Porting Applications in C

Services for UNIX 3.0 Technical Note

Abstract

This paper is in addition to the information presented in the Interix Software Development Kit User's Guide. After a brief discussion of methods used to make code portable and how they affect the porting process, this paper reviews the differences and similarities between Interix and traditional open systems, and discusses some means of dealing with those differences. It ends with a list of APIs supported in Interix 2.2.

On This Page

Porting Applications in C
Examples
Summary of Interfaces
For More Information

Porting Applications in C

This paper is in addition to the information presented in the Windows Services for Unix 3.0 Software Development Kit User's Guide. After a brief discussion of methods used to make code portable and how they affect the porting process, this note reviews the differences and similarities between Interix and traditional open systems, and discusses some means of dealing with those differences. This document provides a number of examples of porting Unix application and ends with a list of APIs supported in Interix 2.2.

The topics of configuration scripts, porting daemons, and X Window System code deserve separate discussions and will be given their own technical notes.

The usual method for porting an application to Interix is to move the source to an Interix system and then recompile. If a Makefile exists, porting may be as simple as typing make.

Although Interix is a POSIX.1-conforming system, it takes different approaches in some areas than traditional systems, and there are differences outside the areas defined by the standard. Some of these are discussed in the User's Guide.

Code Portability Strategies

There are several basic approaches to porting code. Most applications use more than one of these.

• Write a custom library of functions for each platform. The application always calls the private version of the function, which in the compilation stage is linked to a platform-specific library. This is a great deal of work, though it may be the appropriate method for a large body of source.

• Define an extensive set of platform-specific macros that expand to the correct functions and names.

• Use #ifdef statements to isolate sections of code based on the platform. For example, Interix supports POSIX.1, ANSI/ISO C and many interfaces from both historical BSD and SVR4 systems. Older code written using #ifdefs may make assumptions about the platform that aren't valid. For example, code built around #ifdef BSD will usually try to include <sgtty.h> rather than <termios.h>. By labeling blocks of code with the platform name, you're often trying to hit a moving target; for example, BSD4.4 has some different APIs than BSD4.3, but they're both BSD.

If you're porting an application that provides configurations for different platforms, the closest configuration to Interix is another POSIX system. Depending upon which POSIX features are supported, you may find it useful to try the code in an #ifdef BSDI386. If there's a lot of terminal code and no POSIX support, you may find it easier to start with the SVR4 configuration.

• Use #ifdef statements where code compilation is dependent upon features. This is the most common approach.

The advantage of this method is that sets of feature test macros can be gathered into a configuration block; new platforms can be defined with a single block of code in one file. Another advantage is to create a record of changes that can be refered to in other projects.

When porting a package for redistribution, remember that Makefiles may also need modifications.

The Interix Environment

Interix is a POSIX.1 system with extensions taken from both BSD and System V. Interix is "more like" a BSD system in some ways, but the POSIX terminal handling is very much like System V. If the application being ported makes extensive use of terminal interfaces or shared memory IPC, you may want to start with the System V version of the code.

When porting a new program, examine the application for features that aren't yet supported in the Interix subsystem. Depending upon the feature, you can use #ifdef statements to isolate them, or you could use a feature-based macro such as HAVE_MEMPCPY for to avoid an unsupported function (in this case mempcpy).

At the end of this paper is a list of the APIs supported by Service for Unix/Interix 3.0. See the reference pages and the Interix SDK User's Guide for more details.

SDK User's Guide

This note extends information found in the Interix SDK User's Guide. Please also refer there. The SDK User's Guide discusses the following:

• A system overview of Interix

• Writing POSIX-standard code, including

  • compile-time macros and manifest constants

  • getting system information (<limits.h>, confstr(), fpathconf(), pathconf(), sysconf())

  • POSIX APIs that supplant historical APIs: advisory file locking with fcntl() rather than flock() or lockf(); getcwd() rather than getwd(); mkfifo() rather than mknod(); utime() rather than utimes(); regular expressions with the functions declared in <regex.h>; <stdarg.h> rather than <varargs.h>

  • POSIX terminal I/O with the termios interface

  • POSIX signals, including sigaction()

• Extensions to POSIX.1 included in Interix:

  • Header files aligned with the Single UNIX Specification

  • Memory-mapped files, System V IPC mechanisms

  • Sockets

  • Pseudoterminals

  • Controlling terminals

  • dynamic linking

• Interix-specific issues regarding user authentication and absolute pathnames on the Windows NT® operating system.

Unsupported Features

There are several features which Interix does not yet support. They may be supported in future releases. Most of these are mentioned in the SDK User's Guide, but are repeated here.

• Many ioctl() request are not supported, particularly the disk label and magnetic tape I/O requests.

Sample Configuration Blocks

As an example of a configuration block, here is the one added to apache, the Internet http server:

#elif defined(__INTERIX)
#define     USE_FCNTL_SERIALIZED_ACCEPT
#undef     HAS_GMTOFF
#define     NO_SETSID
#define     JMP_BUF sigjmp_buf
#include <sys/time.h>
#define     getwd(d)      getcwd(d,MAX_STRING_LEN)

Dealing with Missing Header Files

If you can't build your application because Interix is missing header files, do not blindly copy header files from another platform. You need to determine why the header file is being used in the application and whether the same information may be in a different header file on Interix. This is the essence of porting.

If the information is supported on Interix in a different header file, you need to #include that header file. If the information is not supported on Interix, you need to figure out a different way to achieve the same ends. In software that has been ported to many platforms, you can usually achieve the same result by changing configuration options. In custom software, you may need to write a new supporting library function.

Reading the Interix documentation and reference pages will be useful.

Handling Filenames

Earlier versions of the Interix environment did not have a single rooted filesystem. On Windows NT, or Windows2000, multiple filesystems are accessible through drive letters. Access to these are available using the syntax "//L" where L is an uppercase letter. This affected configure scripts, a number of popular software packages, Makefiles, and widgets such as file selection boxes in X applications, none of which make allowances for the special "//" prefix.

With the 3.0 version of Services for Unix, this is no longer an issue, Interix has a single rooted filesystem where "/" is the directory where Services for Unix was installed. This change makes porting much, much easier as we will see in the example section below.

Reserved Filenames (only with c89 and cc)

The compiler interfaces (c89 and cc) pass the list of files to the Microsoft Visual C/C++® compiler. Files to be compiled with c89 and cc must have names that are acceptable to the Microsoft compiler in the Win32® environment. This is not a problem when compiling with gcc; the gcc compiler operates entirely in the Interix environment, where the filenames are not reserved.

The following names are reserved under Win32, regardless of case or file extension.

AUX
LCOMn (where n is a digit)
CON
LPTn (where n is a digit)
NUL
PRN

For example, files named AUX.C, aux.c, aux.h, AUx.h or any other permutation will not compile. This is because the compiler is a Win32 program.

If you have files by these names (aux.c is particularly common), you must rename them.

About Configure Scripts

The ability of a configure script to work out the characteristics of Interix depends almost entirely on how clever the script is and what assumptions it makes. It's not possible to give a recipe for "fixing" configure scripts because each script is different—the heuristics used by configure scripts vary with the version.

Suggestions:

• On systems such as SVR4 and BSD, the compiler is silent if there is no problem. The Microsoft Visual C/C++ compiler always writes the name of the compiled file to standard output. Because of this, configure scripts may "decide" that APIs aren't present because there was output from the compiler. The version of cc/c89 included in release 2.1 and later catches this output. This is not a problem for gcc.

• Interix libraries are distributed differently on the system. For example, the ELM configure script searches through libc.a — but many of the interfaces are found in libpsxdll.a. Once the configure script is finished, it has not found many of the APIs that are actually provided. However, searching through libc.so will find these apis because the architecture of libc.so includes functions in libpsxdll.a

Depending on the script, it may be faster to manually change the derived configuration description rather than changing the logic of the configuration script. This is not a portable or reusable solution, but it is sometimes needed. The first apache port for Interix was done this way.

Configure scripts are more thoroughly discussed in the Configure Scripts paper available at https://www.microsoft.com/windows2000/interix.

Interface Conversions

This section describes some of the conversions which may be required in porting code. Most of these are required to run on any POSIX.1 system. The references listed in the Interix SDK User's Guide are a good source of information on these conversions.

ioctl() Calls

The ioctl() interface has many uses. The POSIX.1 committee did not standardize the ioctl() interface because the last argument ca not be type-checked (its type depends upon the request). Instead, the committee broke out certain pieces of functionality and assigned them to other interfaces.

