Sender: rich AT phekda DOT freeserve DOT co DOT uk
Message-ID: <3AA2451A.24DE1704@phekda.freeserve.co.uk>
Date: Sun, 04 Mar 2001 13:37:30 +0000
From: Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>
X-Mailer: Mozilla 4.51 [en] (X11; I; Linux 2.2.17 i586)
X-Accept-Language: de,fr
MIME-Version: 1.0
To: DJGPP workers <djgpp-workers AT delorie DOT com>
Subject: Fileutils 4.0 port and ginstall
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com

Hello.

Eli, a while back you told me of a bug in ginstall from Fileutils. Here is
what you said:

"There is a problem with `install': if you say something like
"install foo /bin/foo", and `foo' already has a stub loader 
prepended to it, you get `foo' installed in the target directory
instead of `foo.exe'.  I think `install' should be changed to 
make sure that executable files in the target directory have an 
executable extension (can be .com or .bat, in addition to .exe), 
since otherwise stock DOS shells will refuse to run them."

I've fixed the problem with stubbed COFF exectuables with no file
extension, i.e.:

    ginstall a b

now creates b.exe not b, if a is a stubbed COFF exe. It also now handles
non-COFF executables in the same way, e.g.

    ginstall c:/windows/command.com b

creates b.exe. I haven't yet written code to distinguish between .exe and
.com formats.

I am unsure how to decide that a file is a DOS batch file. The patches to
ginstall for DJGPP currently modify the target filename, depending on the
source file's file format. Here's some pseudo-code:

if (_check_v2_prog returns valid data) {
  if (unstubbed COFF) {
    stubify
    append .exe to target filename
  } else if (stubbed COFF) {
    append .exe to target filename
} else {
  if (file header contains "MZ") {
    /* FIXME: Distinguish between .exe and .com */
    append .exe to target filename
  }
}

Target filenames for Unix scripts (.sh, .pl, ones using hash-bang
notation, etc.) will be passed straight through, which seems like correct
behaviour.

Can I assume that a non-MZ file is a batch file in the outer else clause?
That is, if a file does not have valid _check_v2_prog() data and is not an
MZ executable, must it be a batch file? It seems to me that this is a bad
assumption - which batch file extension should be used? .bat, .cmd, etc.?
I don't think batch can be treated correctly automatically and it should
be down to the user to sort it out. How many people use ginstall to
install batch files anyway?

Should _check_v2_prog() succeed on ordinary DOS executables such as
command.com? _check_v2_prog() from 2.03 seems to fail for me.

Out of interest, has anyone been testing alpha 2 of Fileutils that I
prepared in October last year? See:

   http://www.phekda.freeserve.co.uk/richdawe/djgpp/fileutils

I've been using it for several months and it seems to work OK. Once I've
finished working on ginstall, I'll test dd, SFN builds and then I think
Fileutils 4.0 port is ready for a beta. Please note that I'm targetting
building with DJGPP 2.03. I think using DJGPP CVS's new goodies will be
quite some work.

Bye, Rich =]

-- 
Richard Dawe <richdawe AT bigfoot DOT com> http://www.bigfoot.com/~richdawe/ 

"The soul is the mirror of an indestructible universe."
--- Gottfried W. Leibniz