Mail Archives: djgpp/1998/01/10/11:45:21
From: | Chris Croughton <crough45 AT amc DOT de>
|
Newsgroups: | comp.os.msdos.djgpp
|
Subject: | GMP v2.0.2 DJGPP sources
|
Date: | Sat, 10 Jan 1998 17:10:23 +0100
|
Message-ID: | <34B79D6F.41C6@amc.de>
|
NNTP-Posting-Host: | bob.bob.bofh.org
|
MIME-Version: | 1.0
|
Lines: | 72
|
To: | djgpp AT delorie DOT com
|
DJ-Gateway: | from newsgroup comp.os.msdos.djgpp
|
GNU Multi-Precision maths library.
OK, I've finally managed to upload a reconstituted
version[1] of these to my web site:
http://www.keris.demon.co.uk/bin/gmp202.zip
Note that it's not linked to from the rest of
the site, you need to go directly to that URL.
Note also that, although (as far as I can tell) it
works it has some interesting differences from the
official release.
(a) the names of some files have had to be changed
to eliminate multiple dots (and name conflicts),
these names have also had to be changed in
references to the files from other files (like
in #includes and in make.bat etc.)
(b) Some code is in the wrong files! This makes no
difference to using the library, but looks strange
if you try to do a diff from the original or look
at the source. See below for why...
The library is still built using MAKE.BAT rather than
a proper makefile. This means that everything is
rebuilt every time instead of only when the object files
are out of date. I find this very annoying but I don't
have time at present to do anything about it. If someone
does and wants to send me the makefile(s) please do...
[1] OK, the reasons for the comments:
I originally got this to work quite a time ago, then
didn't bother with it except for linking with the library
for some of my projects. Then someone (several people,
in fact) asked about it so I said I'd put it on my web
site.
That's where the problems came. In a moment of mental
aberation I managed to delete all of the files instead
of just the objects. Oops - this is a heavily nested
tree! I had a look at my backups (without disturbing
my disk - I'm suspicious!) and found that I didn't have
one with the modified sources. (BAD programmer! No
chocolate chip cookies!)
So, dig out Norton Utilities and undelete - which wants
to know the first letter of each file. Fortunately, I
did have a (pre-modification) directory of the sources,
so I was able to restore all the directories.
In the process, however, names got mixed up. The code
has lots of names like:
f_div.c
t_div.c
x_div.c
and restoring just picked the first one. So in the
sources you'll find files called f_div.c which contain
routines called t_div and vice versa. As I said above,
this doesn't matter for the archive because the linker
looks at the symbols in the files not the file names,
and since all are built they're all there. It just
looks silly, and sometime I'll try to do something
about it. But not now...
Good luck using it...
Chris C
- Raw text -