The Interix API set contains only a few ioctl() operations, including window re-sizing.

The ioctl() interface historically was used to handle disk labels, file I/O, magnetic tape I/O, socket I/O, and terminal I/O. The disk label and magnetic tape I/O requests are not supported in the Interix environment.

File Control

The only ioctl() requests defined for file control are FIONREAD, to get the number of bytes available to read, and FIONBIO, to set and unset non-blocking I/O.

On Interix, the FIONREAD request is equivalent to the fcntl() F_GETNREAD request.

The FIOCLEX and FIONCLEX requests (usually found in <filio.h>) are not provided. They can be replaced with the fcntl() FD_CLOEXEC request.

For example:

#ifndef __INTERIX
(void) ioctl(fd, FIOCLEX, NULL)
#else
(void) fcntl(fd, F_SETFL, fcntl(fd, F_GETFD) |
     FD_CLOEXEC));
#endif

Sockets

The only ioctl() request defined for sockets is SIOCATMARK. Other socket control requests are handled through fcntl() or through functions such as setsockopt().

Terminal Control

In POSIX.1, many of the terminal control calls that used to be handled with ioctl() requests were given separate interfaces (as listed on page 27 of the Interix SDK User's Guide).

For example, the ioctl() requests TIOCGATC, TIOCGATC, TIOCGLTC, and TCGETA all translate to the tcgetattr() call, filling in the termios structure. The requests TIOCSATC, TIOCSATC, TIOCSLTC, and TCSETA translate to the tcsetattr() call.

Allocating a Controlling Terminal

On SVR4, the controlling terminal is allocated when the session leader opens the first terminal device that is not already associated with a session and the open() call does not specify O_NOCTTY. On BSD systems 4.3+, processes allocate the controlling terminal with an ioctl() call, using the request TIOCSCTTY. Interix follows the SVR4 practice.

setpgrp()

Both BSD and System V have a setpgrp() interface, but they have different semantics and behaviors. The System V call changes the caller's process group ID to its own process ID. The BSD call detaches a process from its process group but doesn't change the controlling terminal.

The System V call takes no arguments. It is properly replaced by the POSIX setsid() call. When a process calls setsid() successfully, it creates a new session, which in turn contains a new process group, which contains one process — the calling process. The calling process is now the session and process group leader. The process also disposes of its controlling terminal. (In fact, the setsid() call is used to dispose of a controlling terminal.)

The BSD call takes two arguments, a process ID and a process group ID. It is replaced by the POSIX setpgid() call, which has identical semantics.

Debugging the Ported Code

There are a couple of differences in behavior between Interix and traditional systems such as Solaris that might show up when debugging your ported code.

Segmentation Faults

Some users porting code notice that their application memory faults more often than it did on the original platform. Experience has shown that this usually exposes a defect in the code, not in the Interix subsystem. (A subsystem segmentation fault is a different issue.)

A common cause is trying to de-reference an uninitialized pointer. An uninitialized pointer contains some random bit pattern. An attempt to read or write at the memory address indicated by that bit pattern will succeed or fail, depending on whether the user is allowed access to that address. If it fails, it fails with a segmentation fault. Even if it succeeds, it may cause problems for other applications which are using that chunk of memory.

On Windows NT, Windows 2000 and Windows XP, most of the four-gigabyte address space is off-limits to the user program. Any attempt to read or write in that space causes an application segmentation fault.

For example, this code is incorrect and will behave unexpectedly on a traditional system. It will segmentation-fault on an Interix system:

#include <stdio.h>
struct foo
{
        int one;
        int two;
};
main()
{
     int i1, i2, i3, i4;
        struct foo *t1;
        t1->one = 7;
        printf ("%d\n", t1->one);
}

There are two problems here: first, foo is defined but no actual structure is declared; second, the variable t1 is not initialized. The attempt to de-reference t1, assigning the member one to the value 7, actually points to some random memory location offset to give the access to the structure member. The following two fixes correct the problem:

#include <stdio.h>
struct foo
{
        int one;
} bar;
main()
{
     int i1, i2, i3, i4;
        struct foo *t1;
        t1=&bar;
        t1->one = 7;
        printf ("%d\n", t1->one);
}

A structure (bar) is now declared, and t1 is initialized to its address.

Porting code is often an exercise in discovering assumptions.

This should include some hint as to addresses of shared objects though they should be qualified with "these addresses may change at any time." For example, as of this writing:

The addresses of the other .so's are:

libdb.so	= 0x77c00000	# to 77c30000
libm.so	= 0x77c30000	# to 77c80000
librpc.so	= 0x77c80000	# to 77cb0000
libform.so	= 0x77cb0000	# to 77d00000
libcurses	= 0x77d00000	# to 77d80000
libc.so	= 0x77d80000	# to 77ea0000

SDK User's Guide (and probably plain User's Guide) should mention core files.

SDK User's Guide should mention problems with literal strings in read-only data

By default, gcc generates code for literal strings inline (such as in foo("hithere")) in read-only data. MSVC does not. (They're writeable.)

Such a situation occurred in ps where it was writing into such a string (the old "temporary '\0'" trick), and of course it fell over when compiled with gcc. The command line option -fwritable-strings will convince gcc otherwise. The primary reason for this note, however, is to get it fresh in peoples mind that as we use gcc more if nothing else, to create PIC libc.a) we'll have to be careful about that. (An easy way out of this is to declare the string as a static char[] initialized to that value.)

The GDB Debugger

The GDB debugger is available as part of the Interix Software Development Kit. It is possible

Examples

Introduction

In this section outlines a number of porting examples to show some of the issues that can arise in porting Unix applications to Interix.

zlib-1.1.4

zlib 1.1.4 is a general purpose data compression library. All the code is thread safe. The data format used by the zlib library is described by RFCs (Request for Comments) 1950 to 1952 in the files https://www.ietf.org/rfc/rfc1950.txt (zlib format), rfc1951.txt (deflate format) and rfc1952.txt (gzip format). These documents are also available in other formats from

ftp://ftp.uu.net/graphics/png/documents/zlib/zdoc-index.html

unarchive the package

$ gunzip < zlib-1.1.4.tgz | tar xf -
$ ls
ChangeLog       amiga           infblock.c      infutil.c       zconf.h
FAQ             compress.c      infblock.h      infutil.h       zlib.3
INDEX           configure       infcodes.c      maketree.c      zlib.h
Make_vms.com    contrib         infcodes.h      minigzip.c      zlib.html
Makefile        crc32.c         inffast.c       msdos           zutil.c
Makefile.in     deflate.c       inffast.h       nt              zutil.h
Makefile.riscos deflate.h       inffixed.h      os2
README          descrip.mms     inflate.c       trees.c
adler32.c       example.c       inftrees.c      trees.h
algorithm.txt   gzio.c          inftrees.h      uncompr.c

This archive has a configure script, but it is not the usual GNU configure script, so we need to execute this as follows:

$ ./configure --prefix=/usr/local
Checking for gcc...
Building static library libz.a version 1.1.4 with gcc.
Checking for unistd.h... Yes.
Checking for errno.h...  Yes.
Checking for mmap support... Yes.
$

Now we can try a make

$ make
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c example.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c adler32.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c compress.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c crc32.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c gzio.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c uncompr.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c deflate.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c trees.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c zutil.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c inflate.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c infblock.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c inftrees.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c infcodes.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c infutil.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c inffast.c
ar rc libz.a adler32.o compress.o crc32.o gzio.o uncompr.o deflate.o
trees.o  zutil.o inflate.o infblock.o inftrees.o infcodes.o infutil.o
inffast.o
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -o example example.o -L. -lz
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -c minigzip.c
gcc -O3 -DHAVE_UNISTD_H -DUSE_MMAP -o minigzip minigzip.o -L. -l
$

just to be sure, we will "make test"

$ make test
hello world
uncompress(): hello, hello!
gzread(): hello, hello!
gzgets() after gzseek: hello!
inflate(): hello, hello!
large_inflate(): OK
after inflateSync(): hello, hello!
inflate with dictionary: hello, hello!
_** zlib test OK **_
$

everything looks ok, now we'll make install

$ make install
cp zlib.h zconf.h /usr/local/include
chmod 644 /usr/local/include/zlib.h /usr/local/include/zconf.h
cp libz.a /usr/local/lib
cd /usr/local/lib; chmod 755 libz.a
$

all done, ready for use!

gdbm 1.8.0

This library will be used by PERL in a later example. GNU dbm is a set of database routines that use extendible hashing and works similar to the standard UNIX dbm routines.

