG]: Revised options: @GMTAPI@-S-I-D-V-T-N-000000 -Fg1000 -N0 -Vd ->@GMTAPI@-S-O-D-D-A-Y-000001
filter1d [DEBUG]: Look for file 1000 in /home/sungw937/.gmt
filter1d [DEBUG]: Look for file 1000 in /home/sungw937/.gmt/cache
filter1d [DEBUG]: Look for file 1000 in /home/sungw937/.gmt/server
filter1d [INFORMATION]: Processing input table data
filter1d [DEBUG]: gmtapi_init_import: Passed family = Data Table and geometry = Line
filter1d [DEBUG]: gmtapi_init_import: Added 1 new sources
filter1d [DEBUG]: GMT_Init_IO: Returned first Input object ID = 0
filter1d [DEBUG]: gmtapi_begin_io: Input resource access is now enabled [container]
filter1d [DEBUG]: gmtapi_import_dataset: Passed ID = -1 and mode = 0
filter1d [INFORMATION]: Referencing data table from user 2 column arrays of length 1000
filter1d [DEBUG]: Object ID 2 : Registered Data Table Memory Reference 55a874c98030 as an Input resource with geometry Point [n_objects = 3]
filter1d [DEBUG]: gmtapi_import_dataset processed 1 resources
filter1d [DEBUG]: GMT_End_IO: Input resource access is now disabled
filter1d [DEBUG]: gmtapi_init_export: Passed family = Data Table and geometry = Line
filter1d [DEBUG]: gmtapi_init_export: Added 1 new destination
filter1d [DEBUG]: GMT_Init_IO: Returned first Output object ID = 1
filter1d [DEBUG]: GMT_Begin_IO: Mode value 1 not considered (ignored)
filter1d [DEBUG]: GMT_Begin_IO: Initialize record-by-record access for Output
filter1d [DEBUG]: gmtapi_next_io_source: Selected object 1
filter1d [INFORMATION]: Writing Data Table to memory copy supplied by pointer
filter1d [DEBUG]: GMT_Begin_IO: Output resource access is now enabled [record-by-record]
filter1d [DEBUG]: GMTAPI: GMT_Set_Geometry called but no object available
filter1d [INFORMATION]: Filter the data columns
filter1d [INFORMATION]: Read 1000 records from table 0, segment 0
filter1d [INFORMATION]: F width: 1000 Resolution: 10.01 Start: 500.501 Stop: 9499.5
filter1d [DEBUG]: GMTAPI: GMT_Put_Record (double) called before any segments declared
filter1d [DEBUG]: GMT memory: Initialize 2 temporary column double arrays, each of length : 0
filter1d [DEBUG]: Object ID 3 : Registered Data Table Memory Reference 55a874960570 as an Output resource with geometry Point|Line|Poly [n_objects = 4]
filter1d [DEBUG]: GMT_End_IO: Output resource access is now disabled
filter1d [DEBUG]: gmtlib_garbage_collection: Destroying object: C=0 A=0 ID=2 W=Input F=Data Table M=Memory Reference S=Used P=55a874c98030 N=(null)
filter1d [DEBUG]: GMTAPI_Garbage_Collection freed 1 memory objects
filter1d [DEBUG]: gmtlib_unregister_io: Unregistering object no 2 [n_objects = 3]
filter1d [DEBUG]: gmtlib_unregister_io: Unregistering object no 3 [n_objects = 2]
filter1d [DEBUG]: GMT memory: Free 2 temporary column arrays, each of length : 2097152

Minimal Complete Verifiable Example

import pandas as pd
import pygmt
import numpy as np
import faulthandler
faulthandler.enable()

data = pd.DataFrame(
    data={
        "distance_along_line": np.linspace(0,10000,1000),
        "Free_air":np.linspace(100, 10000, 1000),
    },
)

print(pygmt.show_versions())

filtered = pygmt.filter1d(
    data[["distance_along_line", "Free_air"]],
    end=False,
    time_col=0,
    filter_type="g1000",
    verbose="debug",
)

print(filtered)

