View Issue Details [ Jump to Notes ]

[ Print ]

Project

Category

View Status

Date Submitted

Last Update

0004090

CMake

public

2006-11-29 13:53

2007-02-21 10:52

Reporter

Didier Link

Assigned To

Bill Hoffman

Priority

normal

Severity

major

Reproducibility

always

Status

closed

Resolution

fixed

Platform

OS Version

Product Version

Target Version

Fixed in Version

Summary

0004090: mistake in CMAKE_SYSTEM_PROCESSOR definition ?

Description

Hi,

I work on a gentoo linux amd64 and look at some strange results of a project configuration by cmake.

The project assume to find the architecture of the processor (like x86_64 or powerpc etc etc) in the CMAKE_SYSTEM_PROCESSOR variable to add some specific flags during compilation. In my case they need to have -fPIC.

Now the variable contain the result of uname -p and it is "AMD Athlon(tm) 64 X2 Dual Core Processor 4200+" !? Not very usefull bacause of the variety of processor names ... the uname -m command result is "x86_64" and that's what I need.

I think it's a mistake in the file cmake-2.4.4/Modules/CMakeDetermineSystem.cmake that use uname -p in place of uname -m.

Do you think I'm wrong ?

Thanks to consider this change !

Tags

No tags attached.

Attached Files

system-linux.patch [^] (1,349 bytes) 1969-12-31 19:00 [Show Content] [Hide Content]

--- Modules/CMakeDetermineSystem.cmake	2006-11-20 20:41:59.000000000 +0100
+++ Modules/CMakeDetermineSystem.cmake-new	2006-12-01 08:46:31.000000000 +0100
@@ -31,14 +31,19 @@
   IF(CMAKE_UNAME)
     EXEC_PROGRAM(uname ARGS -s OUTPUT_VARIABLE CMAKE_SYSTEM_NAME)
     EXEC_PROGRAM(uname ARGS -r OUTPUT_VARIABLE CMAKE_SYSTEM_VERSION)
-    EXEC_PROGRAM(uname ARGS -p OUTPUT_VARIABLE CMAKE_SYSTEM_PROCESSOR
-      RETURN_VALUE val)
+    IF(CMAKE_SYSTEM_NAME MATCHES "Linux")
+	EXEC_PROGRAM(uname ARGS -m OUTPUT_VARIABLE CMAKE_SYSTEM_PROCESSOR
+	  RETURN_VALUE val)
+    ELSE(CMAKE_SYSTEM_NAME MATCHES "Linux")
+        EXEC_PROGRAM(uname ARGS -p OUTPUT_VARIABLE CMAKE_SYSTEM_PROCESSOR
+	  RETURN_VALUE val)
+	IF("${val}" GREATER 0)
+		EXEC_PROGRAM(uname ARGS -m OUTPUT_VARIABLE CMAKE_SYSTEM_PROCESSOR
+        	RETURN_VALUE val)
+	ENDIF("${val}" GREATER 0)
+    ENDIF(CMAKE_SYSTEM_NAME MATCHES "Linux")
     IF("${val}" GREATER 0)
-      EXEC_PROGRAM(uname ARGS -m OUTPUT_VARIABLE CMAKE_SYSTEM_PROCESSOR
-        RETURN_VALUE val)
-    ENDIF("${val}" GREATER 0)
-    IF("${val}" GREATER 0)
-      SET(CMAKE_SYSTEM_PROCESSOR "unknown")
+	SET(CMAKE_SYSTEM_PROCESSOR "unknown")
     ENDIF("${val}" GREATER 0)
     SET(CMAKE_UNAME ${CMAKE_UNAME} CACHE INTERNAL "uname command")
     # processor may have double quote in the name, and that needs to be removed

Relationships

Notes
(0005828) Bill Hoffman (manager) 2006-11-29 15:09	I think the problem is that some machines (commerical unix) do not support -m. So the code actually does this: EXEC_PROGRAM(uname ARGS -p OUTPUT_VARIABLE CMAKE_SYSTEM_PROCESSOR RETURN_VALUE val) # if the uname -p failed with an error try -m IF("${val}" GREATER 0) EXEC_PROGRAM(uname ARGS -m OUTPUT_VARIABLE CMAKE_SYSTEM_PROCESSOR RETURN_VALUE val) ENDIF("${val}" GREATER 0) For example on an IRIX machine I get this: laserX 1% uname -p mips laserX 2% uname -m IP35 The -p is better....

(0005829) Didier Link (reporter) 2006-11-29 16:43	Yes the -p is better for IRIX but in the posix specs this option doesn't exist !? Maybe it's necessary to add a test case for use the -m option with posix compliant uname (like the GNU one, present on a lot of hardware ... see debian hardware compatibility for example). What do you think about this point ?

(0005832) Bill Hoffman (manager) 2006-11-29 20:32	http://www.tu-berlin.de/zrz/dienste/www/manpages/uname.html [^] seems to say -p is POSIX. but http://www.opengroup.org/onlinepubs/009695399/utilities/uname.html [^] seems to leave out -p.... So, I am not sure what to do.... The IRIX has both, many linux systems do not have -p so it defaults to -m.

(0005847) Didier Link (reporter) 2006-12-01 03:02	For the FreeBSD uname's command, it's the same behaviour of the Linux ones. The -p option "just" send the "commercial name" of the processor ... I send you a patch for experiment my idea on Linux. Maybe a more in-depth work need to be done for include more unixes flavour in the test case ...

(0006505) Bill Hoffman (manager) 2007-02-21 10:52	I have applied the patch

Issue History
Date Modified	Username	Field	Change