$ gunzip < gdbm-

bzip2 1.0.1

"bzip2" compresses files using the Burrows-Wheeler block-sorting text compression algorithm, and Huffman coding. Compression is generally considerably better than that achieved by more conventional LZ77/LZ78-based compressors, and approaches the performance of the PPM family of statistical compressors. It provides a library, two headers and a set of binaries. The library is used in other packages.

As usual, we will unarchive the package:

$ gunzip < bzip2-1.0.1.tgz | tar xf -
$ cd bzip2-1.0.1
$ ls
CHANGES	bzlib_private.h	manual_4.html
LICENSE	compress.c	manual_toc.html
Makefile	crctable.c	randtable.c
Makefile-libbz2_so	decompress.c	sample1.bz2
README	dlltest.c	sample1.ref
README.COMPILATION.PROBLEMS	dlltest.dsp	sample2.bz2
Y2K_INFO	huffman.c	sample2.ref
blocksort.c	libbz2.def	sample3.bz2
bzip2.1	libbz2.dsp	sample3.ref
bzip2.1.preformatted	makefile.msc	spewG.c
bzip2.c	manual.ps	unzcrash.c
bzip2.txt	manual.texi	words0
bzip2recover.c	manual_1.html	words1
bzlib.c	manual_2.html	words2
bzlib.h	manual_3.html	words3

As we can see, there is no "configure" script, however there is a Makefile, so we will just attempt a make.

$ make
If compilation produces errors, or a large number of warnings,please read
README.COMPILATION.PROBLEMS -- you might be able to adjust the flags in
this Makefile to improve matters.
gcc -Wall -Winline -O2 -fomit-frame-pointer -fno-strength-reduce …
Doing 6 tests (3 compress, 3 uncompress) ...
If there's a problem, things might stop at this point.
./bzip2 -1  < sample1.ref > sample1.rb2
./bzip2 -2  < sample2.ref > sample2.rb2
…
cmp sample2.tst sample2.ref
cmp sample3.tst sample3.ref
If you got this far and the "cmp"s didn't complain, it looks
like you're in business.
To install in /usr/bin, /usr/lib, /usr/man and /usr/include, type
   make install
To install somewhere else, eg, /xxx/yyy/{bin,lib,man,include}, type
   make install PREFIX=/xxx/yyy
If you are (justifiably) paranoid and want to see what 'make install'
is going to do, you can first do
   make -n install                      or
   make -n install PREFIX=/xxx/yyy      respectively.
The -n instructs make to show the commands it would execute, but
not actually execute them.
Instructions for use are in the preformatted manual page, in the file
bzip2.txt.  For more detailed documentation, read the full manual.
It is available in Postscript form (manual.ps) and HTML form
(manual_toc.html).
You can also do "bzip2 --help" to see some helpful information.
"bzip2 -L" displays the software license.
$

That seemed to compile ok, and as we can see by the output, it works correctly too. So we will follow the installation instructions and run "make install". Since we've decided to install everything in /usr/local, we'll set the PREFIX to that.

$ make install PREFIX=/usr/local
if ( test ! -d /usr/local/bin ) ; then mkdir /usr/local/bin ; fi
if ( test ! -d /usr/local/lib ) ; then mkdir /usr/local/lib ; fi
if ( test ! -d /usr/local/man ) ; then mkdir /usr/local/man ; fi
if ( test ! -d /usr/local/man/man1 ) ; then mkdir /usr/local/man/man1 ;
fi
if ( test ! -d /usr/local/include ) ; then mkdir /usr/local/include ; fi
cp -f bzip2 /usr/local/bin/bzip2
cp -f bzip2 /usr/local/bin/bunzip2
cp -f bzip2 /usr/local/bin/bzcat
cp -f bzip2recover /usr/local/bin/bzip2recover
chmod a+x /usr/local/bin/bzip2
chmod a+x /usr/local/bin/bunzip2
chmod a+x /usr/local/bin/bzcat
chmod a+x /usr/local/bin/bzip2recover
cp -f bzip2.1 /usr/local/man/man1
chmod a+r /usr/local/man/man1/bzip2.1
cp -f bzlib.h /usr/local/include
chmod a+r /usr/local/include/bzlib.h
cp -f libbz2.a /usr/local/lib
chmod a+r /usr/local/lib/libbz2.a
$

This package is complete and installed!

gettext 0.10.40

The gettext package is interesting for authors or maintainers of other packages or programs which they want to see internationalized. As one step the handling of messages in different languages should be implemented. For this task GNU gettext provides the needed tools and library functions.

This package configuration is driven by the configure script, like many GNU packages, running configure with the proper arguments ensures a successful configuration.

$ CFLAGS="-D_ALL_SOURCE" ./configure \
--prefix=/usr/local \
--host=intel-pclocal-interix
checking for a BSD compatible install... /bin/install -c
checking whether build environment is sane... yes
checking whether make sets ${MAKE}... yes
checking for working aclocal... missing
checking for working autoconf... missing
checking for working automake... missing
…
updating cache ./config.cache
creating ./config.status
creating Makefile
creating lib/Makefile
creating intl/Makefile
creating src/Makefile
creating po/Makefile.in
creating doc/Makefile
creating man/Makefile
creating tests/Makefile
creating m4/Makefile
creating misc/Makefile
creating misc/gettextize
creating config.h
creating po/POTFILES
creating po/Makefile
$

This looks promising so far, let's attempt a make.

$ make
make  all-recursive
Making all in doc
cd .  && env LANG= LANGUAGE= /bin/sh /dev/fs/C/src/gettext-0.11.1/missing
--run makeinfo    `echo gettext.texi | sed 's,.*/,,'`
Making all in intl
/bin/sh ../libtool --mode=compile gcc -c -
DLOCALEDIR=\"/usr/local/share/locale\"
 -DLOCALE_ALIAS_PATH=\"/usr/local/share/locale\"  -
DLIBDIR=\"/usr/local/lib\" -DIN_LIBINTL -DHAVE_CONFIG_H -I.. -I. -
I../intl -D_ALL_SOURCE -I/usr/local/include -g -D_ALL_SOURCE -
I/usr/local/include  intl-compat.c
gcc -c -DLOCALEDIR=\"/usr/local/share/locale\" -
DLOCALE_ALIAS_PATH=\"/usr/local/share/locale\" -
DLIBDIR=\"/usr/local/lib\" -DIN_LIBINTL -DHAVE_CONFIG_H -I.. -I. -
I../intl -D_ALL_SOURCE -I/usr/local/include -g -D_ALL_SOURCE -
I/usr/local/include intl-compat.c -o intl-compat.o
echo timestamp > intl-compat.lo
…
gcc -g -D_ALL_SOURCE -I/usr/local/include -o tstngettext tstngettext.o
setlocale.o  -L/usr/local/lib ../lib/.libs/libgettextlib.a
/dev/fs/C/src/gettext-0.11.1/intl/.libs/libintl.a ../intl/.libs/libintl.a
-lc
gcc -DLOCALEDIR=\"\" -DHAVE_CONFIG_H -I. -I. -I.. -I.. -I../lib -I../lib
-I../intl   -D_ALL_SOURCE -I/usr/local/include  -g -D_ALL_SOURCE -
I/usr/local/include -c plural-1-prg.c
/bin/sh ../libtool --mode=link gcc  -g -D_ALL_SOURCE -I/usr/local/include
-L/usr/local/lib -o cake  plural-1-prg.o setlocale.o
../lib/libgettextlib.la ../intl/libintl.la
gcc -g -D_ALL_SOURCE -I/usr/local/include -o cake plural-1-prg.o
setlocale.o  -L/usr/local/lib ../lib/.libs/libgettextlib.a
/dev/fs/C/src/gettext-0.11.1/intl/.libs/libintl.a ../intl/.libs/libintl.a
-lc
$

This looks like a successful build, let's attempt a "make check" to see if we have a good build.

