How to apply the NetHack 1.4f patch

The patch to convert NetHack 1.3d to NetHack 1.4f is difficult to apply correctly. Some parts of it seem to be interpreted incorrectly by a modern Unix shell, and the patch for invent.c seems to be created from something different from its counterpart in 1.3d. This article will describe how to correct these problems and apply the NetHack 1.4f patch.

Note to users of Microsoft Windows

The original archives are distributed in a format called a "shell archive" and are most easily unpacked in a Unix environment. Users of Windows can install MinGW or Cygwin to obtain the necessary tools. DJGPP may also work.

The necessary files

Start with a copy of NetHack 1.3d. The Harlow and UUNet distributions are the same, so either may be used.

The patch was originally available at UUNet, but that site has disappeared. Fortunately, UUNet is mirrored at ftp.sunet.se. Download the patch here

The file has the extension ".gz". This means that it has been compressed with gzip. It can be unzipped with "gunzip".

Extracting the patch files

The file patch1 is a shell archive -- a Bourne shell script that, when run, creates the archived files in the current directory. However, it also has Usenet headers and a brief explanation at the top. You'll need to delete all lines before the one that reads "#! /bin/sh".

Run the resulting file as a Bourne shell script, say, as "sh patch1". This will produce the following files:

• README.NEW
• Fixes.1.4
• Manifest.upd
• Makefile.tcc
• Maketcc.ini
• update.patches

Editing the patch for a modern shell

The file update.patches is itself a Bourne shell script that invokes ed to modify the files that have changed in 1.4f. Specifically, it contains a series of shell commands such as these:

cp apply.c apply.c.orig
ed apply.c <<ED_EOF
187d
183,185d
1c
/*	SCCS Id: @(#)apply.c	1.4	87/08/08
.
w
q
ED_EOF

It is not quite compatible with GNU bash as provided, because the text that is passed to ed (between the <<ED_EOF and ED_EOF markers) may contain characters that are meaningful to the shell. We need to edit update.patches to tell bash not to interpret these characters.

Bash will not interpret these characters if the starting marker is quoted in any way. (We must not alter the ending marker.) So it is sufficient to use the replace function of your favorite text editor, and replace all instances of <<ED_EOF with <<'ED_EOF'. For example, in vi, use ":1,$s/<<ED_EOF/<<'ED_EOF'/".

The above block then reads:

cp apply.c apply.c.orig
ed apply.c <<'ED_EOF'
187d
183,185d
1c
/*	SCCS Id: @(#)apply.c	1.4	87/08/08
.
w
q
ED_EOF

and other such blocks are similarly affected.

Editing invent.c to match the patch

Now it is necessary to edit invent.c so that it matches the provided patch. The patch was generated from an altered invent.c that had six extra lines that are not present in the distributed version of 1.3d. Go to line 778; it and the surrounding lines read:

		doinv (stuff);

	return(0);
}

/* look at what is here */
dolook() {

Line 778 is highlighted in green.

You need to insert exactly six lines above this line. It does not matter what the lines contain, as the patch will blindly delete them, but there must be exactly six lines and they must be placed above that comment. You might see something like:

		doinv (stuff);

	return(0);
}

1
2
3
4
5
6
/* look at what is here */
dolook() {

Applying the patch

Once update.patches and invent.c are both edited in this way, you can now apply the patch with "sh update.patches".

When you are satisfied that the patch is applied correctly (you may do well to do "diff invent.c.orig invent.c" in particular), then you can go ahead and delete the files "*.orig", update.patches, and patch1. You now have a NetHack 1.4f that matches the likely intentions of the developers.