• Chris Sauls (65/65) Dec 10 2005 Might be a bug, or might be me doing something silly, but this doesn't w...
• Sean Kelly (7/79) Dec 10 2005 I think this is similar to auto variables not getting cleaned up if
• Thomas Kuehne (12/77) Dec 18 2005 -----BEGIN PGP SIGNED MESSAGE-----

Chris Sauls <ibisbasenji gmail.com> writes:

Might be a bug, or might be me doing something silly, but this doesn't work as
expected:
































The output I expected being:
  | + A
  | - A
  | + B
  | - B
  | + C
  | - C
  | + A2
  | - A2
  | + B2
  | - B2
  | + C2
  | - C2

The output I am getting:
  | + A
  | + B
  | + C
  | + A2
  | - A2
  | + B2
  | - B2
  | + C2
  | - C2
  | - C
  | - B
  | - A

Eh?  Apparently the "auto" variables declared by the mixins aren't getting
collected until 
the end of the entire function, my anonymous blocks be damned, and in reverse
order.  And 
in case anyone wonders, no it doesn't change if I place more code in the
blocks.  I ran 
across this little beaut looking for a way to conditionally compile in my Timer
objects 
with somewhat cleaner reading code. (I like the look of `mixin MTimer!("void
main()");` 
better than `version (Timer) auto Timer t = new Timer("void main()");` ... I
guess I'm crazy.)

Any insights?

-- Chris Sauls

Sean Kelly <sean f4.ca> writes:

Chris Sauls wrote:
 Might be a bug, or might be me doing something silly, but this doesn't 
 work as expected:
 






























 
 The output I expected being:
  | + A
  | - A
  | + B
  | - B
  | + C
  | - C
  | + A2
  | - A2
  | + B2
  | - B2
  | + C2
  | - C2
 
 The output I am getting:
  | + A
  | + B
  | + C
  | + A2
  | - A2
  | + B2
  | - B2
  | + C2
  | - C2
  | - C
  | - B
  | - A
 
 Eh?  Apparently the "auto" variables declared by the mixins aren't 
 getting collected until the end of the entire function, my anonymous 
 blocks be damned, and in reverse order.  And in case anyone wonders, no 
 it doesn't change if I place more code in the blocks.  I ran across this 
 little beaut looking for a way to conditionally compile in my Timer 
 objects with somewhat cleaner reading code. (I like the look of `mixin 
 MTimer!("void main()");` better than `version (Timer) auto Timer t = new 
 Timer("void main()");` ... I guess I'm crazy.)
 
 Any insights?

I think this is similar to auto variables not getting cleaned up if 
scope exits from a throw.  In both cases, the dtor isn't being called as 
it should and they're being cleaned up by the next GC cycle.  In the 
case of mixins however, it seems like a parse issue rather than a 
runtime issue.


Sean

Thomas Kuehne <thomas-dloop kuehne.cn> writes:

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

Chris Sauls schrieb am 2005-12-10:
 Might be a bug, or might be me doing something silly, but this doesn't work as
expected:
































 The output I expected being:
  | + A
  | - A
  | + B
  | - B
  | + C
  | - C
  | + A2
  | - A2
  | + B2
  | - B2
  | + C2
  | - C2

 The output I am getting:
  | + A
  | + B
  | + C
  | + A2
  | - A2
  | + B2
  | - B2
  | + C2
  | - C2
  | - C
  | - B
  | - A

 Eh?  Apparently the "auto" variables declared by the mixins aren't getting
collected until 
 the end of the entire function, my anonymous blocks be damned, and in reverse
order.  And 
 in case anyone wonders, no it doesn't change if I place more code in the
blocks.  I ran 
 across this little beaut looking for a way to conditionally compile in my
Timer objects 
 with somewhat cleaner reading code. (I like the look of `mixin MTimer!("void
main()");` 
 better than `version (Timer) auto Timer t = new Timer("void main()");` ... I
guess I'm crazy.)

 Any insights?

 -- Chris Sauls

Added to DStress as
http://dstress.kuehne.cn/run/a/auto_13_A.d
http://dstress.kuehne.cn/run/a/auto_13_B.d

Thomas


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

iD8DBQFDpgR53w+/yD4P9tIRAmEWAKDSygm+HzdO8DOMCWi80rjJ35Ss0ACgrylU
kQb6vP4rIDdHKnVZg3Vteko=
=tj/u
-----END PGP SIGNATURE-----