This is somewhat obscure, but if you have a switch statement with a structure member as the expression, and you have an undefined identifier in one of your case  statements, the compiler will crash after reporting the error.

Here's code to reproduce the problem (WinXP, DMD 0.108): (sorry about the
formatting, but I haven't figured out how to get the Direct Read Newsreader to
preserve indents).

struct test
{
uint a;
uint b;
}

int main()
{
test tv;
tv.a = 1;

switch( tv.a )
{
case UNDEF_VALUE:
break;
}

return 0;
}


- Geoff Hickey

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

Geoff Hickey schrieb am Fri, 3 Dec 2004 00:52:23 +0000 (UTC):
> This is somewhat obscure, but if you have a switch statement with a structure member as the expression, and you have an undefined identifier in one of your case  statements, the compiler will crash after reporting the error.

This is interesting: if the struct member is an uint the compiler segfaults, if the struct member is an int the compiler doesn't segfault.

Added to DStress as http://svn.kuehne.cn/dstress/nocompile/bug_20041203_A.d http://svn.kuehne.cn/dstress/nocompile/bug_20041203_B.d

Thomas

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.9.12 (GNU/Linux)

iD8DBQFBsGbB3w+/yD4P9tIRAoxZAJ0UjJ98oc7CDaxdENx8wUfleCECPQCeKIdm
pfTQZI3MYHF70am1ke/TQ/k=
=vZN/
-----END PGP SIGNATURE-----

On Fri, 3 Dec 2004 14:14:41 +0100, Thomas Kuehne <thomas-dloop@kuehne.thisisspam.cn> wrote:

>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Geoff Hickey schrieb am Fri, 3 Dec 2004 00:52:23 +0000 (UTC):
>> This is somewhat obscure, but if you have a switch statement with a structure
>> member as the expression, and you have an undefined identifier in one of your
>> case  statements, the compiler will crash after reporting the error.
>
> This is interesting: if the struct member is an uint the compiler
> segfaults, if the struct member is an int the compiler doesn't segfault.
>
> Added to DStress as
> http://svn.kuehne.cn/dstress/nocompile/bug_20041203_A.d
> http://svn.kuehne.cn/dstress/nocompile/bug_20041203_B.d
>
> Thomas
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.9.12 (GNU/Linux)
>
> iD8DBQFBsGbB3w+/yD4P9tIRAoxZAJ0UjJ98oc7CDaxdENx8wUfleCECPQCeKIdm
> pfTQZI3MYHF70am1ke/TQ/k=
> =vZN/
> -----END PGP SIGNATURE-----

Sounds like someone's going to be reducing this... (DMD sure has a lot
of really weird bugs, doesn't it?)

-- 
"Unhappy Microsoft customers have a funny way of becoming Linux,
Salesforce.com and Oracle customers." - www.microsoft-watch.com:
"The Year in Review: Microsoft Opens Up"

"Changes/Additions
--------------------------------------------------------------------------------
HL2DM released! "
- HL2DM changelog
"#$#@% YES!"
- ME

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
NotDashEscaped: You need GnuPG to verify this message


>>> This is somewhat obscure, but if you have a switch statement with a
>>> structure
>>> member as the expression, and you have an undefined identifier in one
>>> of your
>>> case  statements, the compiler will crash after reporting the error.

at least part of the problem can be resolved:

--- expression.org.c    2004-12-03 19:56:15.504405552 +0100
+++ expression.c        2004-12-03 19:56:27.577570152 +0100
@@ -391,7 +391,7 @@
 integer_t Expression::toInteger()
 {
     //printf("Expression %s\n", Token::toChars(op));
-*(char*)0=0;
+    //*(char*)0=0;
     error("Integer constant expression expected instead of %s",
toChars());
     return 0;
 }


Thomas


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.9.12 (GNU/Linux)

iD4DBQFBsLfk3w+/yD4P9tIRAnVWAKCuoq65UEzdaZFHJ9hRhc7Qtsa2uwCSA8iI
wD4R5aXszBupTO0tDKRnFQ==
=yNF6
-----END PGP SIGNATURE-----

It's fixed now. The compiler now puts out:

test.d(14): undefined identifier UNDEF_VALUE
test.d(14): Integer constant expression expected instead of UNDEF_VALUE

as it should.