www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Assertion failure: '!cases' on line 2620 in file 'statement.c'

reply %u <e ee.com> writes:
Assertion failure: '!cases' on line 2620 in file 'statement.c'

This is what I got from trying to build my project(30+ modules) without bud.
Jan 12 2011
parent reply %u <e ee.com> writes:
Should I post it as a bug, even though I have no code to accompany it?
I have no clue as to where to start my directed search for a minimal case.
Jan 12 2011
next sibling parent "Simen kjaeraas" <simen.kjaras gmail.com> writes:
%u <e ee.com> wrote:


 Should I post it as a bug, even though I have no code to accompany it?
 I have no clue as to where to start my directed search for a minimal  
 case.


Please do post it, yes. As for searching, if you are willing to share
the code , others may be willing to do the search.

-- 
Simen
Jan 12 2011
prev sibling parent reply Don <nospam nospam.com> writes:
%u wrote:

 Should I post it as a bug, even though I have no code to accompany it?
 I have no clue as to where to start my directed search for a minimal case.


Can you post the entire source code?
It's important that it be reproducible. It doesn't need to be minimal - 
someone else can reduce it.
Jan 12 2011
parent reply %u <e ee.com> writes:
== Quote from Don (nospam nospam.com)'s article

 %u wrote:

 Should I post it as a bug, even though I have no code to accompany it?
 I have no clue as to where to start my directed search for a minimal case.


 Can you post the entire source code?
 It's important that it be reproducible. It doesn't need to be minimal -
 someone else can reduce it.


I crashed it again :)

Sorry, can't share the code..

I don't really have any time atm, look at it again tomorrow.
Jan 12 2011
parent reply Don <nospam nospam.com> writes:
%u wrote:

 == Quote from Don (nospam nospam.com)'s article

 %u wrote:

 Should I post it as a bug, even though I have no code to accompany it?
 I have no clue as to where to start my directed search for a minimal case.


 Can you post the entire source code?
 It's important that it be reproducible. It doesn't need to be minimal -
 someone else can reduce it.


 
 I crashed it again :)
 
 Sorry, can't share the code..


Bummer.


 I don't really have any time atm, look at it again tomorrow.


It's in a switch statement somewhere.
It sounds as though this is a bug which involves multiple files, so 
it'll be difficult to reduce it.

If you're able to compile DMD, change this line in statement.c line 2620:

Statement *SwitchStatement::semantic(Scope *sc)
{
     //printf("SwitchStatement::semantic(%p)\n", this);
     tf = sc->tf;
+   if (cases) error("xxx");
     assert(!cases);             // ensure semantic() is only run once


and then you'll get the line number where the error is.
If you can provide the function which contains the switch statement, 
there's a chance I could reproduce it.
Jan 13 2011
parent reply %u <e ee.com> writes:
== Quote from Don (nospam nospam.com)'s article

 It's in a switch statement somewhere.
 It sounds as though this is a bug which involves multiple files, so
 it'll be difficult to reduce it.
 If you're able to compile DMD, change this line in statement.c line 2620:
 Statement *SwitchStatement::semantic(Scope *sc)
 {
      //printf("SwitchStatement::semantic(%p)\n", this);
      tf = sc->tf;
 +   if (cases) error("xxx");
      assert(!cases);             // ensure semantic() is only run once
 and then you'll get the line number where the error is.


Yay for first time compiling dmd :)


 If you can provide the function which contains the switch statement,
 there's a chance I could reproduce it.


I've got something better.. a minimal version :)
Which even crashes through bud.

----
module main;

enum E { A = 0 };

struct S{
  int i;

  static S func( S s, E e ){
    switch( e ) //<---------- here
    {
      default:return s;
    }
  }

  static const S s_def = { 1 };
  //static const S A = func(s_def, E.A ); // forward reference error + crash
  //static const S[1] ARR = [ E.A : func(s_def, E.A )]; // error : xxx + crash
}

void main(){}
----
To test all this I switched from 1.065 to 1.066; just to make sure it hadn't
been
fixed already.
And now my project won't compile any more even though bud+1.065 will happily do
so..
bud+ 1.066 gives me the following (no crash though)



























Internal error: backend\cgobj.c 2424

What does that mean?
Jan 13 2011
parent reply Don <nospam nospam.com> writes:
%u wrote:

 == Quote from Don (nospam nospam.com)'s article

 It's in a switch statement somewhere.
 It sounds as though this is a bug which involves multiple files, so
 it'll be difficult to reduce it.
 If you're able to compile DMD, change this line in statement.c line 2620:
 Statement *SwitchStatement::semantic(Scope *sc)
 {
      //printf("SwitchStatement::semantic(%p)\n", this);
      tf = sc->tf;
 +   if (cases) error("xxx");
      assert(!cases);             // ensure semantic() is only run once
 and then you'll get the line number where the error is.


 
 Yay for first time compiling dmd :)


Sorry you had to do that!


 

 If you can provide the function which contains the switch statement,
 there's a chance I could reproduce it.


 
 I've got something better.. a minimal version :)
 Which even crashes through bud.


Awesome!
I'll track this sucker down so it doesn't hit anyone else.


 
 ----
 module main;
 
 enum E { A = 0 };
 
 struct S{
   int i;
 
   static S func( S s, E e ){
     switch( e ) //<---------- here
     {
       default:return s;
     }
   }
 
   static const S s_def = { 1 };
   //static const S A = func(s_def, E.A ); // forward reference error + crash
   //static const S[1] ARR = [ E.A : func(s_def, E.A )]; // error : xxx + crash
 }
 
 void main(){}
 ----
 To test all this I switched from 1.065 to 1.066; just to make sure it hadn't
been
 fixed already.
 And now my project won't compile any more even though bud+1.065 will happily
do so..
 bud+ 1.066 gives me the following (no crash though)
 


























 Internal error: backend\cgobj.c 2424
 
 What does that mean?


The DMD makefile compiles in debug mode by default. It prints some 
useless 'fixups' junk. You need to compile with make -fwin32.mak release
or make -flinux.mak release
to make a release compiler. But, not important, you can go back to using 
an official compiler again. <g>
The internal error means it's now crashing in the backend, rather than 
the frontend. Not an improvement!
Jan 13 2011
next sibling parent Andrej Mitrovic <andrej.mitrovich gmail.com> writes:
On 1/13/11, Don <nospam nospam.com> wrote:

 The DMD makefile compiles in debug mode by default. It prints some
 useless 'fixups' junk. You need to compile with make -fwin32.mak release
 or make -flinux.mak release
 to make a release compiler.


Ugh, I just realized I was using DMD in debug mode by mistake for the
last couple of weeks. Thanks for the info (I should have looked this
up).
Jan 13 2011
prev sibling parent reply %u <e ee.com> writes:
== Quote from Don (nospam nospam.com)'s article

 Yay for first time compiling dmd :)


 Sorry you had to do that!


Had to learn that once anyway :)
Maybe I'll even be able to take a stab at fixing bugs someday..
Jan 13 2011
parent Don <nospam nospam.com> writes:
%u wrote:

 == Quote from Don (nospam nospam.com)'s article

 Yay for first time compiling dmd :)


 Sorry you had to do that!


 
 Had to learn that once anyway :)
 Maybe I'll even be able to take a stab at fixing bugs someday..


<g>
Added your bug as:

http://d.puremagic.com/issues/show_bug.cgi?id=5453
Jan 14 2011