gcc/libf2c/README

1998-04-20

This directory contains the f2c library packaged for use with g77 to configure
and build automatically (in principle!) as part of the top-level configure and
make steps.  This depends on the makefile and configure fragments in ../f.
g77 names this library `libg2c' to avoid conflict with existing copies
of `libf2c' on a system.

Some small changes have been made to the f2c distributions of lib[FI]77 which
come from <ftp:bell-labs.com/netlib/f2c/> and are maintained (excellently) by
David M. Gay <dmg@bell-labs.com>.  See the Notice files for copyright
information.  I'll try to get the changes rolled into the f2c distribution.

Files that come directly from netlib are either maintained in the
gcc/f/runtime/ directory under their original names or, if they
are not pertinent for g77's version of libf2c, under their original
names with `.netlib' appended.  For example, gcc/f/runtime/permissions.netlib
is a copy of f2c's top-level`permissions' file in the netlib distribution.
In this case, it applies only to the relevant portions of the libF77/ and
libI77/ directories; it does not apply to the libU77/ directory, which is
distributed under different licensing arrangements.  Similarly,
the `makefile.netlib' files in libF77/ and libI77/ are copies of
the respective `makefile' files in the netlib distribution, but
are not used when building g77's version of libf2c.

The `README.netlib' files in libF77/ and libI77/ thus might be
interesting, but should not be taken as guidelines for how to
configure and build libf2c in g77's distribution.

The packaging for auto-configuration was done by Dave Love <d.love@dl.ac.uk>.
Minor changes have been made by James Craig Burley <burley@gnu.org>,
who probably broke things Dave had working.  :-)

Among the user-visible changes (choices) g77 makes in its
version of libf2c:

-  f2c.h configured to default to padding unformatted direct reads
   (#define Pad_UDread), because that's the behavior most users
   expect.

-  f2c.h configured to default to outputting leading zeros before
   decimal points in formatted and list-directed output, to be compatible
   with many other compilers (#define WANT_LEAD_0).  Either way is
   standard-conforming, however, and you should try to avoid writing
   code that assumes one format or another.

-  dtime_() and etime_() are from Dave Love's libU77, not from
   netlib's libF77.