www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.bugs - struct initializer bug in dmd.124

reply "Ben Hinkle" <ben.hinkle gmail.com> writes:
I haven't narrowed down some reproduction steps but MinTL is busted by I 
think a faulty struct initializer. The statement
Node* x = new Node;
where
struct Node {
  Node* left, right, parent;
  Key key;
  Value value;
}
allocates a Node where left is 0x20 instead of null. The result is 
eventually a seg-v. It works ok when Key,Value are int,int but with 
Key,Value set to char[],char[] it is wrong. Is there a way to look at the 
actual symbol definition in the object file to see what it has for 
_init_5mintl...
May 19 2005
next sibling parent reply "Ben Hinkle" <ben.hinkle gmail.com> writes:
"Ben Hinkle" <ben.hinkle gmail.com> wrote in message 
news:d6j892$2pk9$1 digitaldaemon.com...

I haven't narrowed down some reproduction steps


here are the reproduction steps:
struct Node {
  Node* left,right,parent;
  int color;
  char[] key;
  char[] value;
}
int main() {
  Node* x = new Node;
  assert( x.left is null );
  return 0;
}

Error: AssertError Failure structbug.d(9)
May 19 2005
parent reply "Walter" <newshound digitalmars.com> writes:
"Ben Hinkle" <ben.hinkle gmail.com> wrote in message
news:d6j9fq$2qdd$1 digitaldaemon.com...

 Error: AssertError Failure structbug.d(9)


I can't reproduce on windows. Are you running on an x86 box? The function in
phobos/internal/gc _d_newarrayi may be the culprit if the stack is different
than x86. I'll change it to use std.c.stdarg.
May 20 2005
parent reply "Ben Hinkle" <bhinkle mathworks.com> writes:
"Walter" <newshound digitalmars.com> wrote in message 
news:d6lc4s$1gse$1 digitaldaemon.com...

 "Ben Hinkle" <ben.hinkle gmail.com> wrote in message
 news:d6j9fq$2qdd$1 digitaldaemon.com...

 Error: AssertError Failure structbug.d(9)


 I can't reproduce on windows. Are you running on an x86 box? The function 
 in
 phobos/internal/gc _d_newarrayi may be the culprit if the stack is 
 different
 than x86. I'll change it to use std.c.stdarg.


I was on WinXP box at home. That was after rebuilding phobos as debug so 
that std.boxer would work. I'll look at it again when I get home. I just 
tried the example I posted here at work (WinXP again) and it worked ok both 
with and without a rebuild of phobos so when I get home I'll whack my dmd 
and redo the setup and see if it fixes the problem.
May 20 2005
parent reply "Walter" <newshound digitalmars.com> writes:
"Ben Hinkle" <bhinkle mathworks.com> wrote in message
news:d6lclr$1hbj$1 digitaldaemon.com...

 "Walter" <newshound digitalmars.com> wrote in message
 news:d6lc4s$1gse$1 digitaldaemon.com...

 "Ben Hinkle" <ben.hinkle gmail.com> wrote in message
 news:d6j9fq$2qdd$1 digitaldaemon.com...

 Error: AssertError Failure structbug.d(9)


 I can't reproduce on windows. Are you running on an x86 box? The




function

 in
 phobos/internal/gc _d_newarrayi may be the culprit if the stack is
 different
 than x86. I'll change it to use std.c.stdarg.


 I was on WinXP box at home. That was after rebuilding phobos as debug so
 that std.boxer would work. I'll look at it again when I get home. I just
 tried the example I posted here at work (WinXP again) and it worked ok


both

 with and without a rebuild of phobos so when I get home I'll whack my dmd
 and redo the setup and see if it fixes the problem.


The 0x20 value is instructive. It's the size of the struct, and is pushed on
the stack as the argument to _d_newarrayi just before the ..., which is what
led me to believe it's a stack issue.
May 20 2005
parent "Ben Hinkle" <ben.hinkle gmail.com> writes:
"Walter" <newshound digitalmars.com> wrote in message 
news:d6lfuv$1jqq$1 digitaldaemon.com...

 "Ben Hinkle" <bhinkle mathworks.com> wrote in message
 news:d6lclr$1hbj$1 digitaldaemon.com...

 "Walter" <newshound digitalmars.com> wrote in message
 news:d6lc4s$1gse$1 digitaldaemon.com...

 "Ben Hinkle" <ben.hinkle gmail.com> wrote in message
 news:d6j9fq$2qdd$1 digitaldaemon.com...

 Error: AssertError Failure structbug.d(9)


 I can't reproduce on windows. Are you running on an x86 box? The




 function

 in
 phobos/internal/gc _d_newarrayi may be the culprit if the stack is
 different
 than x86. I'll change it to use std.c.stdarg.


 I was on WinXP box at home. That was after rebuilding phobos as debug so
 that std.boxer would work. I'll look at it again when I get home. I just
 tried the example I posted here at work (WinXP again) and it worked ok


 both

 with and without a rebuild of phobos so when I get home I'll whack my dmd
 and redo the setup and see if it fixes the problem.


 The 0x20 value is instructive. It's the size of the struct, and is pushed 
 on
 the stack as the argument to _d_newarrayi just before the ..., which is 
 what
 led me to believe it's a stack issue.


I couldn't reproduce it on my home machine after resetting my dmd 
installation either. Looks like I had a bogus install. I'm all set now.
May 20 2005
prev sibling next sibling parent "Walter" <newshound digitalmars.com> writes:
"Ben Hinkle" <ben.hinkle gmail.com> wrote in message
news:d6j892$2pk9$1 digitaldaemon.com...

 I haven't narrowed down some reproduction steps but MinTL is busted by I
 think a faulty struct initializer. The statement
 Node* x = new Node;
 where
 struct Node {
   Node* left, right, parent;
   Key key;
   Value value;
 }
 allocates a Node where left is 0x20 instead of null. The result is
 eventually a seg-v. It works ok when Key,Value are int,int but with
 Key,Value set to char[],char[] it is wrong. Is there a way to look at the
 actual symbol definition in the object file to see what it has for
 _init_5mintl...


obj2asm is the right tool for that.
May 19 2005
prev sibling parent Thomas Kuehne <thomas-dloop kuehne.this-is.spam.cn> writes:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ben Hinkle schrieb am Thu, 19 May 2005 19:42:25 -0400:

 I haven't narrowed down some reproduction steps but MinTL is busted by I 
 think a faulty struct initializer. The statement
 Node* x = new Node;
 where
 struct Node {
   Node* left, right, parent;
   Key key;
   Value value;
 }
 allocates a Node where left is 0x20 instead of null. The result is 
 eventually a seg-v. It works ok when Key,Value are int,int but with 
 Key,Value set to char[],char[] it is wrong.


Seems to be a Windows-only bug. Added to DStress as
http://dstress.kuehne.cn/run/s/struct_initialization_07.d
http://dstress.kuehne.cn/run/s/struct_initialization_07_B.d

Thomas


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

iD8DBQFCjdNt3w+/yD4P9tIRAu2eAJ42fJ7+UQdGoTaU35GV3r9H2S3OYQCeL1oA
sMdN6xiQvnqnpcLPTQLFBbU=
=rgKG
-----END PGP SIGNATURE-----
May 20 2005