[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[netCDF #IQO-210917]: Fortran 90 manpages missing in netcdf 3.6.2 beta3

> Hi Ed,
> 
> > I believe that this is correct and usual with autoconf/automake/
> > libtool packages. There is no way for me to know what optimization
> > flags might work on your platform. So how can I set them?
> 
> Actually I later noticed that the optimization flags were set
> properly by configure (all -O2). I was premature here. Sorry.
> However, the -g flag is automatically added to.

BTW autoconf does turn on -g -O2 for any GNU C compiler automatically. It's not 
me! 
So if you don't want that you must use CPPFLAGS.

> 
> >> 2) There is no easy way to switch off debugging (unless I do the
> >> above AND specify AM_CPPFLAGS=../libsrc in calling make)
> >
> > You mean the -g option? That should no longer be turned on anywhere
> > by default. If it is, that's a bug. Please let me know. All flags
> > must be set by the user, for debugging or optimization.
> 
> The files ncdump/Makefile.in and fortran/Makefile.in do have the
> following variable set:
> 
> AM_FLAGS = -g .....(etc)...
> 
> This makes at least the fortran sources compile with the -g flag on,
> irrespective of "configure".

Thanks for pointing this out. I've removed those now and that will be in
the next release.

> Furthermore, "configure" sets the -g flag on all of CFLAGS, CXXFLAGS,
> FFLAGS and FCFLAGS. This leads to debugging code to be installed in
> all objects.
> 
> I let that be for the time being, and used "make install-strip" to
> get rid of it again. I haven't figured out to avoid that "configure"
> adds the -g flags. That's something for you to figure out.
> 
> >> 3) The manpages for Fortran 90 are missing (there is no file netcdf.
> >> 3f90)
> >
> > Thanks for pointing this out! I never noticed that the F90 man
> > pages wasn't being installed. I have fixed this for the next release.
> 
> I saw that netcdf.3 and netcdf.3f are both made out of netcdf.m4. But
> netcdf.3f90 is completely separate and does not seem to come from the
> same source. Wouldn't you want to include the making of netcdf.3f90
> out of netcdf.m4?

No, the fortran and C APIs are nearly identical. So we generate those from a
common source. The F90 man page is totally different and exists on it's own.

> 
> >> Furthermore, since man will only find netcdf.3f and netcdf.3f90
> >> when they are in directories man3f and man3f90, it may be wise to add
> >> symbolic links man3f -> man3 and man3f90 -> man3
> >
> > Sorry, I don't follow you here. Could you elaborate?
> 
> netcdf.3, netcdf.3f, and netcdf.3f90 all end up in the directory
> $prefix/share/man/man3
> 
> If you do "man netcdf" you always get the C manual.
> To get the fortran manual, you can use "man 3f netcdf" or "man 3f90
> netcdf" for the F90 manual.
> But this ONLY works when they are in a directory $prefix/share/man/
> man3f and $prefix/share/man/man3f90, respectively.
> 
> So the installation should either make the man pages directly in the
> corresponding directories, or all in man3 and make symbolic links in
> $prefix/share/man:
> man3f -> man3
> man3f90 -> man3
> 
> Does that make sense?

Yes, thanks. I am looking into all this right now. I think the correct answer
will be to have the man pages installed in their proper directories, and I am 
now trying to figure out how to do that in automake.

> 
> Thanks for looking into this all. The installation on Mas OS X has
> already immensely improved since 3.6.1! These are the only small
> lingering issues with the installation, and then it will work
> beautifully.
> 

Wonderful!

Ed

Ticket Details
===================
Ticket ID: IQO-210917
Department: Support netCDF
Priority: Normal
Status: Closed