From: Crispen, Bob (Robert.Crispen@HSV.Boeing.com)
Date: Thu Oct 14 1999 - 15:17:59 CDT
Kent Landfield[SMTP:kent@landfield.com] sez:
>As I know very little about cygwin other than where to get it...
>Could someone write up a bulleted "how-to" such as...
>
> Getting cigwin32:
> Installing cigwin32:
> Unpacking hypermail:
> Building Hypermail:
> - run configure
> - Compile hypermail
> - If it does not compile
> Installing hypermail:
> Concerns running hypermail on a win32 system:
>
>This would be VERY helpful to a large section of the community.
Here's some more -- note: correction from last time.
Getting cygwin:
* Go to http://sourceware.cygnus.com/cygwin/ (I'll call this
the "cygwin home").
* Click "Download Sites" in the box on the left.
* Choose a download site close to you.
* When you click on a link on that page, you'll always be
taken to a directory. Click on "latest" to go to the
directory containing the latest release.
* There are several packages there. I recommend getting
"full.exe". This is the package that has everything
(compiler, shell tools, etc.)
Installing cygwin
* While you *could* simply double click on full.exe, I
recommend that you read the documentation first.
* Go to the Cygwin home and click "Documentation" in the
box on the left. Then click on Cygwin User's Guide.
* Read the documentation. The more you read, the happier
you'll be.
* Double click on full.exe and install cygwin.
* In the current version, you need to do a few things with
the shortcut that's created in
C:\Windows\Start Menu\Programs\Cygnus Solutions
before you can fire up the shell.
Using Windows Explorer, go to that folder and right
mouse on Cygwin B20. Select "Properties". You'll
probably want to make the following changes:
* On the "Program" tab, check "Close on exit"
* On the "Memory" tab set "Initial Environment"
to some high number (e.g., 4096). This avoids
a bug in Windows' DOS shell where auto
allocating memory fails.
* On the "Misc" tab, uncheck "Fast Pasting".
This will let you use copy and paste to paste
text into the bash shell.
* On the "Screen" tab, select Initial size: 50 lines.
This will give you as big a window as you can get.
* On the "Font" tab, select TT 8 x 16 to start with.
This gives you full-sized characters for the window.
* Double-click on Cygwin B20 or use the Start menu to bring
up the shell.
* You'll get a warning message that bash can't find the
/tmp directory. Fix that by typing:
mkdir /tmp /bin /usr /etc
This is in the instructions you should have read.
NOTE: CORRECTION!!!
* In bash, type:
ln -s /cygnus/cygwin~1/h-i586~1/bin/sh.exe /bin/sh.exe
(If you've installed cygwin somewhere other than the default
location, you'll have to make the left-hand side of the link
expression something else).
This creates a soft link that's recognizeable by bash. If
you use Explorer to make a shortcut, what you'll get in /bin
is a file called "she.exe.lnk" which unfortunately bash
doesn't recognize as executable.
You need to make the link (or a copy of sh.exe) because many
scripts (e.g., configure, many makefiles) will call /bin/sh.
This is also in the instructions you should have read.
* NOTE: if you don't like the idea of creating all those
directories in the C: root, you absolutely can use links
and mount to put them somewhere else. On my system at
home, I've got /usr/bin pointed to the Cygnus bin
directory and /usr/local pointed somewhere entirely
different, and I've got /bin pointing to /usr/bin. This
falls under advanced customization, and I don't recommend
it unless you have a lot of time to play around with the
mount command, which is somewhat different in cygwin (both
from Unix and often from their own documentation).
* To exit from bash type
exit
You should *not* just click the "X" box in the top right-hand
corner of the window. bash can leave your system in a
blue-screen-of-death state if you don't exit properly.
It's a lot better than it used to be about this, but it's
still dangerous to just kill off bash.
Testing the Compiler:
* Create a "hello.c" file somewhere. You know what you need
to type in that file.
