(0024694)
Markus Elfring (reporter)
2011-01-14 15:08

My story seems to be related to the bug report "IF() command behaviour contradicts examples in documentation".
http://public.kitware.com/Bug/view.php?id=10773 [^]

(0024696)
Brad King (manager)
2011-01-14 17:21

Closing as duplicate of 0010773. Please visit that bug and click the "Monitor" button to track its progress.

(0024704)
Markus Elfring (reporter)
2011-01-15 02:30

I find that your documentation adjustment is still not enough.
http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=e4e14e85 [^]

Would you like to mention the "magic evaluation" like the condition application "if(DEFINED variable)" in any clearer way?
Where is this behaviour completely described at the moment?

Are any further clarifications needed for handling of variables without "${}" in conditions?

(0024709)
Brad King (manager)
2011-01-15 08:07

The full behavior is documented well above the example that was replaced:

       Possible expressions are:

         if(<constant>)

       True if the constant is 1, ON, YES, TRUE, Y, or a non-zero number.
       False if the constant is 0, OFF, NO, FALSE, N, IGNORE, "", or ends in
       the suffix '-NOTFOUND'. Named boolean constants are case-insensitive.
       If the argument is not one of these constants, it is treated as a
       variable:

         if(<variable>)

       True if the variable is defined to a value that is not a false
       constant. False otherwise.

       ...

         if(DEFINED variable)

       True if the given variable is defined. It does not matter if the
       variable is true or false just if it has been set.

I don't see how that can be any clearer.

(0024712)
Markus Elfring (reporter)
2011-01-15 09:49

I agree that the functionality "if(<variable>)" is documented by itself. But I find that the explanation for the combination with "old style" variable evaluation is still an open issue.
Does the "magic" result in any differences in comparison with the current ${} syntax?

Modestas Vainius and I have got different expectations about the usable behaviour ...

(0024715)
Brad King (manager)
2011-01-15 10:26

> Does the "magic" result in any differences in comparison with the current ${} syntax?

The ${} syntax is evaluated before if() sees the arguments. Then the arguments that if() does see are evaluated as specified in the documentation I quoted. Please ignore the old "behave as if you wrote ${}" example that was just plain wrong. Read the documentation from scratch as it appears now. What is missing?

(0024719)
Markus Elfring (reporter)
2011-01-15 13:00

Please look at my script example for this issue (again):
A variable was set to a value that I would not interpret as the boolean status "true" according to the updated documentation.

Why was the message "Good day." displayed?

(0024746)
Brad King (manager)
2011-01-17 09:28

In the line

if(MY_TEXT)

"MY_TEXT" is not one of the named boolean constants. Therefore the command moves on to the "if(<variable>)" form, which is documented as

True if the variable is defined to a value that is not a false constant. False otherwise.

The value of the variable is "condition-test1.cxx", which is not a false constant. Therefore it is TRUE.

(0024747)
Brad King (manager)
2011-01-17 09:47

This makes the signatures where auto-dereferencing occurs more explicit:

http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=7d9b9037 [^]

(0024848)
Brad King (manager)
2011-01-18 07:26

Closing as "fixed" since the commit mentioned in 0011701:0024747 clarifies the documentation further.

(0024850)
Markus Elfring (reporter)
2011-01-18 07:31

Thanks for your explanation and the fine-tuning of the documentation.

I have got the impression that a syntax diagram (or decision tree) might be needed to understand the dependencies of parameter evaluation a bit easier here.
http://en.wikipedia.org/wiki/File:Syntax_Diagrams.png [^]

Are the displays the same from the execution of the following script example by a couple of CMake versions?

cmake_minimum_required(VERSION 2.6)
project(my_condition_test2)
set(my_text "condition-test2")
if(my_text)
   set(result1 one)
else()
   set(result1 two)
endif()
if(${my_text})
   set(result2 one)
else()
   set(result2 two)
endif()
if("Bluff")
   set(result3 one)
else()
   set(result3 two)
endif()
if( (${result1} STREQUAL ${result2})
   AND (${result1} STREQUAL ${result3}))
   message(STATUS "This result was expected.")
else()
   message(FATAL_ERROR "Results:
1. ${result1}
2. ${result2}
3. ${result3}
Check your expectations again!")
endif()

I get the following log display from CMake-GUI 2.8.3.
"CMake Error at CMakeLists.txt:23 (message):
  Results:

  1. one

  2. two

  3. two

  Check your expectations again!