$ make check
Making check in doc
…
make  check-TESTS
PASS: test-names.sh
==================
All 1 tests passed
==================
Making check in src
Making check in po
Making check in man
Making check in m4
Making check in projects
Making check in misc
WARNING: Warnings can be ignored. :-)
if test no != no; then  EMACS=no /bin/sh ./elisp-comp po-mode.el;  else :
; fi
Making check in tests
make  check-TESTS
PASS: gettext-1
…
PASS: msgcat-1
msgcat: Cannot convert from "ISO-8859-1" to "UTF-8". msgcat relies on
iconv(). This version was built without iconv().
FAIL: msgcat-2
msgcat: Cannot convert from "ISO-8859-1" to "UTF-8". msgcat relies on
iconv(). This version was built without iconv().
FAIL: msgcat-3
msgcat: Cannot convert from "ISO-8859-1" to "UTF-8". msgcat relies on
iconv(). This version was built without iconv().
FAIL: msgcat-4
PASS: msgcat-5
…
PASS: msgcomm-3
msgcomm: Cannot convert from "ASCII" to "ISO-8859-1". msgcomm relies on
iconv(). This version was built without iconv().
FAIL: msgcomm-4
msgcomm: Cannot convert from "ASCII" to "ISO-8859-1". msgcomm relies on
iconv(). This version was built without iconv().
FAIL: msgcomm-5
msgcomm: Cannot convert from "ASCII" to "ISO-8859-1". msgcomm relies on
iconv(). This version was built without iconv().
FAIL: msgcomm-6
msgcomm: Cannot convert from "ASCII" to "ISO-8859-1". msgcomm relies on
iconv().
 This version was built without iconv().
FAIL: msgcomm-7
PASS: msgcomm-8
…
PASS: msgcomm-15
msgcomm: Cannot convert from "ISO-8859-1" to "UTF-8". msgcomm relies on
iconv(). This version was built without iconv().
FAIL: msgcomm-16
PASS: msgcomm-17
…
PASS: msgcomm-23
msgconv: Cannot convert from "BIG5" to "UTF-8". msgconv relies on
iconv(). This version was built without iconv().
FAIL: msgconv-1
msgconv: Cannot convert from "UTF-8" to "BIG5". msgconv relies on
iconv(). This version was built without iconv().
FAIL: msgconv-2
PASS: msgconv-3
…
PASS: plural-2
Couldn't set locale.
SKIP: lang-c
Couldn't set locale.
SKIP: lang-c++
…
SKIP: lang-pascal
PASS: lang-ycp
msgfmt: Cannot convert from "ISO-8859-1" to "UTF-8". msgfmt relies on
iconv(). This version was built without iconv().
FAIL: lang-tcl
PASS: lang-po
…
PASS: rpath-2bbd
======================
11 of 159 tests failed
======================
*** Error code 1
Stop.
*** Error code 1
Stop.
*** Error code 1
Stop.
$

This looks less promising, and we need to investigate the failures. In this case, most of the failures are a result of Interix not providing the iconv API. The other issues are due to the fact that Interix currently only supports the C and POSIX locales (check the setlocale manpage), the lang-c, lang-c++ and lang-pascal attempt to set the locale to fr_FR which is not possible with Interix. This is found by examining the lang-* files for the tests that were SKIPPED. It is likely that future releases of Interix will support multiple locales. However, more than 90% of this package is operational. It may be that some issues may arise as we continue to use this package, but we can measure this as a qualified success.

xpm 3.4k

XPM (X PixMap) is a format for storing/retrieving X pixmaps to/from files. If we look at the file listing, we will notice that there is no configure script at all. This is a package built for the X-Windows system and it uses a different process for building. With X-Windows applications, the Imakefile is the "configure" counterpart. A different tool is used to construct the Makefiles, xmkmf (create a Makefile from an Imakefile). That tool is also part of the Interix utility set, so we should be able to use this to build this package.

$ ls -CF
CHANGES         FILES           README.AMIGA    cxpm/           namecvt*
COPYRIGHT       Imakefile       README.MSW      doc/            sxpm/
FAQ.html        Makefile.noX    README.html     lib/

There is a README.html, we can read this file to help us understand what the package does and whether there are any instructions on compiling the package. The instructions in the README.html state that we should use "xmkmf -a" to start if the platform supports it, we will try!

$ xmkmf -a
imake -DUseInstalled -I/usr/X11R5/lib/X11/config
make Makefiles
making Makefiles in ./lib...
rm -f lib/Makefile.bak
cd lib; imake -DUseInstalled -I/usr/X11R5/lib/X11/config  -DTOPDIR=../. -
DCURDIR=./lib;  make  UPPREFIX=../ NEWTOP=../ MAKEFILE_SUBDIR=lib
NEW_CURRENT_DIR=./lib
 Makefiles
making Makefiles in ./sxpm...
rm -f sxpm/Makefile.bak
cd sxpm; imake -DUseInstalled -I/usr/X11R5/lib/X11/config  -DTOPDIR=../.
-DCURDIR=./sxpm;  make  UPPREFIX=../ NEWTOP=../ MAKEFILE_SUBDIR=sxpm
NEW_CURRENT_DIR=./
sxpm Makefiles
making Makefiles in ./cxpm...
rm -f cxpm/Makefile.bak
cd cxpm; imake -DUseInstalled -I/usr/X11R5/lib/X11/config  -DTOPDIR=../.
-DCURDIR=./cxpm;  make  UPPREFIX=../ NEWTOP=../ MAKEFILE_SUBDIR=cxpm
NEW_CURRENT_DIR=./
cxpm Makefiles
make includes
including in ./lib...
including in ./sxpm...
including in ./cxpm...
make depend
depending in ./lib...
makedepend  -s "# DO NOT DELETE" -- -I.  -I/usr/X11R5/include  -Dopennt
-- data.c create.c misc.c rgb.c scan.c parse.c hashtab.c
CrBufFrI.c CrDatFrP.c CrPFrBuf.c RdFToI.c WrFFrI.c      CrBufFrP.c
CrIFrBuf.c CrPFrDat.c RdFToP
.c WrFFrP.c      CrDatFrI.c CrIFrDat.c RdFToDat.c WrFFrDat.c     Attrib.c
CrIFrP.c CrPFrI.c Image.c Info.c RdFToBuf.c WrFFrBuf.c
depending in ./sxpm...
makedepend  -s "# DO NOT DELETE" -- -I.././X11 -I..  -I/usr/X11R5/include
-Dopennt     -- sxpm.c
depending in ./cxpm...
makedepend  -s "# DO NOT DELETE" -- -I.././X11 -I..  -I/usr/X11R5/include
-Dopennt     -- cxpm.c

As we can see, the Interix xmkmf utility does support the "-a" flag, so we should now just be able to build the package with "make"

$ make
making all in ./lib...
rm -f data.o
…
making all in ./sxpm...
…
making all in ./cxpm...
…
gcc -fstrength-reduce -fpcc-struct-return -O -O  -static -L/usr/X11R5/lib
-o cxpm cxpm.o
$

Everything built ok, so now, we can install by "make install"

$ make install
installing in ./lib...
install -c -m 0644 libXpm.a /usr/X11R5/lib
+ install -c -m 0444 xpm.h /usr/X11R5/include/X11
install in ./lib done
installing in ./sxpm...
install -c   sxpm /usr/X11R5/bin
install in ./sxpm done
installing in ./cxpm...
install -c   cxpm /usr/X11R5/bin
install in ./cxpm done
$

Another package built and installed!

groff 1.17.2

Many UNIX packages deliver documentation in nroff format. GNU groff is an Open Source replacement for the AT&T nroff package. As usual, we unarchive the package, and run the configure command. If you are building many of the GNU packages, you may want to create a shell script that provides the parameters to the configuration file. This can be very useful for consistency and avoid typing mistakes. The following is an example:

$ cat /usr/local/bin/runconfig
set -x
CPPFLAGS="-D_ALL_SOURCE -I/usr/local/include" \
CXXFLAGS="-D_ALL_SOURCE -I/usr/local/include" \
CFLAGS="-D_ALL_SOURCE -I/usr/local/include" \
LDFLAGS="-L/usr/local/lib" \
./configure --prefix=/usr/local \
--host=intel-pclocal-interix $*

This script sets the compilation flags for C, C++ and CPP, it also includes the /usr/local directory for both include and library files, so the libraries we built above, will be found. Using this script to run the groff configure script we get a good configure and we should be able to make without difficulties.

