www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Quasi debug mode

reply "Chris Miller" <chris dprogramming.com> writes:
DMD has a quasi debug mode when you don't use -debug and -release.


import std.stdio;
int main()
{
    debug
    {
       writefln("debug");
       assert("test" is null);
    }
    else
    {
       writefln("release?");
       assert("test" is null);
    }
    return 0;
}


Compiled 3 times:

"dmd test" outputs:
release?
Error: AssertError Failure test(12)

"dmd -debug test" outputs:
debug
Error: AssertError Failure test(7)

"dmd -release test" outputs:
release?


Notice the first one (quasi debug) not saying debug, yet it does asserts.


There are at least 2 problems with this quasi debug mode:

    1) If you wish to distribute lib files for debug and release, you will  
actually have to distribute 3 due to this quasi debug mode. If you use  
either a debug lib or release lib with the quasi debug mode, things  
sometimes don't line up when linking. (phobos:std.boxer is home proof).

    2) There's no way to know if it's a release compilation using version  
or debug statements.


The best move would probably be to make -debug default unless -release is  
specified. It might be nice to add a version(release) but won't be  
necessary if debug{}else{} worked.

- Chris
Jan 07 2006
next sibling parent reply "Ameer Armaly" <ameer_armaly hotmail.com> writes:
"Chris Miller" <chris dprogramming.com> wrote in message 
news:op.s205x4rxpo9bzi moe...

 DMD has a quasi debug mode when you don't use -debug and -release.


 import std.stdio;
 int main()
 {
    debug
    {
       writefln("debug");
       assert("test" is null);
    }
    else
    {
       writefln("release?");
       assert("test" is null);
    }
    return 0;
 }


 Compiled 3 times:

 "dmd test" outputs:
 release?
 Error: AssertError Failure test(12)

 "dmd -debug test" outputs:
 debug
 Error: AssertError Failure test(7)

 "dmd -release test" outputs:
 release?


 Notice the first one (quasi debug) not saying debug, yet it does asserts.


 There are at least 2 problems with this quasi debug mode:

    1) If you wish to distribute lib files for debug and release, you will 
 actually have to distribute 3 due to this quasi debug mode. If you use 
 either a debug lib or release lib with the quasi debug mode, things 
 sometimes don't line up when linking. (phobos:std.boxer is home proof).

    2) There's no way to know if it's a release compilation using version 
 or debug statements.


 The best move would probably be to make -debug default unless -release is 
 specified. It might be nice to add a version(release) but won't be 
 necessary if debug{}else{} worked.


I don't know; I do understand the problem with not beeing able to 
differentiate between -debug and no switch, but I do see some uses for 
having the no switch option.  The way I see it, -debug is to turn on 
optional debugging functionality, and -release is to take out *all* of it. 
If you don't specify either, you're basically asking for the default 
package; that which does what is considered "standard" testing, without 
having to necessarily give you verbose reports or whatever you have it do 
in -debug mode.

 - Chris
Jan 07 2006
parent reply =?ISO-8859-1?Q?Anders_F_Bj=F6rklund?= <afb algonet.se> writes:
Ameer Armaly wrote:


 The way I see it, -debug is to turn on optional debugging
 functionality, and -release is to take out *all* of it.


In D, -release doesn't affect debugging but it does affect contracts...

The confusion comes from:
"Two versions of programs are commonly built,
a release build and a debug build. The debug
build includes extra error checking code,
test harnesses, pretty-printing code, etc."
(http://www.digitalmars.com/d/version.html)

versus:
"-release
     compile release version, which means not
     generating code for contracts and asserts"
(http://www.digitalmars.com/d/dcompiler.html)

Some people think it would have been better named "-nocontracts" or so.


This in turn comes from an age-old debate whether asserts* should be
left in the release builds or not. C/C++/D tends to side with "not":
"If I wanted runtime assertions in my release code, I'd use Java" :-P
(as Java has no means of stripping out assertions or bounds checking)


I believe that Matthew is running an article series on the subject ?
http://www.artima.com/cppsource/deepspace.html

By naming the option "release", I think Walter's position is clear :-)
(as -release in D is the equivalent of -DNDEBUG in C, i.e. no asserts)

--anders

* assuming here that "assert" is the basic form of a code contract
Jan 07 2006
parent "Ameer Armaly" <ameer_armaly hotmail.com> writes:
"Anders F Björklund" <afb algonet.se> wrote in message 
news:dppq61$2h50$1 digitaldaemon.com...

 Ameer Armaly wrote:


 The way I see it, -debug is to turn on optional debugging
 functionality, and -release is to take out *all* of it.


 In D, -release doesn't affect debugging but it does affect contracts...

 The confusion comes from:
 "Two versions of programs are commonly built,
 a release build and a debug build. The debug
 build includes extra error checking code,
 test harnesses, pretty-printing code, etc."
 (http://www.digitalmars.com/d/version.html)

 versus:
 "-release
     compile release version, which means not
     generating code for contracts and asserts"
 (http://www.digitalmars.com/d/dcompiler.html)

 Some people think it would have been better named "-nocontracts" or so.


 This in turn comes from an age-old debate whether asserts* should be
 left in the release builds or not. C/C++/D tends to side with "not":
 "If I wanted runtime assertions in my release code, I'd use Java" :-P
 (as Java has no means of stripping out assertions or bounds checking)


 I believe that Matthew is running an article series on the subject ?
 http://www.artima.com/cppsource/deepspace.html

 By naming the option "release", I think Walter's position is clear :-)
 (as -release in D is the equivalent of -DNDEBUG in C, i.e. no asserts)


That's what I'm saying; -debug implies you want maximum checking 
done, -release says you want none of it, and nothing just means you want 
those checks not designated as "debug."  Personally I find it very useful.

 --anders

 * assuming here that "assert" is the basic form of a code contract
Jan 07 2006
prev sibling parent =?UTF-8?B?QW5kZXJzIEYgQmrDtnJrbHVuZA==?= <afb algonet.se> writes:
Chris Miller wrote:


 DMD has a quasi debug mode when you don't use -debug and -release.


I usually call this the D "contracts" mode, myself.

It's useful to have such a Phobos version around...
(If you use my RPM, such a beast is in a subpackage)

The default Phobos doesn't show contract violations.


-release is just a short-hand for "strip asserts / contracts" :
(if you want to optimize, you need to look at -O and -inline)

     if (global.params.release)
     {	global.params.useInvariants = 0;
	global.params.useIn = 0;
	global.params.useOut = 0;
	global.params.useAssert = 0;
	global.params.useArrayBounds = 0;
	global.params.useSwitchError = 0;
     }

Kinda confusing though. Catches all newcomers to D, for sure...


Just guessing, you'd think that -debug and -release are related ?

But -debug is part of the version-system*, and it has variants :
-debug=string, -debug=zlib, or whatever that you want to debug...

And it does *not* add the debugging symbols, for that you use -g.

--anders

* see http://www.digitalmars.com/d/version.html#debug :
   "The debug statement conditionally compiles in its statement body.
    It is D's way of what in C is done with #ifdef DEBUG / #endif pairs."
Jan 07 2006