Configuring incomplete, errors occurred!"

How do you think about this situation now?

By the way:
Does it also show an open issue that blank lines are added to the original output text?

(0024851)
Brad King (manager)
2011-01-18 07:54

The results *exactly* match the documentation!

Part 1: "my_text" is a variable defined to a value that is not a false constant -> true

Part 2: "${my_text}" becomes "condition-test2" before if() even sees it. The latter is not a defined variable -> false

Part 3: "Bluff" is not a defined variable -> false

Please, this issue tracker is not a place to ask for tutorials. Go to the mailing list and ask for help there. I don't mean to be dismissive but I've already spent way too much time on this. Only a few people read the issue tracker and we don't have time to give everyone detailed tutorial sessions. There are may people on the mailing list that answer questions. It distributes the load.

The documentation is correct and consistent with the implementation and the behavior that we intend. The examples are correct. This issue is resolved. If you open it again without a link to discussion on the mailing list that involves posts other than yours, I will close it without answering again.

Issue History
Date Modified	Username	Field	Change
2011-01-14 14:51	Markus Elfring	New Issue
2011-01-14 15:08	Markus Elfring	Note Added: 0024694
2011-01-14 17:20	Brad King	Relationship added	duplicate of 0010773
2011-01-14 17:21	Brad King	Note Added: 0024696
2011-01-14 17:21	Brad King	Status	new => closed
2011-01-14 17:21	Brad King	Assigned To	=> Brad King
2011-01-14 17:21	Brad King	Resolution	open => duplicate
2011-01-15 02:30	Markus Elfring	Note Added: 0024704
2011-01-15 02:30	Markus Elfring	Status	closed => feedback
2011-01-15 02:30	Markus Elfring	Resolution	duplicate => reopened
2011-01-15 08:07	Brad King	Note Added: 0024709
2011-01-15 09:49	Markus Elfring	Note Added: 0024712
2011-01-15 09:49	Markus Elfring	Status	feedback => assigned
2011-01-15 10:26	Brad King	Note Added: 0024715
2011-01-15 13:00	Markus Elfring	Note Added: 0024719
2011-01-17 09:28	Brad King	Note Added: 0024746
2011-01-17 09:47	Brad King	Note Added: 0024747
2011-01-18 07:26	Brad King	Note Added: 0024848
2011-01-18 07:26	Brad King	Status	assigned => closed
2011-01-18 07:26	Brad King	Resolution	reopened => fixed
2011-01-18 07:31	Markus Elfring	Note Added: 0024850
2011-01-18 07:31	Markus Elfring	Status	closed => feedback
2011-01-18 07:31	Markus Elfring	Resolution	fixed => reopened
2011-01-18 07:54	Brad King	Note Added: 0024851
2011-01-18 07:54	Brad King	Status	feedback => closed
2011-01-18 07:54	Brad King	Resolution	reopened => fixed
2011-01-31 16:01	David Cole	Fixed in Version	=> CMake 2.8.4
2011-01-31 16:01	David Cole	Target Version	=> CMake 2.8.4

Notes
(0024694) Markus Elfring (reporter) 2011-01-14 15:08	My story seems to be related to the bug report "IF() command behaviour contradicts examples in documentation". http://public.kitware.com/Bug/view.php?id=10773 [^]

(0024696) Brad King (manager) 2011-01-14 17:21	Closing as duplicate of 0010773. Please visit that bug and click the "Monitor" button to track its progress.

(0024704) Markus Elfring (reporter) 2011-01-15 02:30	I find that your documentation adjustment is still not enough. http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=e4e14e85 [^] Would you like to mention the "magic evaluation" like the condition application "if(DEFINED variable)" in any clearer way? Where is this behaviour completely described at the moment? Are any further clarifications needed for handling of variables without "${}" in conditions?

