Figure.meca causes segmentation fault when given "event_name" via a DataFrame

[wasjabloch]

Description of the problem

When calling fig.meca(data, scale=".3") where data is a pandas DataFrame with a column called "event_name", a segmentation fault occurs.

Minimal Complete Verifiable Example

import pandas as pd
import pygmt
R = [-125, -122.25, 46, 48.5]
J = "M5c"
dat = pd.DataFrame(
    {"longitude": [-122.597],
     "latitude": [47.014],
     "depth": [50.0],
     "strike": [14.0],
     "dip": [82.0],
     "rake": [-135.0],
     "magnitude":  [6.7],
     "year": [1949],
     "size": [0.67],
     "event_name": ["1949 M6.7"]
     }
)
fig = pygmt.Figure()
fig.meca(dat, scale=".3", region=R, projection=J)

Full error message

free(): double free detected in tcache 2
Aborted (core dumped)

System information

PyGMT information:
  version: v0.9.0
System information:
  python: 3.10.10 | packaged by conda-forge | (main, Mar 24 2023, 20:08:06) [GCC 11.3.0]
  executable: /home/xxx/.conda/envs/gmt/bin/python
  machine: Linux-5.4.0-121-generic-x86_64-with-glibc2.31
Dependency information:
  numpy: 1.24.3
  pandas: 2.0.1
  xarray: 2023.4.2
  netCDF4: 1.6.3
  packaging: 23.1
  contextily: None
  geopandas: None
  ghostscript: 9.54.0
GMT library information:
  binary version: 6.4.0
  cores: 32
  grid layout: rows
  image layout: 
  library path: /home/xxx/.conda/envs/gmt/lib/libgmt.so
  padding: 2
  plugin dir: /home/xxx/.conda/envs/gmt/lib/gmt/plugins
  share dir: /home/xxx/.conda/envs/gmt/share/gmt
  version: 6.4.0

[welcome]

👋 Thanks for opening your first issue here! Please make sure you filled out the template with as much detail as possible. You might also want to take a look at our contributing guidelines and code of conduct.

[wasjabloch]

P.S.: I just saw that pygmt v0.9.0 is out. I'll now try to confirm the bug with the most recent version.

[wasjabloch]

Confirmed with the most recent version. The bug report is now up to date.

[seisman]

Thanks. I can reproduce the issue. Will see why it happens later.

[seisman]

@PaulWessel This looks like an upstream bug to me, could you please see if you can verify and it?

Here is a minimal example to reproduce the issue:

import pandas as pd
import pygmt

pygmt.show_versions()
dat = pd.DataFrame(
    {
        "longitude": [-122.597],
        "latitude": [47.014],
        "depth": [50.0],
        "strike": [14.0],
        "dip": [82.0],
        "rake": [-135.0],
        "magnitude": [6.7],
        "event_name": ["Eventname"],
    }
)
fig = pygmt.Figure()
fig.meca(dat, scale="0.3", region=[-125, -122.25, 46, 48.5], projection="M5c")
fig.show()

Here are the verbose messages:

PyGMT information:
  version: v0.9.1.dev1+gaaa0ca006.d20230405
System information:
  python: 3.11.2 | packaged by conda-forge | (main, Mar 31 2023, 17:54:27) [Clang 14.0.6 ]
  executable: /Users/seisman/opt/miniconda/bin/python
  machine: macOS-13.3.1-x86_64-i386-64bit
Dependency information:
  numpy: 1.24.2
  pandas: 2.0.0
  xarray: 2022.12.0
  netCDF4: 1.6.3
  packaging: 23.0
  contextily: 1.2.0
  geopandas: 0.12.2
  IPython: 8.7.0
  rioxarray: 0.13.3
  ghostscript: 9.54.0
GMT library information:
  binary version: 6.5.0_f6d1b51_2023.06.16
  cores: 8
  grid layout: rows
  image layout:
  library path: /Users/seisman/opt/GMT-debug/lib/libgmt.dylib
  padding: 2
  plugin dir: /Users/seisman/opt/GMT-debug/lib/gmt/plugins
  share dir: /Users/seisman/opt/GMT-debug/share
  version: 6.5.0
gmtset [DEBUG]: gmtlib_get_graphics_item: Fig: 0 Subplot: 2 Panel: () Inset: 0
pygmt-session [DEBUG]: Exit:  gmt_reload_settings
pygmt-session [DEBUG]: Enter: gmtlib_plot_C_format
pygmt-session [DEBUG]: Exit:  gmtlib_plot_C_format
pygmt-session [DEBUG]: Enter: gmtinit_get_history
pygmt-session [DEBUG]: gmtlib_get_graphics_item: Fig: 0 Subplot: 2 Panel: () Inset: 0
pygmt-session [DEBUG]: Initialize FFTW with 8 threads.
pygmt-session [DEBUG]: GMT_Create_Session initialized GMT structure
pygmt-session [DEBUG]: Shared Library # 0 (core). Path = /Users/seisman/opt/GMT-debug/lib/libgmt.dylib
pygmt-session [DEBUG]: Loading GMT plugins from: /Users/seisman/opt/GMT-debug/lib/gmt/plugins
pygmt-session [DEBUG]: Shared Library # 1 (supplements). Path = /Users/seisman/opt/GMT-debug/lib/gmt/plugins/supplements.so
pygmt-session [DEBUG]: GMT now running in modern mode [Session ID = 61391]
pygmt-session [DEBUG]: GMT_Call_Command string: gmt figure d140f05165714c95a3a42cd454074ce1 -
pygmt-session [DEBUG]: Revised options: d140f05165714c95a3a42cd454074ce1 -
figure [DEBUG]: No figure file /Users/seisman/.gmt/sessions/gmt_session.61391/gmt.figures - nothing to do
figure (gmt_init.c:19445(gmt_add_figure)): FYI: gmt_M_free given a NULL pointer - ignored
figure [DEBUG]: New figure: 1	d140f05165714c95a3a42cd454074ce1	-
pygmt-session [DEBUG]: Exit:  gmt_reload_settings
pygmt-session [DEBUG]: Enter: gmtlib_plot_C_format
pygmt-session [DEBUG]: Exit:  gmtlib_plot_C_format
pygmt-session [DEBUG]: Enter: gmtinit_get_history
pygmt-session [DEBUG]: gmtlib_get_graphics_item: Fig: 1 Subplot: 2 Panel: () Inset: 0
pygmt-session [DEBUG]: Initialize FFTW with 8 threads.
pygmt-session [DEBUG]: GMT_Create_Session initialized GMT structure
pygmt-session [DEBUG]: Shared Library # 0 (core). Path = /Users/seisman/opt/GMT-debug/lib/libgmt.dylib
pygmt-session [DEBUG]: Loading GMT plugins from: /Users/seisman/opt/GMT-debug/lib/gmt/plugins
pygmt-session [DEBUG]: Shared Library # 1 (supplements). Path = /Users/seisman/opt/GMT-debug/lib/gmt/plugins/supplements.so
pygmt-session [DEBUG]: GMT now running in modern mode [Session ID = 61391]
pygmt-session [DEBUG]: GMT_Call_Command string: gmt figure d140f05165714c95a3a42cd454074ce1 -
pygmt-session [DEBUG]: Revised options: d140f05165714c95a3a42cd454074ce1 -
