Re: Octave 3.6.2-rc0 release candidate available for ftp

[John W. Eaton]

On 12-May-2012, Thomas Yengst wrote:

| Scientific Linux 6.2 64bit. Routinely got warnings at each stage of 
configure, make, make check
| 
| Makefile:2510: warning: overriding commands for target `check'
| Makefile:2096: warning: ignoring old commands for target `check'
| Makefile:2510: warning: overriding commands for target `check'
| Makefile:2096: warning: ignoring old commands for target `check'
| Makefile:2510: warning: overriding commands for target `check'
| Makefile:2096: warning: ignoring old commands for target `check'

Those warnings are due to this change that Rik made:

  http://hg.savannah.gnu.org/hgweb/octave/rev/454bd50a5308

His reason for making the change:

  
https://mailman.cae.wisc.edu/pipermail/octave-maintainers/2011-February/023247.html

I don't know how to override the check target in automake without
getting the warning.

| During make - got a whole host of warnings, some of them look serious enough 
(e.g., casting uint to int). File attached.

The old-style casts warnings come from C code in header files (macros)
or generated code flex.  I don't know how much effort it is worth to
try to eliminate them, but I would prefer that we use C++-style casts
in C++ code, so we have the warning enabled.

We don't maintain gl2ps.c and it is not released as a library, so the
recommended way to use it is to include the sources directly in other
packages.  So the right place to fix the code and eliminate the
warnings from it would be upstream.

That really just leaves some signed/unsigned comparison warnings due
to comparing octave_idx_type (which must be signed, because values of
this type are passed to Fortran code) and size_t objects.  I'd welcome
patches to fix that problem in a good way.

jwe