$ runconfig
+ ./configure --prefix=/usr/local --host=intel-pclocal-interix
+ CPPFLAGS=-D_ALL_SOURCE -I/usr/local/include CXXFLAGS=-g -D_ALL_SOURCE -
I/usr/l
ocal/include CFLAGS=-g -D_ALL_SOURCE -I/usr/local/include LDFLAGS=-
L/usr/local/lib
configure: WARNING: If you wanted to set the --build type, don't use –
host.
    If a cross compiler is detected then cross compile mode will be used.
checking for intel-pclocal-interix-gcc... no
checking for gcc... gcc
checking for C compiler default output... a.out
checking whether the C compiler works... yes
…
checking for prefix of system macro packages...
checking which system macro packages should be made available...
configure: creating ./config.status
config.status: creating Makefile
config.status: creating src/xditview/Imakefile
$

This looks fine, so let's attempt a make

$ make
g++ -I. -I/dev/fs/C/src/groff-1.17.2/src/libs/libgroff  -
I/dev/fs/C/src/groff-1.17.2/src/include -I/dev/fs/C/src/groff-
1.17.2/src/include -DHAVE_STDLIB_H=1 -DHAVE_UNISTD_H=1 -DHAVE_DIRENT_H=1
-DHAVE_LIMITS_H=1 -DHAVE_SYS_DIR_H=1 -DHAVE_STRING_H=1 -DHAVE_STRINGS_H=1
-DHAVE_MATH_H=1 -DRET_TYPE_SRAND_IS_VOID=1 -DHAVE_CC_OSFCN_H=1 -
DHAVE_CC_LIMITS_H=1 -DRETSIGTYPE=void -DHAVE_STRUCT_EXCEPTION=1 -DHAVE
_STDLIB_H=1 -DHAVE_UNISTD_H=1 -DHAVE_GETPAGESIZE=1 -DHAVE_FMOD=1 -
DHAVE_STRTOL=1 -DHAVE_GETCWD=1 -DHAVE_STRERROR=1 -DHAVE_PUTENV=1 -
DHAVE_RENAME=1 -DHAVE_STRCASECMP=1 -DHAVE_STRNCASECMP=1 -DHAVE_STRSEP=1 -
DHAVE_STRDUP=1 -DHAVE_MKSTEMP=1 -DHAVE_MKSTEMP_PROTO=1 -g -D_ALL_SOURCE -
I/usr/local/include  -c assert.cc
g++ -I. -I/dev/fs/C/src/groff-1.17.2/src/libs/libgroff  -
I/dev/fs/C/src/groff-1.17.2/src/include -I/dev/fs/C/src/groff-
1.17.2/src/include -DHAVE_STDLIB_H=1 -DHAVE_UNISTD_H=1 -DHAVE_DIRENT_H=1
-DHAVE_LIMITS_H=1 -DHAVE_SYS_DIR_H=1 -DHAVE_STRING_H=1 -DHAVE_STRINGS_H=1
-DHAVE_MATH_H=1 -DRET_TYPE_SRAND_IS_VOID=1 -DHAVE_CC_OSFCN_H=1 -
DHAVE_CC_LIMITS_H=1 -DRETSIGTYPE=void -DHAVE_STRUCT_EXCEPTION=1 -
DHAVE_STDLIB_H=1 -DHAVE_UNISTD_H=1 -DHAVE_GETPAGESIZE=1 -DHAVE_FMOD=1 -
DHAVE_STRTOL=1 -DHAVE_GETCWD=1 -DHAVE_STRERROR=1 -DHAVE_PUTENV=1 -
DHAVE_RENAME=1 -DHAVE_STRCASECMP=1 -DHAVE_STRNCASECMP=1 -DHAVE_STRSEP=1 -
DHAVE_STRDUP=1 -DHAVE_MKSTEMP=1 -DHAVE_MKSTEMP_PROTO=1 -g -D_ALL_SOURCE -
I/usr/local/include  -c change_lf.cc
…
Making pfbtops.n from pfbtops.man
rm -f DESC
cat /dev/fs/C/src/groff-1.17.2/font/devps/DESC.in >DESC
echo broken 7 >>DESC
if test "letter" = A4; then  echo "paperlength 841890" >>DESC;  else
echo "paperlength 792000" >>DESC;  fi
test -z 'lp' || echo print 'lp' >>DESC
rm -f prologue
sed -f /dev/fs/C/src/groff-1.17.2/font/devps/psstrip.sed prologue.ps
>prologue
rm -f symbolsl.pfa
sed -f /dev/fs/C/src/groff-1.17.2/font/devps/psstrip.sed symbolsl.ps
>symbolsl.pfa
rm -f zapfdr.pfa
sed -f /dev/fs/C/src/groff-1.17.2/font/devps/psstrip.sed zapfdr.ps
>zapfdr.pfa
cat /dev/fs/C/src/groff-1.17.2/font/devdvi/DESC.in >DESC
test -z 'lp' || echo print 'lp' >>DESC
rm -f DESC
echo "res 600" >DESC
echo "unitwidth `expr 7620000 / 600`" >>DESC
cat /dev/fs/C/src/groff-1.17.2/font/devlj4/DESC.in >>DESC
if test "letter" = A4; then  echo "papersize a4" >>DESC;  else  echo
"papersize
letter" >>DESC;  fi
test -z 'lp' || echo print 'lp' >>DESC
Making R
…
chmod +x nroff
Making nroff.n from nroff.man
rm -f mmroff
sed -e 's;/usr/bin/perl;/bin/perl;' /dev/fs/C/src/groff-
1.17.2/contrib/mm/mmroff.pl >mmroff
chmod +x mmroff
Making groff_mm.n from groff_mm.man
Making groff_mmse.n from groff_mmse.man
Making mmroff.n from mmroff.man
$

It looks like we've got a good build, but there some things that we should take the time to look at. Sometimes, GNU packages create a config.h file that contains different constant definitions. In this case, those constants are defined in the compilation. By reviewing these compile flags we can learn some things about the Interix. For example, the "-DHAVE_*_H" flags indicate that we have a particular header file. "-DHAVE_STRSEP" means that we have the strsep(2) API (separate strings). Sometimes, we may have functionality not found by the configure script. This would be the case for the makedev() function. For example, on Solaris8, this is implemented as a function, but in Interix, it is defined as a macro in /usr/include/sys/makedev.h. A configure script that is testing for makedev, may not include the correct header file and thus not determine that makedev() is available with Interix. In this case, you would need to either alter the CFLAGS (as in this build) or modify the config.h file.

Since we have a good build, let's try a "make check"

$ make check
$

That didn't seem to do anything, however, there is a test-groff file in the directory, by inspection, it looks like it could be used, and there is a description in the INSTALL file that describes how to use it.

$ ./test-groff -man -Tascii src/roff/groff/groff.n | less

Running this command, we get what looks like a man page, so we've probably got a good build. At this point, we should attempt a make install.

$ make install
test -d /usr/local/man || /dev/fs/C/src/groff-1.17.2/mkinstalldirs
/usr/local/man
test -d /usr/local/man/man1 || /dev/fs/C/src/groff-1.17.2/mkinstalldirs
/usr/local/man/man1
test -d /usr/local/man/man5 || /dev/fs/C/src/groff-1.17.2/mkinstalldirs
/usr/local/man/man5
test -d /usr/local/man/man7 || /dev/fs/C/src/groff-1.17.2/mkinstalldirs
/usr/local/man/man7
test -d /usr/local/bin || /dev/fs/C/src/groff-1.17.2/mkinstalldirs
/usr/local/bin
rm -f /usr/local/bin/groff
/bin/install -c groff /usr/local/bin/groff
…
rm -f /usr/local/bin/mmroff
/bin/install -c mmroff /usr/local/bin/mmroff
test -d /usr/local/share/groff/1.17.2/tmac || /dev/fs/C/src/groff-
1.17.2/mkinstalldirs /usr/local/share/groff/1.17.2/tmac
rm -f /usr/local/share/groff/1.17.2/tmac/tmac.m
rm -f /usr/local/share/groff/1.17.2/tmac/m.tmac
/bin/install -c -m 644 /dev/fs/C/src/groff-1.17.2/contrib/mm/m.tmac
/usr/local/share/groff/1.17.2/tmac/m.tmac
test -d /usr/local/share/groff/1.17.2/tmac/mm || /dev/fs/C/src/groff-
1.17.2/mkinstalldirs /usr/local/share/groff/1.17.2/tmac/mm
for f in 0.MT 5.MT 4.MT ms.cov se_ms.cov; do  rm -f
/usr/local/share/groff/1.17.2/tmac/mm/$f;  /bin/install -c -m 644
/dev/fs/C/src/groff-1.17.2/contrib/mm/mm/$
f /usr/local/share/groff/1.17.2/tmac/mm/$f;  done
for f in locale se_locale; do  test -f
/usr/local/share/groff/1.17.2/tmac/mm/$f
|| touch /usr/local/share/groff/1.17.2/tmac/mm/$f;  done