Full error message

pygmt-session [DEBUG]: Revised options: @GMTAPI@-S-I-D-V-T-N-000000 -Fg1000 -N0 -Vd ->@GMTAPI@-S-O-D-D-A-Y-000001
filter1d [DEBUG]: Look for file 1000 in /home/sungw937/.gmt
filter1d [DEBUG]: Look for file 1000 in /home/sungw937/.gmt/cache
filter1d [DEBUG]: Look for file 1000 in /home/sungw937/.gmt/server
filter1d [INFORMATION]: Processing input table data
filter1d [DEBUG]: gmtapi_init_import: Passed family = Data Table and geometry = Line
filter1d [DEBUG]: gmtapi_init_import: Added 1 new sources
filter1d [DEBUG]: GMT_Init_IO: Returned first Input object ID = 0
filter1d [DEBUG]: gmtapi_begin_io: Input resource access is now enabled [container]
filter1d [DEBUG]: gmtapi_import_dataset: Passed ID = -1 and mode = 0
filter1d [INFORMATION]: Referencing data table from user 2 column arrays of length 1000
filter1d [DEBUG]: Object ID 2 : Registered Data Table Memory Reference 55e628994960 as an Input resource with geometry Point [n_objects = 3]
filter1d [DEBUG]: gmtapi_import_dataset processed 1 resources
filter1d [DEBUG]: GMT_End_IO: Input resource access is now disabled
filter1d [DEBUG]: gmtapi_init_export: Passed family = Data Table and geometry = Line
filter1d [DEBUG]: gmtapi_init_export: Added 1 new destination
filter1d [DEBUG]: GMT_Init_IO: Returned first Output object ID = 1
filter1d [DEBUG]: GMT_Begin_IO: Mode value 1 not considered (ignored)
filter1d [DEBUG]: GMT_Begin_IO: Initialize record-by-record access for Output
filter1d [DEBUG]: gmtapi_next_io_source: Selected object 1
filter1d [INFORMATION]: Writing Data Table to memory copy supplied by pointer
filter1d [DEBUG]: GMT_Begin_IO: Output resource access is now enabled [record-by-record]
filter1d [DEBUG]: GMTAPI: GMT_Set_Geometry called but no object available
filter1d [INFORMATION]: Filter the data columns
filter1d [INFORMATION]: Read 1000 records from table 0, segment 0
filter1d [INFORMATION]: F width: 1000 Resolution: 10.01 Start: 500.501 Stop: 9499.5
filter1d [DEBUG]: GMTAPI: GMT_Put_Record (double) called before any segments declared
filter1d [DEBUG]: GMT memory: Initialize 2 temporary column double arrays, each of length : 0
filter1d [DEBUG]: Object ID 3 : Registered Data Table Memory Reference 55e628a8fd20 as an Output resource with geometry Point|Line|Poly [n_objects = 4]
filter1d [DEBUG]: GMT_End_IO: Output resource access is now disabled
filter1d [DEBUG]: gmtlib_garbage_collection: Destroying object: C=0 A=0 ID=2 W=Input F=Data Table M=Memory Reference S=Used P=55e628994960 N=(null)
filter1d [DEBUG]: GMTAPI_Garbage_Collection freed 1 memory objects
filter1d [DEBUG]: gmtlib_unregister_io: Unregistering object no 2 [n_objects = 3]
filter1d [DEBUG]: gmtlib_unregister_io: Unregistering object no 3 [n_objects = 2]
filter1d [DEBUG]: GMT memory: Free 2 temporary column arrays, each of length : 2097152
free(): invalid next size (fast)
Aborted (core dumped)

System information

PyGMT information:
  version: v0.18.0
System information:
  python: 3.14.3 | packaged by conda-forge | (main, Feb  9 2026, 21:56:02) [GCC 14.3.0]
  executable: /home/sungw937/miniforge3/envs/pygmt_mkl/bin/python
  machine: Linux-5.15.167.4-microsoft-standard-WSL2-x86_64-with-glibc2.39
Dependency information:
  numpy: 2.4.3
  pandas: 3.0.1
  xarray: 2026.2.0
  packaging: 26.0
  contextily: None
  geopandas: None
  IPython: None
  pyarrow: None
  rioxarray: None
  gdal: 3.12.3
  ghostscript: 10.06.0
GMT library information:
  version: 6.6.0
  padding: 2
  share dir: /home/sungw937/miniforge3/envs/pygmt_mkl/share/gmt
  plugin dir: /home/sungw937/miniforge3/envs/pygmt_mkl/lib/gmt/plugins
  library path: /home/sungw937/miniforge3/envs/pygmt_mkl/lib/libgmt.so
  cores: 8
  grid layout: rows
  image layout: 
  binary version: 6.6.0

Activity

  1. mdtanker
    added
    bugSomething isn't working
    on Mar 26, 2026

  2. weiji14 commented on Mar 26, 2026

    @weiji14
    weiji14
    on Mar 26, 2026
    Member

    Oof, this mkl vs openblas situation again. I was fighting this a couple of week ago at conda-forge/lapack-feedstock#81 (comment) (try not to go down the rabbit hole), and my general impression is that even though these different blas variants should be compatible (https://conda-forge.org/docs/maintainer/knowledge_base/#blas), the whole situation is still a mess. Can you share the output of conda list | grep mkl for your pygmt_mkl environment, and conda list | grep blas for the pygmt_openblas environment? Want to check what versions of blas dependent libraries are installed.

    Cc @seisman who might know a bit more if there's an issue with the C code as well.

  3. mdtanker commented on Mar 27, 2026

    @mdtanker
    mdtanker
    on Mar 27, 2026
    ContributorAuthor

    Yeah I think I'll stay out of that rabbit hole 😂

    Here's the output from the pygmt_mkl env:

      libblas              3.11.0        6_h5875eb1_mkl         conda-forge
  libcblas             3.11.0        6_hfef963f_mkl         conda-forge
  liblapack            3.11.0        6_h5e43f62_mkl         conda-forge
  mkl                  2025.3.1      h0e700b2_10            conda-forge

    and the output from the pygmt_openblas env:

      libblas              3.11.0        6_h4a7cf45_openblas   conda-forge
  libcblas             3.11.0        6_h0358290_openblas   conda-forge
  liblapack            3.11.0        6_h47877c9_openblas   conda-forge
  libopenblas          0.3.32        pthreads_h94d23a6_0   conda-forge

  4. weiji14 commented on Mar 31, 2026

    @weiji14
    weiji14
    on Mar 31, 2026
    Member

    Hmm yeah, I can reproduce the segfault on my Linux computer with your mamba env create -n pygmt_mkl ... + Python commands. The error goes away if I do a mamba install "libblas=3.11.*=*openblas" (could be a temporary workaround) to use the openblas variant.

    I also tried upgrading to the GMT dev version using mamba install -c conda-forge/label/dev gmt=6.7.0.dev2+8931e47, and got the same error. So might be something that needs to be fixed upstream in GMT, similar to what was done in GenericMappingTools/gmt#8755, but somewhere else?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions