This requires two modules.

    [a.d]
    struct TC(T)
    {
        void method()
        {
            void inline_function()
            {
            }
        }
    }

    template TB(T)
    {
        alias TC!(T) tc_instance;
    }

    struct TA(T)
    {
        mixin TB!(T);
    }

    [b.d]
    import a;

    void main()
    {
        TA!(int) variable;
    }

Attempting to compile this causes the compiler to do an access violation; it should be acceptable code.  The following are variants known to cause the problem to disappear.  First variant, to remove the inline function:

    [a.d]
    struct TC(T)
    {
        void method()
        {
--            void inline_function()
--            {
--            }
        }
    }

    template TB(T)
    {
        alias TC!(T) tc_instance;
    }

    struct TA(T)
    {
        mixin TB!(T);
    }

    [b.d]
    import a;

    void main()
    {
        TA!(int) variable;
    }

Second variant, to remove an abstraction stage:

    [a.d]
    struct TC(T)
    {
        void method()
        {
            void inline_function()
            {
            }
        }
    }

--    template TB(T)
--    {
--        alias TC!(T) tc_instance;
--    }

    struct TA(T)
    {
--        mixin TB!(T);
++        alias TC!(T) tc_instance;
    }

    [b.d]
    import a;

    void main()
    {
        TA!(int) variable;
    }

Third variant, instantiate the template outside of the "main" function:

    [a.d]
    struct TC(T)
    {
        void method()
        {
            void inline_function()
            {
            }
        }
    }

    template TB(T)
    {
        alias TC!(T) tc_instance;
    }

    struct TA(T)
    {
        mixin TB!(T);
    }

    [b.d]
    import a;

++  alias TA!(int) TA_int;

    void main()
    {
--        TA!(int) variable;
++        TA_int variable;
    }

I couldn't reproduce the fault, but I put it in the test suite anyway.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Walter schrieb am Wed, 18 May 2005 22:52:41 -0700:
> I couldn't reproduce the fault, but I put it in the test suite anyway.
>
>

(gdb) run -v a.d b.d
Starting program: dmd-0.123 -v a.d b.d
parse     a
parse     b
semantic  a
semantic  b
semantic2 a
semantic2 b
semantic3 a
semantic3 b
code      a
generating code for function 'method'

Program received signal SIGSEGV, Segmentation fault.
0x080a9d16 in FuncDeclaration::toObjFile() ()
(gdb) bt
#0  0x080a9d16 in FuncDeclaration::toObjFile() ()
#1  0x080aa192 in FuncDeclaration::toObjFile() ()
#2  0x080e358e in StructDeclaration::toObjFile() ()
#3  0x080dadfa in TemplateInstance::toObjFile() ()
#4  0x080a9ce9 in Module::genobjfile() ()
#5  0x080b814b in main ()

Thomas


-----BEGIN PGP SIGNATURE-----

iD8DBQFCjFR+3w+/yD4P9tIRAkN6AKCvNyVgjHJdpne+ldeqzCueVM2uLwCeKLP0
37AVAbol2yLhZm4Ssim9GTI=
=r3O2
-----END PGP SIGNATURE-----

"Thomas Kuehne" <thomas-dloop@kuehne.this-is.spam.cn> wrote in message news:d8ol3u.43o.thomas-dloop@laermschleuder.kuehne.cn...
> Program received signal SIGSEGV, Segmentation fault.

Doesn't happen for me, either on windows or linux.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Walter schrieb am Thu, 19 May 2005 14:46:00 -0700:
>
> "Thomas Kuehne" <thomas-dloop@kuehne.this-is.spam.cn> wrote in message news:d8ol3u.43o.thomas-dloop@laermschleuder.kuehne.cn...
>> Program received signal SIGSEGV, Segmentation fault.
>
> Doesn't happen for me, either on windows or linux.
>

The files are http://dstress.kuehne.cn/run/b/bug_20050510_A2.d http://dstress.kuehne.cn/compile/b/bug_20050510_A1.d

no bug:
dmd run/b/bug_20050510_A2.d compile/b/bug_20050510_A1.d

bug:
dmd compile/b/bug_20050510_A1.d run/b/bug_20050510_A2.d

Thomas

-----BEGIN PGP SIGNATURE-----

iD8DBQFCjS6v3w+/yD4P9tIRAmIrAJ9qgxh72kss+rOMpjjVdorC0AgiBgCeML3L
Akf2qcyY9dE5wEHstM+mx7w=
=tNEF
-----END PGP SIGNATURE-----