Another application compiled, tested and installed in minutes!

perl 5.6.1

Now that we have a number of applications and libraries built, we should look at something more complicated. PERL is a fairly complicated package that runs on a large number of UNIX systems as well as a native Win32 port. However, there are some differences between the UNIX and Windows releases. For example, the Windows release does not support the UNIX model for shared memory and does not provide an "exec" function. PERL also has a different configure script, that is somewhat different than the GNU configure script. This Configure script is interactive, and allows for choices to be made as the configuration progresses. For purposes of this document, I have decided to use the non-interactive mode and describe the flags that are set to provide the best PERL port. In order to configure PERL, we will use the following flags:

$ ./Configure -dOes -Ddosuid=undef -Dprefix=/usr/local \
-Dd_bincompat3=undef -Dcc=gcc -Dcflags="-D_ALL_SOURCE" -Dd_poll=undef \
-Dldflags="--dynamic -Wl,-E" -Dusenm=define -Dld="gcc" \
-Dlddlflags="--shared" -Ddont_use_nlink=define -Dusemymalloc="undef" \
-Dcccdlflags="none" libpth='/usr/local/lib /usr/lib' \
-Dinstallusrbinperl=undef -Dcf_email="support@microsoft.com" \
libs='-ldb -lm -lcrypt -lgdbm -lrpclib -ldl' \
perllibs='-ldb -lm -lcrypt -lgdbm -lrpclib -ldl'
$

Going through the flags one by one:

Oes
dosuid	Do not provide a setuid perl
prefix	set the installation directory to /usr/local
d_bincompat3=undef	Don't be binary compatible with earlier releases
cc=gcc	Use the gcc compiler
cflags="-D_ALL_SOURCE"	Add "_ALL_SOURCE" to CFLAGS. This flag is used to get access to more than the standard POSIX APIs.
d_poll=undef	Do not use the poll API. With Interix, this API is only usefull with the /proc filesystem. The reason that this is disabled in PERL is because PERL will try to use this as a general purpose poll API.
ldflags="--dynamic -Wl,-E"	In order to take advantage of dynamic linking, we need to provide the appropriate flags to gcc. The "--dynamic" flag instructs gcc to link against the dynamic libaries provided by Inteirx. the "-Wl, -E" ensures that the linker will add all symbols to the dynamic-symbol table. This will be required for proper operation of the dynamic modules.
ld="gcc"	Use gcc as a linker. For Interix, this is the easiest way to build shared libraries.
usenm=define	Use NM to determine what symbols and routines are available
lddlflags="--shared"	add "--shared" to the command line when building shared libraries
dont_use_nlink=define	PERL will check the nlink member of the stat structure to determine whether the file is a directory. Interix does not use the nlink member to designate a directory.
usemymalloc="undef"	Do not use the PERL malloc
cccdlflags="none"	no additional flags are necessary for dynamic linking
installusrbinperl=undef	do not install /usr/bin/perl
cf_email="myname@somewhere.com"	set the email address of the builder
libpth='/usr/local/lib /usr/lib'	use this as the library path when looking for libraries.
libs='-ldb -lm -lcrypt -lgdbm -lrpclib -ldl'	Use these libraries when linking PERL

If these flags are used, a build of perl should be successful.

$ ./Configure -dOes -Ddosuid=undef -Dprefix=/usr/local \
-Dd_bincompat3=undef -Dcc=gcc -Dcflags="-D_ALL_SOURCE" -Dd_poll=undef \
-Dldflags="--dynamic -Wl,-E" -Dusenm=define -Dld="gcc" \
-Dlddlflags="--shared" -Ddont_use_nlink=define -Dusemymalloc="undef" \
-Dcccdlflags="none" libpth='/usr/local/lib /usr/lib' \
-Dinstallusrbinperl=undef -Dcf_email="support@microsoft.com" \
libs='-ldb -lm -lcrypt -lgdbm -lrpclib -ldl' \
perllibs='-ldb -lm -lcrypt -lgdbm -lrpclib -ldl'
(I see you are using the Korn shell.  Some ksh's blow up on Configure,
mainly on older exotic systems.  If yours does, try the Bourne shell
instead.)
First let's make sure your kit is complete.  Checking...
Locating common programs...
Checking compatibility between /bin/echo and builtin echo (if any)...
Symbolic links are supported.
…
Finding dependencies for hash.o.
Finding dependencies for str.o.
Finding dependencies for util.o.
Finding dependencies for walk.o.
echo Makefile.SH cflags.SH | tr ' ' '\n' >.shlist
Updating makefile...
Now you must run 'make'.
If you compile perl5 on a different machine or from a different object
directory, copy the Policy.sh file from this object directory to the
new one before you run Configure -- this will help you with most of
the policy defaults.

OK, now let's try to build PERL

$ make
`sh  cflags libperl.a miniperlmain.o`  miniperlmain.c
          CCCMD =  gcc -DPERL_CORE -c -I/usr/local/include -O
`sh  cflags libperl.a perl.o`  perl.c
          CCCMD =  gcc -DPERL_CORE -c -I/usr/local/include -O
perl.c: In function `S_open_script':
perl.c:2573: warning: assignment makes pointer from integer without a
cast
`sh  cflags libperl.a gv.o`  gv.c
          CCCMD =  gcc -DPERL_CORE -c -I/usr/local/include -O
`sh  cflags libperl.a toke.o`  toke.c
          CCCMD =  gcc -DPERL_CORE -c -I/usr/local/include -O
…
chmod 644 re.bs
rm -f ../../lib/auto/re/re.so
LD_RUN_PATH="" gcc  --shared -L/usr/local/lib re_exec.o re_comp.o re.o  -
o ../..
/lib/auto/re/re.so
chmod 755 ../../lib/auto/re/re.so
cp re.bs ../../lib/auto/re/re.bs
chmod 644 ../../lib/auto/re/re.bs
        Making Errno (nonxs)
Writing Makefile for Errno
../../miniperl -I../../lib -I../../lib -I../../lib -I../../lib
Errno_pm.PL Errno
.pm
cp Errno.pm ../../lib/Errno.pm
        Everything is up to date. 'make test' to run test suite.
$

This looks very promising, now let's attempt a test

$ make test
        AutoSplitting perl library
./miniperl -Ilib -e 'use AutoSplit;  aut
*/*.pm
        Making DynaLoader (static)
gcc -o perl --dynamic -Wl,-E -L/usr/loca
naLoader.a  libperl.a `cat ext.libs` -ls
        Making utilities
…
        Making attrs (dynamic)
        Making re (dynamic)
        Making Errno (nonxs)
cd t && (rm -f perl; /bin/ln -s ../perl
if (true </dev/tty) >/dev/null 2>&1; the
  cd t && PERL_SKIP_TTY_TEST=1  ./perl T
base/cond............ok
…
io/openpid...........Hangup
$

Oh, oh. This could be a problem. The PERL test is exiting and the test is reporting a Hangup. By inspecting the test, we see that SIGPIPE is being ignored, perhaps we should also ignore SIGHUP.

To attempt a work-around, we add

         $SIG{HUP} = 'IGNORE';

after the

         $SIG{PIPE} = 'IGNORE';

and rerun our make test.

$ make test
        AutoSplitting perl library
…
op/goto_xs...........ok
op/grent.............FAILED at test 1
op/grep..............ok

apache

rpm

ssh

Summary of Interfaces

These are the interfaces documented with Interix 2.2. This list is a guide, not an absolute reference. Some of the APIs are provided for compatibility and may follow either the BSD or System V behavior. See the reference pages for more information.

This list does not include macro definitions. For example, getdtablesize() is available as a macro in <unistd.h>; it is not listed here. (The macro actually calls sysconf(_SC_OPEN_MAX).)

This list also does not include the X11R5 routines or the Motif 1.2.4 routines. The X11R5 libraries come with the Interix SDK.

Because the list spans several pages, the table is organized alphabetically across rows.