(0024709) Brad King (manager) 2011-01-15 08:07	The full behavior is documented well above the example that was replaced: Possible expressions are: if(<constant>) True if the constant is 1, ON, YES, TRUE, Y, or a non-zero number. False if the constant is 0, OFF, NO, FALSE, N, IGNORE, "", or ends in the suffix '-NOTFOUND'. Named boolean constants are case-insensitive. If the argument is not one of these constants, it is treated as a variable: if(<variable>) True if the variable is defined to a value that is not a false constant. False otherwise. ... if(DEFINED variable) True if the given variable is defined. It does not matter if the variable is true or false just if it has been set. I don't see how that can be any clearer.

(0024712) Markus Elfring (reporter) 2011-01-15 09:49	I agree that the functionality "if(<variable>)" is documented by itself. But I find that the explanation for the combination with "old style" variable evaluation is still an open issue. Does the "magic" result in any differences in comparison with the current ${} syntax? Modestas Vainius and I have got different expectations about the usable behaviour ...

(0024715) Brad King (manager) 2011-01-15 10:26	> Does the "magic" result in any differences in comparison with the current ${} syntax? The ${} syntax is evaluated before if() sees the arguments. Then the arguments that if() does see are evaluated as specified in the documentation I quoted. Please ignore the old "behave as if you wrote ${}" example that was just plain wrong. Read the documentation from scratch as it appears now. What is missing?

(0024719) Markus Elfring (reporter) 2011-01-15 13:00	Please look at my script example for this issue (again): A variable was set to a value that I would not interpret as the boolean status "true" according to the updated documentation. Why was the message "Good day." displayed?

(0024746) Brad King (manager) 2011-01-17 09:28	In the line if(MY_TEXT) "MY_TEXT" is not one of the named boolean constants. Therefore the command moves on to the "if(<variable>)" form, which is documented as True if the variable is defined to a value that is not a false constant. False otherwise. The value of the variable is "condition-test1.cxx", which is not a false constant. Therefore it is TRUE.

(0024747) Brad King (manager) 2011-01-17 09:47	This makes the signatures where auto-dereferencing occurs more explicit: http://cmake.org/gitweb?p=cmake.git;a=commitdiff;h=7d9b9037 [^]

(0024848) Brad King (manager) 2011-01-18 07:26	Closing as "fixed" since the commit mentioned in 0011701:0024747 clarifies the documentation further.

(0024850) Markus Elfring (reporter) 2011-01-18 07:31	Thanks for your explanation and the fine-tuning of the documentation. I have got the impression that a syntax diagram (or decision tree) might be needed to understand the dependencies of parameter evaluation a bit easier here. http://en.wikipedia.org/wiki/File:Syntax_Diagrams.png [^] Are the displays the same from the execution of the following script example by a couple of CMake versions? cmake_minimum_required(VERSION 2.6) project(my_condition_test2) set(my_text "condition-test2") if(my_text) set(result1 one) else() set(result1 two) endif() if(${my_text}) set(result2 one) else() set(result2 two) endif() if("Bluff") set(result3 one) else() set(result3 two) endif() if( (${result1} STREQUAL ${result2}) AND (${result1} STREQUAL ${result3})) message(STATUS "This result was expected.") else() message(FATAL_ERROR "Results: 1. ${result1} 2. ${result2} 3. ${result3} Check your expectations again!") endif() I get the following log display from CMake-GUI 2.8.3. "CMake Error at CMakeLists.txt:23 (message): Results: 1. one 2. two 3. two Check your expectations again! Configuring incomplete, errors occurred!" How do you think about this situation now? By the way: Does it also show an open issue that blank lines are added to the original output text?

(0024851) Brad King (manager) 2011-01-18 07:54	The results exactly match the documentation! Part 1: "my_text" is a variable defined to a value that is not a false constant -> true Part 2: "${my_text}" becomes "condition-test2" before if() even sees it. The latter is not a defined variable -> false Part 3: "Bluff" is not a defined variable -> false Please, this issue tracker is not a place to ask for tutorials. Go to the mailing list and ask for help there. I don't mean to be dismissive but I've already spent way too much time on this. Only a few people read the issue tracker and we don't have time to give everyone detailed tutorial sessions. There are may people on the mailing list that answer questions. It distributes the load. The documentation is correct and consistent with the implementation and the behavior that we intend. The examples are correct. This issue is resolved. If you open it again without a link to discussion on the mailing list that involves posts other than yours, I will close it without answering again.