Memory leak in TDFieldExporter2D::export_field

[samcunliffe]

Not sure why yet, but we have a segmentation violation (I think from the 2D field exporter)...

➤ tdms in_pstd_cyl.mat output.mat
[2022-10-10 16:47:51.417] [debug] Parsing 3 command line arguments
infile:[in_pstd_cyl.mat], outfile:[output.mat], m=0
[2022-10-10 16:47:51.417] [debug] Finished parsing arguments
[2022-10-10 16:47:51.425] [info] Using 8 OMP threads
[2022-10-10 16:47:51.425] [info] Isource is empty
[2022-10-10 16:47:51.425] [info] Jsource is empty
[2022-10-10 16:47:51.425] [debug] f_ex_vec has ndims=2 N=1
[2022-10-10 16:47:51.425] [info] interpolation method = 1
[2022-10-10 16:47:51.425] [info] exi not present
[2022-10-10 16:47:51.425] [info] eyi not present
[2022-10-10 16:47:51.425] [info] Ni_tdf=276 Nk_tdf=276
[2022-10-10 16:47:51.425] [info] Refractive index = 1.35001
[2022-10-10 16:47:52.470] [debug] init_diff_shift_op, delta=-0.5
[2022-10-10 16:47:52.470] [debug] init_diff_shift_op, delta=-0.5
[2022-10-10 16:47:52.470] [debug] init_diff_shift_op, delta=-0.5
[2022-10-10 16:47:52.470] [debug] init_diff_shift_op, delta=0.5
[2022-10-10 16:47:52.470] [debug] init_diff_shift_op, delta=0.5
[2022-10-10 16:47:52.470] [debug] init_diff_shift_op, delta=0.5
[2022-10-10 16:47:52.470] [info] Np=2 Nt=500 Npe=250 f_max=2.94577e+14 Npraw=0.4606
dims:(254,1,254)
[2022-10-10 16:47:52.476] [debug] Starting main loop
[2022-10-10 16:47:52.488] [info] Saving field
[2022-10-10 16:47:52.488] [debug] Entered TDFieldExporter2D::export_field
[2022-10-10 16:47:52.488] [debug] Entered field written to array, creating output filename
[2022-10-10 16:47:52.488] [debug] output folder name: ??
[2022-10-10 16:47:52.488] [debug] output file name: ??/ex_000000.mat
time domain output: ??/ex_000000.mat
[2022-10-10 16:47:52.488] [debug] About to write matlab_array to the MATLAB output file
zsh: segmentation fault  tdms in_pstd_cyl.mat output.mat

Seems to only appear once debug logging was turned on. Which is rather alarming.

[samcunliffe]

[Notes to myself so I don't forget]

Seems to be restricted to my machine. Cannot reproduce on Myriad, and does not happen on the GitHub actions build.

The problem seems to be the TDFieldExporter2D.folder_name which gets given garbage.

(lldb) thread backtrace
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x20)
  * frame #0: 0x00000001092f3690 libmat.dylib`___lldb_unnamed_symbol1400 + 16
    frame #1: 0x00000001092f366f libmat.dylib`mat_developer_api::matPutVariable(MATFile*, char const*, mxArray_tag const*) + 63
    frame #2: 0x0000000100037640 tdms`TDFieldExporter2D::export_field(SplitField&, int, int) const + 736
    frame #3: 0x000000010005c801 tdms`mexFunction + 66385
    frame #4: 0x00000001000057a9 tdms`main + 601
    frame #5: 0x000000020013552e dyld`start + 462

[t-young31]

could just get rid of MATLAB..?

[samcunliffe]

In this case, I think it's our fault. For some reason, we're giving matPutVariable some memory garbage filename.