System Interfaces, Standard C, Math, DB, RPC, Lex, Yacc and dl Libraries

a64l	floorf	isalnum	realpath	strcoll	clntudp_bufcreate
abort	fmod	isalpha	recv	strcpy	clntudp_create
abs	fnmatch	isascii	recvfrom	strcspn	dbm_clearerr
accept	fopen	isatty	regcomp	strdup	dbm_close
access	fork	isblank	regerror	strerror	dbm_delete
acos	fpathconf	iscntrl	regexec	strftime	dbm_error
acosh	fprintf	isdigit	regfree	strlcat	dbm_fetch
alarm	fpurge	isgraph	remainder	strlcpy	dbm_firstkey
asctime	fputc	isinf	remove	strlen	dbm_nextkey
asin	fputs	isinff	rename	strmode	dbm_open
asinh	fread	islower	renamewtmpx	strncasecmp	dbm_store
assert	free	isnan	rewind	strncat	endrpcent
atan	freopen	isnanf	rewinddir	strncmp	get_myaddress
atan2	frexp	isprint	rexec	strncpy	getrpcbyname
atan2f	fscanf	ispunct	rindex	strpbrk	getrpcbynumber
atanf	fseek	isspace	rint	strptime	getrpcent
atanh	fsetpos	isupper	rmdir	strrchr	getrpcport
atexit	fstat	isxdigit	rresvport	strsep	longjmp
atof	fstatvfs	j0	ruserok	strsignal	pmap_getmaps
atoi	fsync	j1	sbrk	strsigname	pmap_getport
atol	ftell	jn	scalb	strspn	pmap_rmtcall
basename	ftime	jrand48	scalbn	strstr	pmap_set
bcmp	ftok	kill	scalbnf	strtod	pmap_unset
bcopy	ftruncate	killpg	scanf	strtok	registerrpc
bind	fts_children	l64a	seed48	strtol	setjmp
brk	fts_close	labs	seekdir	strtoul	setrpcent
bsearch	fts_open	lchown	select	strunvis	svc_getreq
bzero	fts_read	lcong48	semctl	strvis	svc_getreqset
cabs	fts_set	ldexp	semget	strvisx	svc_register
calloc	ftw	ldiv	semop	strxfrm	svc_run
catclose	fwrite	lfind	send	swab	svc_sendreply
catgets	gamma	lgamma	sendto	symlink	svc_unregister
catopen	gamma_r	lgamma_r	setbuf	sysconf	svcerr_auth
cbrt	gcvt	link	setbuffer	syslog	svcerr_decode
ceil	getbsize	listen	setegid	system	svcerr_noproc
cfgetispeed	getc	localeconv	setenv	tan	svcerr_noprog
cfgetospeed	getchar	localtime	seteuid	tanf	svcerr_progvers
cfsetispeed	getcwd	lockf	setgid	tanh	svcerr_systemerr
cfsetospeed	getdtablesize	log	setgrent	tcdrain	svcerr_weakauth
chdir	getegid	log10	sethostent	tcflow	svcfd_create
chmod	getenv	log1p	setitimer	tcflush	svcraw_create
chown	geteuid	logb	setjmp	tcgetattr	svctcp_create
chroot	getgid	login	setlinebuf	tcgetpgrp	svcudp_bufcreate
clearerr	getgrent	longjmp	setlocale	tcsendbreak	t_accept
clock	getgrgid	lrand48	setlogmask	tcsetattr	t_alloc
close	getgrnam	lsearch	setmode	tcsetpgrp	t_bind
closedir	getgroups	lseek	setnetent	tdelete	t_close
closelog	gethostbyaddr	lstat	setpgid	telldir	t_connect
confstr	gethostbyname	madvise	setpriority	tempnam	t_error
connect	gethostent	malloc	setprotoent	tfind	t_free
copysign	gethostname	mctl	setpwent	time	t_getinfo
copysignf	getitimer	memccpy	setregid	times	t_getprotaddr
cos	getlogin	memchr	setreuid	tmpfile	t_getstate
cosf	getmode	memcmp	setrlimit	tmpnam	t_listen
cosh	getnetbyaddr	memcntl	setservent	toascii	t_look
creat	getnetbyname	memcpy	setsid	truncate	t_open
ctermid	getnetent	memmove	setsockopt	tsearch	t_optmgmt
ctime	getopt	memset	setstate	ttyname	t_rcv
cuserid	getpass	mkdir	settimeofday	ttyslot	t_rcvconnect
daemon	getpeername	mkfifo	setuid	twalk	t_rcvdis
difftime	getpgrp	mknod	setutxent	tzset	t_rcvrel
dirfd	getpid	mkstemp	setvbuf	ualarm	t_rcvudata
dirname	getppid	mktemp	shmat	ulimit	t_rcvuderr
div	getpriority	mktime	shmctl	umask	t_snd
drand48	getprotobyname	mmap	shmdt	uname	t_snddis
drem	getprotobynumber	modf	shmget	ungetc	t_sndrel
dup	getprotoent	mprotect	shutdown	unlink	t_sndudata
dup2	getpwent	mrand48	sigaction	unlockpt	t_strerror
ecvt	getpwnam	msgctl	sigaddset	unsetenv	t_sync
endgrent	getpwuid	msgget	sigblock	unvis	t_unbind
endhostent	getrlimit	msgrcv	sigdelset	usleep	xdr_accepted_reply
endnetent	getrusage	msgsnd	sigemptyset	utime	xdr_array
endprotoent	gets	msync	sigfillset	verr	xdr_authunix_parms
endpwent	getservbyname	munmap	sighold	verrx	xdr_bool
endservent	getservbyport	nextafter	sigignore	vfork	xdr_bytes
endutxent	getservent	nftw	sigismember	vfprintf	xdr_callhdr
erand48	getsockname	nice	siglongjmp	vis	xdr_callmsg
erf	getsockopt	nrand48	signal	vprintf	xdr_char
erfc	gettimeofday	ntohl	significand	vscanf	xdr_double
err	getuid	ntohs	sigpause	vsnprintf	xdr_enum
errx	getutxent	open	sigpending	vsprintf	xdr_float
execl	getutxid	opendir	sigprocmask	vsscanf	xdr_free
execle	getutxline	openlog	sigrelse	vsyslog	xdr_int
execlp	glob	openpty	sigset	vwarn	xdr_long
execv	globfree	pathconf	sigsetjmp	vwarnx	xdr_opaque
execve	gmtime	pause	sigsetmask	wait	xdr_opaque_auth
execvp	grantpt	pclose	sigsuspend	waitpid	xdr_pmap
exp	hcreate	perror	sigvec	warn	xdr_pmaplist
expm1	hcreate_r	pipe	sin	warnx	xdr_pointer
fabs	hdestroy	poll	sinf	write	xdr_reference
fabsf	hdestroy_r	popen	sinh	writev	xdr_rejected_reply
fchdir	hsearch	pow	sleep	y0	xdr_replymsg
fchmod	hsearch_r	printf	snprintf	y1	xdr_short
fchown	htonl	ptsname	socket	yn	xdr_string
fclose	htons	putc	socketpair	z_abs	xdr_u_char
fcntl	hypot	putchar	sprintf	authnone_create	xdr_u_int
fcvt	ilogb	putenv	sqrt	authunix_create	xdr_u_long
fdopen	index	puts	sqrtf	authunix_create_default	xdr_u_short
feof	inet_addr	pututxline	sradixsort	bindresvport	xdr_union
ferror	inet_aton	qsort	srand	callrpc	xdr_vector
fflush	inet_lnaof	radixsort	srand48	clnt_broadcast	xdr_void
ffs	inet_makeaddr	raise	srandom	clnt_create	xdr_wrapstring
fgetc	inet_netof	rand	sscanf	clnt_pcreateerror	xdrmem_create
fgetln	inet_network	random	stat	clnt_perrno	xdrrec_create
fgetpos	inet_ntoa	rcmd	statvfs	clnt_perror	xdrrec_endofrecord
fgets	init	read	strcasecmp	clnt_spcreateerror	xdrrec_eof
fileno	initstate	readdir	strcasestr	clnt_sperrno	xdrrec_skiprecord
finite	ioctl	readlink	strcat	clnt_sperror	xdrstdio_create
finitef	ipcs	readv	strchr	clntraw_create	xprt_register
flock	iruserok	realloc	strcmp	clnttcp_create	xprt_unregister
floor	isalnum	realpath	strcoll	clntudp_bufcreate	dlopen
dlsym	dlerror	dlclose