pygmt-session [DEBUG]: Exit:  gmt_reload_settings
pygmt-session [DEBUG]: Enter: gmtlib_plot_C_format
pygmt-session [DEBUG]: Exit:  gmtlib_plot_C_format
pygmt-session [DEBUG]: Enter: gmtinit_get_history
pygmt-session [DEBUG]: gmtlib_get_graphics_item: Fig: 1 Subplot: 2 Panel: () Inset: 0
pygmt-session [DEBUG]: Initialize FFTW with 8 threads.
pygmt-session [DEBUG]: GMT_Create_Session initialized GMT structure
pygmt-session [DEBUG]: Shared Library # 0 (core). Path = /Users/seisman/opt/GMT-debug/lib/libgmt.dylib
pygmt-session [DEBUG]: Loading GMT plugins from: /Users/seisman/opt/GMT-debug/lib/gmt/plugins
pygmt-session [DEBUG]: Shared Library # 1 (supplements). Path = /Users/seisman/opt/GMT-debug/lib/gmt/plugins/supplements.so
pygmt-session [DEBUG]: Pad argument (2) ignored in initialization of Data Table
pygmt-session [DEBUG]: Initializing a vector for handing external Input
pygmt-session [DEBUG]: Object ID 0 : Registered Vector Memory Reference 7fd069174830 as an Input resource with geometry Point [n_objects = 1]
pygmt-session [DEBUG]: Successfully created a new Vector container to represent a Data Table
pygmt-session [DEBUG]: Set_Object for family: 0
==> 1 API Objects at end of GMT_Create_Data
--------------------------------------------------------
K.. ID RESOURCE.... FAMILY.... ACTUAL.... DIR... S O M L
--------------------------------------------------------
* 0  0 7fd069174830 Data Table Vector     Input  0 Y N 0
--------------------------------------------------------
pygmt-session [DEBUG]: VirtualFile name created: @GMTAPI@-S-I-D-V-T-N-000000
pygmt-session [DEBUG]: GMT now running in modern mode [Session ID = 61391]
pygmt-session [DEBUG]: GMT_Call_Command string: gmt meca @GMTAPI@-S-I-D-V-T-N-000000 -JM5c -R-125/-122.25/46/48.5 -Sa.3
pygmt-session [DEBUG]: Use PS filename /Users/seisman/.gmt/sessions/gmt_session.61391/gmt_1.ps-
pygmt-session [DEBUG]: gmtinit_get_current_panel: No current panel selected so not in subplot mode
pygmt-session [DEBUG]: Revised options: @GMTAPI@-S-I-D-V-T-N-000000 -JM5c -R-125/-122.25/46/48.5 -Sa.3
meca [DEBUG]: History: Process -JM5c
meca [DEBUG]: History: Process -R-125/-122.25/46/48.5
meca [DEBUG]: Map distance calculation will be using great circle approximation with authalic auxiliary latitudes and authalic (R_2) radius = 6371007.1809 m, in meter.
meca [DEBUG]: Look for file -125/-122.25/46/48.5 in /Users/seisman/.gmt
meca [DEBUG]: Look for file -125/-122.25/46/48.5 in /Users/seisman/.gmt/cache
meca [DEBUG]: Look for file -125/-122.25/46/48.5 in /Users/seisman/.gmt/server
meca [DEBUG]: Got regular w/e/s/n for region (-125/-122.25/46/48.5)
meca [DEBUG]: Reset MAP_ANNOT_OBLIQUE to anywhere
meca [INFORMATION]: Central meridian not given, default to -123.625
meca [DEBUG]: Projected values in meters: -153064 153064 5.7496e+06 6.15843e+06
meca [DEBUG]: Computed automatic parameters using dimension scaling: 0.843708
meca [INFORMATION]: Map scale is 61.2257 km per cm or 1:6.12257e+06.
meca [DEBUG]: Running in PS mode modern
