sysarch - defines system specific variables using the uname system command
sysarch [OPTION]... [VARIABLE] [VARIABLE]
sysarch provides a uniform description of a system, including the underlying hardware, the operating system, and the unique operating system characteristics. Its intent is to identify important system characteristics which determine the interoperability of software with hardware/operating system platforms.
Compatibility between operating systems sometimes occurs. Solaris is a true champion of this, retaining compatibility across more than a decade in some cases. It's also often possible to run programs compiled on one flavor of Linux on another. It is rare for software compiled on newer platforms to run on older ones, however. While sysarch can't determine which systems are compatible, it tries to present enough salient information so that it is possible to provide a hierarchy of compatibility.
sysarch encodes system information into a set of variables. It will generate code for various shells and GNU make which can be used to incorporate those variables into scripts and Makefiles.
sysarch takes an optional argument specifying the name of a variable to output. If no options or arguments are specified they are all output.
Output code which creates environmental, rather than shell, variables. Only releveant for shells.
Prepend pfx to variable names.
Print this usage information and exit.
Output formats are specified either using the --output option, or
simply by prefixing the format name with -- and treating it as an
option:
--output csh --csh
If no variables are specified on the command line, the default output
format is nice. If variables are specified, it defaults to value.
The following formats are supported:
Output commands in the appropriate shell to create the variables.
Create static make compatible macro assignments. Macro names are in upper case.
Create dynamic GNU make compatible macro assignments using a call to sysarch. Macro names are in upper case.
Create the output required by the code output by --output
gmake_extract.
Generate make target commands to output the sysarch generated
macros. Useful as part of a make debug style target.
Output Perl code generating scalar variables.
Output Perl code for the contents of a hash. It does not generate the wrapper code required to actually create the hash.
Output only the variables' values.
A more human readable format.
The most specific description of the actual CPU.
The generic description for the CPU. This is not the most generic one with which the CPU is compatible. For example,
athlon_64 => x86_64, not i686
core2 => x86_64, not i686
pentium4 => i686
ultra3i => sun4u, not sparc
A colon separated list of models or families of CPU's with which the hardware is compatible, from least to most generic.
For example, for a SunBlade 1500,
hw_cpu = ultra3i hw_cpus = ultra3i:sun4u:sun4m:sparc
For an AMD Athlon64 CPU,
hw_cpu = athlon_64 hw_cpus = athlon_64:x86_64:i686
The system designation represents the combination of operating system, kernel, and optionally vendor or distribution name.
The operating system name. OS has historical significance for Linux and Solaris, and retains the following value for them:
Linux SunOS
For other operating systems, OS and os_type are equivalent.
os_type is the lowercased more "obvious" name (i.e. solaris
rather than sunos) and is typically equivalent to the OSTYPE
environmental variable.
Note that Linux represents both the kernel and the accompanying
system software. This is often represented as GNU/Linux, similar to
Debian GNU/kFreeBSD, Debian GNU/NetBSD, GNU/Hurd and Debian
GNU/Solaris. The latter set aren't available to the authors, so we're
not worrying about them yet.
os_vendor is the vendor or distribution name. Some vendors provide
multiple distributions (e.g. Red Hat Fedora Core, Red Hat Enterprise
Linux) -- os_vendor is then the combination (e.g. RHFC, RHEL).
System version information is complex. It might refer to the version given to a particular distribution or release (such as Fedora Core 3 or 4). For some applications the kernel version is more important, or perhaps the version of glibc. sysarch provides all of that information.
kern_version is the kernel version. On Linux it is the base kernel, without any vendor specific version information.
kern_version_maj contains just the most significant kernel version (e.g. 2.6 for Linux, 8.8 for Darwin).
A colon separated list of kernels with which this is compatible.
os_version is the vendor supplied operation system version. os_version_maj is the most significant part (e.g, for OSX 10.4 vs. 10.4.8).
A colon separated list of operating system versions with which this operating system is backwards compatible.
The version of the operating system component most important for user space applications. For Linux this is chosen to be the version of glibc. This may not be restrictive enough for some applications (e.g. if one distribution ships with a different version of libreadline). os_version_maj is the most significant part (e.g., 2.3 vs. 2.3.6).
A colon separated list of user space component versions with which this operating system is backwards compatible. For Linux it is assumed that previous minor (in the glibc sense) releases are backwards compatible.
A platform is the combination of hardware and operating system specifications. It's a triad with one of the following combinations:
<hw_cpu>-<os_type>-<os_uversion> <hw_cpu>-<OS>-<os_version>
where <OS> is
<os_type>_<os_vendor>
except for Solaris and OSX (Darwin) where it is
<os_type>
It is assumed that os_uversion is constant for a given os_version.
These variables describe the host system's platform
Aliases for platform_os_native.
An alias for platform_uos_generic.
The most detailed specification including the OS vendor/distribution. Equivalent to
<hw_cpu>-<OS>-<os_version>
This specification allows for a more generic hardware and OS specification, allowing for compatibility within minor OS and hardware variations. This is equivalent to
<hw_cpu_generic>-<OS>-<os_version_maj>
The most detailed specification using the user space component version. Equivalent to
<hw_cpu>-<os_type>-<os_uversion>
Equivalent to
<hw_cpu_generic>-<OS>-<os_uversion_maj>
These variables provide lists of compatible platforms
A colon separated list of compatible platforms, with hardware being the most important element for compatibility. This is a combination of platforms_os and platforms_uos.
A colon separated list of compatible platforms of the form
<hw_cpu>-<OS>-<os_version>
With Hardware being the most important element for compatibility.
A colon separated list of compatible platforms of the form
<hw_cpu>-<os_type>-<os_uversion>
with hardware being the most important element for compatibility.
Well, that depends upon how its being used. First off, use platforms and sysarch's accompanying syspathsubst program to create compatibility hierarchies for executable and library search paths.
platform_generic is not the lowest common denominator. It's essentially the most generic specification that can be natively created (i.e. compiled) on the platform. Because it purports to specify user space compatibility, use it on Linux if you're sure that your application will run on multiple Linux distributions based solely on glibc compatibility. If, however, your application links against libraries in a combination that differs between distributions, you might as well use platform_os_generic.
This software is Copyright 2007 The Smithsonian Astrophysical Observatory and is released under the GNU General Public License. You may find a copy at: http://www.fsf.org/copyleft/gpl.html
M. Tibbetts <mtibbetts@cfa.harvard.edu>
D. Jerius <djerius@cfa.harvard.edu>