Curses, Menus, Panels ,Forms and Terminfo APIs

The following APIs make up the ported ncurses libraries:

addch	init_color	mvinsch	ripoffline	ungetch
addchnstr	init_pair	mvinsnstr	savetty	ungetmouse
addchstr	initscr	mvinsstr	scanw	unpost_menu
addnstr	innstr	mvinstr	scr_dump	untouchwin
addstr	insch	mvprintw	scr_init	update_panels
attr_get	insdelln	mvscanw	scr_restore	use_env
attr_off	insertln	mvwaddch	scr_set	vidattr
attr_on	insnstr	mvwaddchnstr	scrl	vidputs
attr_set	insstr	mvwaddchstr	scroll	vline
attroff	instr	mvwaddnstr	scrollok	vw_printw
attron	intrflush	mvwaddstr	set_current_item	vw_scanw
attrset	is_linetouched	mvwchgat	set_curterm	vwprintw
baudrate	is_wintouched	mvwdelch	set_item_init	vwscanw
beep	isendwin	mvwgetch	set_item_opts	waddch
bkgd	item_count	mvwgetstr	set_item_term	waddchnstr
bkgdset	item_description	mvwin	set_item_userptr	waddchstr
border	item_index	mvwinch	set_item_value	waddnstr
bottom_panel	item_init	mvwinchnstr	set_menu_back	waddstr
box	item_name	mvwinchstr	set_menu_fore	wattr_get
can_change_color	item_opts	mvwinnstr	set_menu_format	wattr_off
cbreak	item_opts_off	mvwinsch	set_menu_grey	wattr_on
chgat	item_opts_on	mvwinsnstr	set_menu_init	wattr_set
clear	item_term	mvwinsstr	set_menu_items	wattroff
clearok	item_userptr	mvwinstr	set_menu_mark	wattron
clrtobot	item_value	mvwprintw	set_menu_opts	wattrset
clrtoeol	item_visible	mvwscanw	set_menu_pad	wbkgd
color_content	keyname	napms	set_menu_pattern	wbkgdset
copywin	keypad	new_item	set_menu_sub	wborder
current_item	killchar	new_menu	set_menu_term	wchgat
curs_set	leaveok	new_panel	set_menu_userptr	wclear
def_prog_mode	longname	newpad	set_panel_userptr	wclrtobot
def_shell_mode	menu_back	newterm	set_term	wclrtoeol
del_curterm	menu_driver	newwin	set_top_row	wcursyncup
del_panel	menu_fore	nl	setscrreg	wdelch
delay_output	menu_format	nocbreak	setupterm	wdeleteln
delch	menu_grey	nodelay	show_panel	wechochar
deleteln	menu_init	noecho	slk_clear	werase
delscreen	menu_items	nonl	slk_init	wgetch
delwin	menu_mark	noqiflush	slk_label	wgetstr
derwin	menu_opts	noraw	slk_noutrefresh	whline
doupdate	menu_opts_off	notimeout	slk_refresh	winch
dupwin	menu_opts_on	overlay	slk_restore	winchnstr
echo	menu_pad	overwrite	slk_set	winchstr
echochar	menu_pattern	pair_content	slk_touch	winnstr
endwin	menu_sub	panel_above	standend	winsch
erasechar	menu_term	panel_below	standout	winsdelln
filter	menu_userptr	panel_hidden	start_color	winsertln
flash	menu_win	panel_userptr	subpad	winsnstr
flushinp	meta	panel_window	subwin	winsstr
free_item	mousemask	pechochar	syncok	winstr
free_menu	move	pnoutrefresh	termattrs	wmove
getch	move_panel	pos_menu_cursor	termname	wnoutrefresh
getmouse	mvaddch	post_menu	tgetent	wprintw
getstr	mvaddchnstr	prefresh	tgetflag	wredrawln
getwin	mvaddchstr	printw	tgetnum	wrefresh
halfdelay	mvaddnstr	putp	tgetstr	wresize
has_colors	mvaddstr	putwin	tgoto	wscanw
has_ic	mvchgat	qiflush	tigetflag	wscrl
has_il	mvcur	raw	tigetnum	wsetscrreg
hide_panel	mvdelch	redrawwin	tigetstr	wstandend
hline	mvderwin	refresh	timeout	wstandout
idcok	mvgetch	replace_panel	top_panel	wsyncdown
idlok	mvgetstr	reset_prog_mode	top_row	wsyncup
immedok	mvinch	reset_shell_mode	tparm	wtimeout
inch	mvinchnstr	resetty	tputs	wtouchln
inchnstr	mvinchstr	restartterm	typeahead	wvline
inchstr	mvinnstr	ripoffline	unctrl

Interix-Specific APIs

The following APIs are unique to Interix:

_getInstallPath	_getInstallPath_Win	_prefixInstallPath
_prefixInstallPath_Win	authenticateuser	chpass
env_alloc	env_array	env_cron
env_expand_win	env_free	env_get
env_login	env_put	env_putarray
env_set	env_strfree	env_unset
env_winlogin	execl_asuser	execle_asuser
execlp_asuser	execv_asuser	execve_asuser
execvp_asuser	fullqualname	getgrent_nomembers
getgrgid_ex	getgrgid_nomembers	getgrnam_nomembers
getloginenv	getpdomain	getpwuid_ex
getreg_strvalue	getreg_wstrvalue	getsecret
gettzenv	group_from_gid	ntpath2posix
path_casesensitive	path_convert	setsecret
setuser	stripdomainprefix	uidtontsid
unixpath2win	user_from_uid	winpath2unix

The routines _getInstallPath() and _prefixInstallPath() retrieve the Interix install root ($INTERIX_ROOT) and prefix it to a string, respectively. They are discussed in the Interix SDK User's Guide.

The authenticateuser() and exec*_asuser() routines are used to login as another user. They are discussed in the Interix SDK User's Guide. These APIs are now deprecated in favor of the setuser() call; the man page for the setuser() call is included in this note, because it was not included in the 2.2 Service Pack 1 release.

The uidtontsid() routine converts a POSIX-style user ID number into the user's Windows NT security ID.

The remaining routines are convenience routines:

The _ntpath2posix() routine converts a Windows NT-style pathname such as C:\Interix to its Interix equivalent, //C/Interix.

The stripdomainprefix() routine removes the domain prefix from a user name, so that a string such as "ADMIN+jo" becomes "jo".

The gettzenv() routine retrieves the value of the Win32 system's TZ environment variable.

For more information about the routines, see the relevant reference pages using the man command or using the Adobe Acrobat documents.

setuser() Reference

NAME

setuser() — change effective and real uid and gid of process

SYNOPSIS

#include <opennt/security.h>
int setuser(char *username, char *password, int flags)

DESCRIPTION

The setuser() function changes the effective and real uid and gid of the current process to that of the specified username. All of the security attributes and permissions become those of username. However, the process' current working directory does not change.

The setuser() utility takes the following arguments:

username

The name of the user. If the username is not fully qualified with a domain name (that is, "domain+name"), it is not changed. On a system configured for workgroups, specify the domain as NULL: that is, the string "+name".

password

The plaintext password for the specified username.

flags

Control flags. These possible values for flags are defined in <opennt/security.h>:

SU_COMPLETE

Change the real and effective user and group IDs and all security attributes to the default for the specified user.

SU_CHECK

Check that the process will be able to perform a setuser() action using SU_COMPLETE for the specified username and password. This is a quick way to verify a password for a user. This action is equivalent to the older Interix call, authenticateuser().

There can be a performance degradation if a process changes identity to a user who does not have permission to be located in the current working directory. The best solution is, after a call to setuser(), to chdir() to a directory known to be permitted for the new identity.

RETURN VALUE

On success, the setuser() function returns 0. On failure, it returns -1 and sets errno.

ERRORS

The setuser() function can fail for the following reasons:

[EINVAL]

An argument is invalid or is missing.

[ENOMEM]

Not enough memory to complete the request.

[EPERM]

The requesting process does not have permission to change to the new user. One or both of the username or password were incorrectly specified.

NOTES:

This function replaces authenticateuser(). In fact, the version of authenticateuser() in this release is a wrapper around setuser().

SEE ALSO

setuid(), setgid(), exec_asuser(), authenticateuser()

For More Information

For more information about Interix, see the Interix Web site at: https://www.microsoft.com/windows2000/interix.