ok. Now the dir.m is included in this bug report. Please see below.

yes, I retitled to be specifically about dir because the glob issue was a secondary issue at the user level while dir appeared to be an actual compatibility issue where matlab code would not execute as expected in octave.

If I understand correctly, Matlab doesn't have a 'glob' function.  The one referenced in this report is a function contributed to matlab central.  I don't think we need to worry about compatibility with random functions from that archive.  Octave's glob function was intended to provide the functionality available in the Unix function of the same name.  As far as I know, it does that job.

To get recursive expansion for "**" (in bash) I need to set

$ shopt -s globstar

I was not able to do the same withing octave.

Dmitri.
--

In bash shell it works differently. From octave build directory:

$ ls ../**
<...all octave files...>
$ ls ../**/*.m
../test/fntests.m  ../test/mk_bc_overloads_expected.m
$ ls ../*/*.m
../test/fntests.m  ../test/mk_bc_overloads_expected.m

in octave

octave:1> glob("../**")
ans =
{
  [1,1] = ../CITATION
  [2,1] = ../COPYING
  [3,1] = ../COPYRIGHT.md
  [4,1] = ../INSTALL
  [5,1] = ../Makefile.am
  [6,1] = ../Makefile.in
  [7,1] = ../README
  [8,1] = ../aclocal.m4
  [9,1] = ../autogen.sh
  [10,1] = ../autom4te.cache
  [11,1] = ../autopull.sh
  [12,1] = ../bootstrap
  [13,1] = ../bootstrap-funclib.sh
  [14,1] = ../bootstrap.conf
  [15,1] = ../build-aux
  [16,1] = ../config.in.h
  [17,1] = ../configure
  [18,1] = ../configure.ac
  [19,1] = ../doc
  [20,1] = ../etc
  [21,1] = ../examples
  [22,1] = ../gcc_opt
  [23,1] = ../gcc_tsan
  [24,1] = ../gcc_tsanO0
  [25,1] = ../gnulib
  [26,1] = ../libgnu
  [27,1] = ../libgui
  [28,1] = ../libinterp
  [29,1] = ../liboctave
  [30,1] = ../m4
  [31,1] = ../oct-conf-post-private.in.h
  [32,1] = ../oct-conf-post-public.in.h
  [33,1] = ../octave.doap
  [34,1] = ../run-octave.in
  [35,1] = ../scripts
  [36,1] = ../src
  [37,1] = ../test
}

octave:2> glob("../**/*.m")
ans =
{
  [1,1] = ../test/fntests.m
  [2,1] = ../test/mk_bc_overloads_expected.m
}

Dmitri.
--

Ok, confirming that in Matlab 2022b dir('**\*m') does recursion to full depth (seeing at least five levels deep in that working folder, so I'll assume full depth).  Octave still only does one level deep.  So that seems to be a legit compatibility request.

Regarding glob, I have no idea where that blogger got a working version of glob.  But searching Mathworks you get no function referenc, just that blog, and in Matlab 2022b with a pile of toolboxes:

>> glob('*.m')
Undefined function 'glob' for input arguments of type 'char'.

>> glob("*.m")
Unrecognized function or variable 'glob'.

>> help glob
--- glob not found. Showing help for globe instead. ---

 globe  Earth as sphere in 3-D graphics
...

>> help @string/glob

@string/glob not found.

(fixed markup)

ok.  by itself, glob is not recursive.

glob ('*.m') gives me all of the m files in the same folder.

glob ('**\*.m') appears to recurse one level deep.  so for example, in my project folder:

>> ls
octave-statistics  pkg  repelem  scatteredInterpolant  temp

all have m files, but only repelem and scatteredInterpolant have m files one level down. So:

>> glob('*.m')
ans = {}(0x0)

>> glob('**/*.m')
ans =
{
  [1,1] = repelem\repelem.m
  [2,1] = scatteredInterpolant\TriScatteredInterp.m
  [3,1] = scatteredInterpolant\delaunayn.m
  [4,1] = scatteredInterpolant\extraptest.m
  [5,1] = scatteredInterpolant\find_convex_vis_pts.m
  [6,1] = scatteredInterpolant\scatteredInterpolant.m
  [7,1] = scatteredInterpolant\testHDR.m

I'm not familiar with "filename globbing patterns .. as used by the Unix shells" so unless that pattern with glob is expected to do full recursion, this seems to be behaving okay

just to be clear, I think what you're saying is that you expect the glob function to recursively search the path below the octave source directory, and show all 6937 m-files, just as it shows the m files in your project directory.  But for the Octave program directory, it's returning empty.  similar for the dir command? (although you ran different dir commands for the two locations.)

is glob supposed to do a recursive search by default?

I'm struggling to understand which part of your examples is wrong.
Which output do you expect for which command?

Are there any .m files in 'C:\Program Files\GNU Octave\Octave-7.3.0\' (or in the immediate subfolders) for you? (There aren't any by default.)

What is the problem you'd like to solve?

Found some problem in dir / glob:

dir

cd('C:\Program Files\GNU Octave\Octave-7.3.0\');
pwd
ans = C:\Program Files\GNU Octave\Octave-7.3.0
dir('*.m')
warning: dir: nonexistent directory '*.m'
warning: called from
    dir at line 193 column 5


cd('D:\Projekte1\');
dir('**\*.m')
script001.m
script002.m
script003.m
...

glob

cd('C:\Program Files\GNU Octave\Octave-7.3.0\');
glob('**\*.m')
ans = {}(0x0)
----------------------------------------------------
cd('D:\Projekte1\');
glob('**\*.m')
ans =
{
  [1,1] = dir1\script001.m
  [2,1] = dir1\script002.m
  [3,1] = dir2\script003.m
...
}

traced it down to...

_wglob_('C:\Program Files\GNU Octave\Octave-7.3.0\**\*.m')
ans = {}(0x0)

_wglob_('D:\Projekte1\**\*.m')
ans =
{
  [1,1] = D:\Projekte1\dir1\script001.m
  [2,1] = D:\Projekte1\dir1\script002.m
  [3,1] = D:\Projekte1\dir2\script003.m
...
}

Is there a solution to this problem?

I also set all directory rights to full access for the directory
C:\Program Files\GNU Octave\Octave-7.3.0\