* In your bash shell, go to the folder where you've put
hello.c (use cd) and type:
gcc hello.c
* When the code compiles and links successfully (you can read
what it's doing on the caption of the bash window), type
./a.exe
and you should get the familiar "Hello, world!".
* Create another file called "hello.cpp". You know what you need
to type in that file.
Incidentally, you may #include <iostream> or <iostream.h>
* Type in
g++ -o hello hello.cpp
* Execute hello.exe:
./hello
* If you haven't got Quick View installed on your system, do the
following:
* Open the control panel
* Select Add/Remove Programs
* Go to the Windows Setup tab
* Select Accessories, and press the "Details..." button
* Look for Quick View, check it, and insert your Windows
9X CD-ROM when prompted.
* In exporer, right mouse on a.exe and select Quick View.
Notice that the executable requires cygwin1.dll. Do the
same for hello.exe.
* Recompile the first file, "hello.c", as follows:
gcc -mno-cygwin hello.c
* Now use Quick View and notice that the executable requires
crtdll.dll and kernel32.dll, but not cygwin1.dll.
* Now recompile the second file, "hello.cpp", the same way:
g++ -mno-cygwin -o hello hello.cpp
Disaster! A huge number of error messages.
To solve it, go out and get
ftp://ftp.xraylith.wisc.edu/pub/khan/gnu-win32/cygwin/egcs-1.1.2/egcs-1.1.2-
mingw-extra.tar.gz
(that should be all one line -- I hate Exchange!)
and extract it to:
C:\usr\mingw32
In this tarball, libm.a seems to be a soft link which got
copied by mistake into the distribution. Either extract
everything but libm.a or extract everything and delete libm.a
from C:\usr\mingw32\lib
Now try again, using the following command line arguments:
g++ -mno-cygwin -I/usr/mingw32/include -L/usr/mingw32/lib
hello.cpp
This time it should link and run correctly.
The whole issue is explained in:
http://www.xraylith.wisc.edu/~khan/software/gnu-win32/ -- click
on the article "-mno-cygwin"
* Now test gcc on a Windows program. Copy the following file
http://www.geocities.com/Tokyo/Towers/6162/win32/simple.txt
to the directory you're working in and rename it to simple.c
* Compile it as follows:
gcc -mwindows simple.c or
gcc -mno-cygwin -mwindows simple.c
* Execute it from your shell:
./a.exe
* If you compiled it -mno-cygwin, you can execute it from
Explorer by double-clicking on the icon. If you compiled it
without that option, you'll need to make a copy of cygwin1.dll
somewhere in your executable path before you can execute it
by double-clicking on the icon.
* A somewhat more complex Win32 program is at
http://www.geocities.com/Tokyo/Towers/6162/win32/test.txt
Rename it to test.c and compile as you did simple.c
Extracting hypermail:
* Save the tarball in a local folder. If you're using Netscape,
don't forget to make the dot that Netscape turned into an
underscore a dot again. That is, the file should be named
hypermail-2b25.tar.gz, not hypermail-2b25_tar.gz.
* Using WinZip, open the file and extract it into some other
folder. Make sure that "Use folder names" is checked on
the Extract menu.
* If you don't have WinZip, you'll need to copy the tarball
to a Unix machine on which you have a shell account, extract
the files, and copy them all back to your Windows machine.
Configuring and compiling hypermail:
* In bash, go to the directory where you extracted the files
in the tarball. e.g.,
cd //c/aawork/hypermail-2b25
* In bash, type
sh ./configure
This should give you a lot of messages and complete
successfully. If you've used configure on a Unix system,
the only surprise is how familiar the outputs look.
* In bash, type
make
This should also complete successfully, with one warning
message:
mail.c: In function `lookupnumaddr':
mail.c:40: warning: return discards `const' from
pointer target type
* Use a plain text editor to edit the makefiles. I suppose
there is a good way to do this in the configuration, but
it's so easy to do by hand, why not?