meca [DEBUG]: Use PS filename /Users/seisman/.gmt/sessions/gmt_session.61391/gmt_1.ps-
meca [DEBUG]: Create hidden PS file /Users/seisman/.gmt/sessions/gmt_session.61391/gmt_1.ps-
meca [DEBUG]: Got session name as pygmt-session and default graphics formats as pdf
meca [DEBUG]: Basemap order: Frame = above  Grid = below  Tick/Annot = below
meca [INFORMATION]: Linear magnitude scaling selected, normalizing by 5.
meca [DEBUG]: gmtapi_init_import: Passed family = Data Table and geometry = Point
meca (gmt_api.c:7824(gmtapi_init_import)): FYI: gmt_M_free given a NULL pointer - ignored
meca [DEBUG]: gmtapi_init_import: Added 1 new sources
meca [DEBUG]: GMT_Init_IO: Returned first Input object ID = 0
meca [DEBUG]: gmtapi_begin_io: Input resource access is now enabled [container]
meca [DEBUG]: gmtapi_import_dataset: Passed ID = -1 and mode = 0
meca [INFORMATION]: Referencing data table from user 7 column arrays of length 1
meca [DEBUG]: Object ID 1 : Registered Data Table Memory Reference 60000203ad00 as an Input resource with geometry Point [n_objects = 2]
meca [DEBUG]: gmtapi_import_dataset processed 1 resources
meca [DEBUG]: GMT_End_IO: Input resource access is now disabled
meca [DEBUG]: Set_Object for family: 6
==> 2 API Objects at end of GMT_Read_Data
--------------------------------------------------------
K.. ID RESOURCE.... FAMILY.... ACTUAL.... DIR... S O M L
--------------------------------------------------------
* 0  0 7fd069174830 Vector     Vector     Input  2 Y N 0
* 1  1 60000203ad00 Data Table Data Table Input  2 Y N 1
--------------------------------------------------------
meca [DEBUG]: GMT_End_IO: Input resource access is now disabled
meca [INFORMATION]: Number of records read: 1
meca [DEBUG]: Current size of half-baked PS file /Users/seisman/.gmt/sessions/gmt_session.61391/gmt_1.ps- = 22710.
==> 2 API Objects at end of GMTAPI_Garbage_Collection entry
--------------------------------------------------------
K.. ID RESOURCE.... FAMILY.... ACTUAL.... DIR... S O M L
--------------------------------------------------------
* 0  0 7fd069174830 Vector     Vector     Input  2 Y N 0
* 1  1 60000203ad00 Data Table Data Table Input  2 Y N 1
--------------------------------------------------------
meca [DEBUG]: gmtlib_garbage_collection: Destroying object: C=0 A=0 ID=1 W=Input F=Data Table M=Memory Reference S=Used P=60000203ad00 N=(null)
meca [DEBUG]: GMTAPI_Garbage_Collection freed 1 memory objects
meca [DEBUG]: gmtlib_unregister_io: Unregistering object no 1 [n_objects = 1]
==> 1 API Objects at end of GMTAPI_Garbage_Collection exit
--------------------------------------------------------
K.. ID RESOURCE.... FAMILY.... ACTUAL.... DIR... S O M L
--------------------------------------------------------
* 0  0 7fd069174830 Vector     Vector     Input  2 Y N 0
--------------------------------------------------------
python(61391,0x7ff85d8e6340) malloc: *** error for object 0x24bfb385c9b0: pointer being freed was not allocated
python(61391,0x7ff85d8e6340) malloc: *** set a breakpoint in malloc_error_break to debug
[1]    61391 abort      python test.py

