Compiling code that worked fine with DMD 0.160 now fails on DMD 0.161 with the following compiler error:
   Assertion failure: 'next' on line 2226 in file 'mtype.c'

Looking at mtype.c I notice the assert makes sure 'next' is not null, but on the next line it checks 'next' for null and acts appropriately; also, in the same function of DMD 0.160 this null check is not there, only the assert, so it leads me to believe that perhaps the new compiler now allows a null 'next' but the assert was left accidentally.

I tried simplifying my code to make a small reproducable example, but this project is huge and the changes were causing template mixin forward reference errors. Compiling with -v shows it fails on the first semantic pass.

- Chris

Chris Miller wrote:
> Compiling code that worked fine with DMD 0.160 now fails on DMD 0.161 with the following compiler error:
>    Assertion failure: 'next' on line 2226 in file 'mtype.c'
> 
> Looking at mtype.c I notice the assert makes sure 'next' is not null, but on the next line it checks 'next' for null and acts appropriately; also, in the same function of DMD 0.160 this null check is not there, only the assert, so it leads me to believe that perhaps the new compiler now allows a null 'next' but the assert was left accidentally.
> 
> I tried simplifying my code to make a small reproducable example, but this project is huge and the changes were causing template mixin forward reference errors. Compiling with -v shows it fails on the first semantic pass.

There's no way to track this down without a reproducible test case.

Walter Bright wrote:
> Chris Miller wrote:
>> Compiling code that worked fine with DMD 0.160 now fails on DMD 0.161
>> with the following compiler error:
>>    Assertion failure: 'next' on line 2226 in file 'mtype.c'
>>
>> Looking at mtype.c I notice the assert makes sure 'next' is not null, but on the next line it checks 'next' for null and acts appropriately; also, in the same function of DMD 0.160 this null check is not there, only the assert, so it leads me to believe that perhaps the new compiler now allows a null 'next' but the assert was left accidentally.
>>
>> I tried simplifying my code to make a small reproducable example, but this project is huge and the changes were causing template mixin forward reference errors. Compiling with -v shows it fails on the first semantic pass.
> 
> There's no way to track this down without a reproducible test case.

Have a look at Carlos Santander's post "ice returning a delegate".

Walter Bright wrote:
> Chris Miller wrote:
>> Compiling code that worked fine with DMD 0.160 now fails on DMD 0.161 with the following compiler error:
>>    Assertion failure: 'next' on line 2226 in file 'mtype.c'
>>
>> Looking at mtype.c I notice the assert makes sure 'next' is not null, but on the next line it checks 'next' for null and acts appropriately; also, in the same function of DMD 0.160 this null check is not there, only the assert, so it leads me to believe that perhaps the new compiler now allows a null 'next' but the assert was left accidentally.
>>
>> I tried simplifying my code to make a small reproducable example, but this project is huge and the changes were causing template mixin forward reference errors. Compiling with -v shows it fails on the first semantic pass.
> 
> There's no way to track this down without a reproducible test case.


  template tpl() {
    typeof( {} ) foo;
  }

  int main(char[][] args)
  {
    tpl!();
    return 0;
  }

» Assertion failure: 'next' on line 2226 in file 'mtype.c'

Not sure if this is the same as bug 220, since that one asserts on a different line (altough on the same file), and the test case seems different.


-- 
Bruno Medeiros - CS/E student
http://www.prowiki.org/wiki4d/wiki.cgi?BrunoMedeiros#D

Bruno Medeiros schrieb am 2006-06-23:
>

[snip]

>    template tpl() {
>      typeof( {} ) foo;
>    }
>
>    int main(char[][] args)
>    {
>      tpl!();
>      return 0;
>    }
>
> » Assertion failure: 'next' on line 2226 in file 'mtype.c'
>
> Not sure if this is the same as bug 220, since that one asserts on a different line (altough on the same file), and the test case seems different.

Added to DStress as http://dstress.kuehne.cn/run/b/bug_mtype_2226_A.d http://dstress.kuehne.cn/run/b/bug_mtype_2226_B.d

Thomas

Thomas Kuehne wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Bruno Medeiros schrieb am 2006-06-23:
> 
> [snip]
> 
>>    template tpl() {
>>      typeof( {} ) foo;
>>    }
>>
>>    int main(char[][] args)
>>    {
>>      tpl!();
>>      return 0;
>>    }
>>
>> � Assertion failure: 'next' on line 2226 in file 'mtype.c'
>>
>> Not sure if this is the same as bug 220, since that one asserts on a different line (altough on the same file), and the test case seems different.
> 
> Added to DStress as
> http://dstress.kuehne.cn/run/b/bug_mtype_2226_A.d
> http://dstress.kuehne.cn/run/b/bug_mtype_2226_B.d
> 
> Thomas
> 
> 
> -----BEGIN PGP SIGNATURE-----
> 
> iD8DBQFEo7ja3w+/yD4P9tIRApoZAKCkzviTOmEau6ZnPqYwlHVKtnwqfACfbqbh
> apH8t/ytCqHn2K2//Ens2JU=
> =oPTq
> -----END PGP SIGNATURE-----

Seems this was fixed in DMD 162.

-- 
Bruno Medeiros - CS/E student
http://www.prowiki.org/wiki4d/wiki.cgi?BrunoMedeiros#D