Edit src/Makefile
Look for a line in the file that calls install (not the
command, but the program). Change "hypermail" to
"hypermail.exe" on that line. In my src/Makefile
it's line 107.
Edit archive/Makefile
There are several lines in that file to change: the first
is "hmail" and the last is "na-rdmsg". In each case, add
".exe" to the name in the line that calls install. In my
archive/Makefile, these statements start at line 99 and
there's 10 of them.
* In bash, type
mkdir /usr/local /usr/local/bin /usr/local/man
mkdir /usr/local/man/man1 /usr/local/man/man4
This creates the directories your programs and documents
will reside in. Yes, I know install is supposed to do
a mkdir if it can't find those files, but that seems to
be broken.
* In bash, type
make install
If you've changed all the makefiles correctly, you should
see the following directory contents:
BASH.EXE-2.02$ ls -F /usr/local/bin
hmail.exe* na-mail.exe* nu-rdmsg.exe*
hrdmsg.exe* na-rdmsg.exe* wu-rdmsg.exe*
hypermail.exe* nu-mail.exe* wuftpdmail.exe*
BASH.EXE-2.02$ ls -F /htdocs
customizing.html hr.yellow.gif hypermail.html
hmrc.html hypermail.gif
BASH.EXE-2.02$ ls -RF /usr/local/man
man1/ man4/
/usr/local/man/man1:
hypermail.1
/usr/local/man/man4:
hmrc.4
Testing Hypermail:
* Create an /etc/passwd file that has one entry:
yourlogin:x:500:10:Your Name://c/Bob:/usr/bin/bash
Note that I define HOME in C:\autoexec.bat:
SET HOME=C:\Bob
so vim doesn't end up writing its scratch files everywhere.
It *appears* that the routine in file.c uses this entry to
find $HOME and that no other routine finds $HOME any other
way. Still, it probably isn't a bad idea to go ahead and
define HOME in C:\autoexec.bat and reboot.
In the line above, the user number "500" is important. The
group number "10" probably isn't (I don't have an /etc/group),
and (obviously) //c/Bob isn't important unless your name is Bob.
Just make sure the directory exists.
Incidentally, creating /etc/passwd was the *only* thing I had
to do to get hypermail to run. Before I did this, it crashed
on line 231 of file.c
sprintf(tmppath, "%s%s", pp->pw_dir, path + 1);
because pp was null.
CODE GURUS: You might want to think about checking the return of
getpwuid() and getting $HOME out of the environment if it's NULL.
That would avoid a crash for Win32 users who don't bother to set up
/etc/passwd
* Copy the example .hmrc file to your home directory:
cp hmrc.example ~/.hmrc
* Get an example mbox file. If you have Netscape Navigator,
it turns out that in C:\Program Files\Netscape\Users\you\Mail
all the files that don't have extensions are mbox-style files.
In the example below, I used "Sent" as a simple test case.
* Copy the mbox file to some test directory.
* Go to that directory and create a directory (say, call it foo).
cd test
mkdir foo
* From bash, execute hypermail to build your archive.
/usr/local/bin/hypermail -m Sent -d foo
* Admire the results.
(speaking of admire, included mail files seem to lack the
<PRE> and look awful!)
(I also noticed that a JPEG file that was attached to one of
the
messages didn't seem to translate properly, though the JPEG
header looked good.) If nobody else runs into this, forget
it
and chalk it up to a bad file. Otherwise, you might want
to
look at Jorgen Hagg's base64 routine, which I got
permission
to use in foxynews:
http://home.hiwaay.net/~crispen/src/#foxynews
I'm sure he'd give you permission to use it in hypermail.
Now if you want to know about receiving mail on Win9x/NT and
using hypermail interactively, I've run out of things I know
how to do. Somebody else want to pick up the ball?
Bob Crispen
bob.crispen@boeing.com