[PaulWessel]

I ran this via Xcode etc but no crash, at least it did not stop at a line like yours. It ended with these message:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/pwessel/UH/RESEARCH/CVSPROJECTS/GMTdev/pygmt/pygmt/helpers/decorators.py", line 598, in new_module
    return module_func(*args, **kwargs)
  File "/Users/pwessel/UH/RESEARCH/CVSPROJECTS/GMTdev/pygmt/pygmt/helpers/decorators.py", line 738, in new_module
    return module_func(*args, **kwargs)
  File "/Users/pwessel/UH/RESEARCH/CVSPROJECTS/GMTdev/pygmt/pygmt/src/meca.py", line 383, in meca
    lib.call_module(module="meca", args=build_arg_string(kwargs, infile=fname))
  File "/Users/pwessel/UH/RESEARCH/CVSPROJECTS/GMTdev/pygmt/pygmt/clib/session.py", line 621, in call_module
    raise GMTCLibError(
pygmt.exceptions.GMTCLibError: Module 'meca' failed with status code 45:

Unfortunately, if not crashing on macOS you may have to run in ddd in Linux...

[seisman]

Here are my understanding of the issue:

In PyGMT, we first call

GMT_Create_Data(API, GMT_IS_DATASET|GMT_VIA_VECTOR, GMT_IS_POINT, GMT_CONTAINER_ONLY, [7, 1, 1, 0], NULL, NULL, NULL, NULL, NULL)

to create the data container, and then call GMT_Put_Vectors and GMT_Put_Strings (family=GMT_IS_VECTOR|GMT_IS_DUPLICATE) to put the vectors/strings into the container. This creates the API object 0:

--------------------------------------------------------
K.. ID RESOURCE.... FAMILY.... ACTUAL.... DIR... S O M L
--------------------------------------------------------
* 0  0 7fd069174830 Vector     Vector     Input  2 Y N 0

Then in GMT, the psmeca module calls the GMT_Init_IO and GMT_Read_Data functions (L780 in psmeca.c https://github.com/GenericMappingTools/gmt/blob/8bf949a0ece1a3af56ba64e00cd979010d751854/src/seis/psmeca.c#L780). The GMT_Read_Data function calls gmtapi_import_dataset and then we go to the case GMT_IS_REFERENCE|GMT_VIA_VECTOR branch (https://github.com/GenericMappingTools/gmt/blob/8bf949a0ece1a3af56ba64e00cd979010d751854/src/gmt_api.c#L4091).

In the case GMT_IS_REFERENCE|GMT_VIA_VECTOR branch, the function creates a GMT_DATASET container and hooks up the vectors and the trailing text

...
S->data[col_pos_out] = V_obj->data[col_pos].f8;
SH->alloc_mode[col_pos_out] = GMT_ALLOC_EXTERNALLY;	
...
if (smode) S->text = V_obj->text;	/* Hook up trailing text array */

Now we have two API objects:

--------------------------------------------------------
K.. ID RESOURCE.... FAMILY.... ACTUAL.... DIR... S O M L
--------------------------------------------------------
* 0  0 7fd069174830 Vector     Vector     Input  2 Y N 0
* 1  1 60000203ad00 Data Table Data Table Input  2 Y N 1
--------------------------------------------------------

At the end of the psmeca module, we call the gmtlib_garbage_collection (and the gmtlib_free_dataset_ptr function) which frees the API object 1 (i.e., the Dataset container). The vector columns S->data[col] are not freed because alloc_mode[col] was set to GMT_ALLOC_EXTERNALLY, but the trailing text S->text are freed. It also frees the trailing texts of the original GMT_VECTOR object (i.e., V_obj->text) because S->text is a reference of V_obj->text.

Thus, at the end of the GMT session, when we try to free the API object 0 (i.e., V_obj->text), we see the pointer being freed was not allocated error.

I hope my understanding is correct and it can help you find a fix for this issue.

[PaulWessel]

Hm, since the data columns have their flag

SH->alloc_mode[col_pos_out] = GMT_ALLOC_EXTERNALLY; /* Not this objects job to free what was passed in by reference */

it would seem we need a similar alloc_mode for the trailing text to protect it. In gmt_hidden.h the GMT_DATASEGMENT_HIDDEN struct for datasegments has

`enum GMT_enum_alloc *alloc_mode;	/* Allocation mode per column [GMT_ALLOC_INTERNALLY] */`

So I will add

`enum GMT_enum_alloc alloc_mode_text;	/* Allocation mode for trailing text [GMT_ALLOC_INTERNALLY] */`

and then consult this variable before free. Sees sensible?

[seisman]

So I will add

`enum GMT_enum_alloc alloc_mode_text;	/* Allocation mode for trailing text [GMT_ALLOC_INTERNALLY] */`

and then consult this variable before free. Sees sensible?

Yes, it will very likely fix the issue!

[PaulWessel]

Not so. And, if I try to place a stop point in psmeca it never stops there, It stops in the places I asked (Put_Strings etc in the API) but pretty soon we are in the GarbageCollector and no evidence it ever went into meta. Not sure what to do.

[seisman]

Your changes in PR GenericMappingTools/gmt#7557 work well for me after applying the changes in https://github.com/GenericMappingTools/gmt/pull/7557/files#r1234079875.