www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Howto build a repo case for a ICE?

reply Benjamin Thaut <code benjamin-thaut.de> writes:
The last dmd version I could use to actually build my project was dmd 
2.056. Now I want to help finding those bugs that keep me from compiling 
and I'm wondering how I can build repro cases for the ICEs I encounter. 
I tried compiling with -v but that does not really give much information 
about where the crash actually happens. So I wonder what can actually be 
done to find the part of the code or even only the module it crashes in?

The current ICE I'm encountering is:
Assertion failure: 't1->ty == t2->ty' on line 2082 in file 'cast.c'

Kind Regards
Benjamin Thaut
Feb 19 2012
next sibling parent "Daniel Murphy" <yebblies nospamgmail.com> writes:
There are only a few options:

- Reduce your code using dustmite.  I've never tried this, but I hear it 
works.
- Use a debug version of dmd, or post the code and hope someone else debugs 
it.
- Reduce the code manually.

The error you're getting occurs when merging two integral types together, 
and something goes horribly wrong.  It shouldn't be happening, but I'd guess 
it's something to do with alias this with integral types (huge guess).  How 
big is the project?

"Benjamin Thaut" <code benjamin-thaut.de> wrote in message 
news:jhqt5p$1uel$1 digitalmars.com...

 The last dmd version I could use to actually build my project was dmd 
 2.056. Now I want to help finding those bugs that keep me from compiling 
 and I'm wondering how I can build repro cases for the ICEs I encounter. I 
 tried compiling with -v but that does not really give much information 
 about where the crash actually happens. So I wonder what can actually be 
 done to find the part of the code or even only the module it crashes in?

 The current ICE I'm encountering is:
 Assertion failure: 't1->ty == t2->ty' on line 2082 in file 'cast.c'

 Kind Regards
 Benjamin Thaut
Feb 19 2012
prev sibling next sibling parent reply Timon Gehr <timon.gehr gmx.ch> writes:
On 02/19/2012 02:24 PM, Benjamin Thaut wrote:

 The last dmd version I could use to actually build my project was dmd
 2.056. Now I want to help finding those bugs that keep me from compiling
 and I'm wondering how I can build repro cases for the ICEs I encounter.
 I tried compiling with -v but that does not really give much information
 about where the crash actually happens. So I wonder what can actually be
 done to find the part of the code or even only the module it crashes in?

 The current ICE I'm encountering is:
 Assertion failure: 't1->ty == t2->ty' on line 2082 in file 'cast.c'

 Kind Regards
 Benjamin Thaut



DustMite is an useful tool for generating reduced cases that cause the 
same ICE:

https://github.com/CyberShadow/DustMite
Feb 19 2012
next sibling parent reply Benjamin Thaut <code benjamin-thaut.de> writes:
Am 19.02.2012 14:44, schrieb Timon Gehr:

 On 02/19/2012 02:24 PM, Benjamin Thaut wrote:

 The last dmd version I could use to actually build my project was dmd
 2.056. Now I want to help finding those bugs that keep me from compiling
 and I'm wondering how I can build repro cases for the ICEs I encounter.
 I tried compiling with -v but that does not really give much information
 about where the crash actually happens. So I wonder what can actually be
 done to find the part of the code or even only the module it crashes in?

 The current ICE I'm encountering is:
 Assertion failure: 't1->ty == t2->ty' on line 2082 in file 'cast.c'

 Kind Regards
 Benjamin Thaut



 DustMite is an useful tool for generating reduced cases that cause the
 same ICE:

 https://github.com/CyberShadow/DustMite



Which dmd version should I use to build DustMite?
I tried 2.055 and 2.058 but DustMite crashes half way through the 
reductino process without a message (I'm using windows)
Feb 19 2012
next sibling parent reply Benjamin Thaut <code benjamin-thaut.de> writes:
Am 19.02.2012 17:06, schrieb Vladimir Panteleev:

 On Sunday, 19 February 2012 at 15:31:32 UTC, Benjamin Thaut wrote:

 Which dmd version should I use to build DustMite?
 I tried 2.055 and 2.058 but DustMite crashes half way through the
 reductino process without a message (I'm using windows)



 I've never heard of a problem like this. Is the DustMite crash
 reproducible?



Yes, it never crashes at the extact same reduction step, but it always 
crashes somewhere in between. I assume that it is a segfault because it 
does not print a stack trace. I could however only test it on 1 machine 
so it could be machine specific.
Feb 19 2012
parent Benjamin Thaut <code benjamin-thaut.de> writes:
Am 19.02.2012 18:25, schrieb Vladimir Panteleev:

 On Sunday, 19 February 2012 at 16:11:35 UTC, Benjamin Thaut wrote:

 Am 19.02.2012 17:06, schrieb Vladimir Panteleev:

 On Sunday, 19 February 2012 at 15:31:32 UTC, Benjamin Thaut wrote:

 Which dmd version should I use to build DustMite?
 I tried 2.055 and 2.058 but DustMite crashes half way through the
 reductino process without a message (I'm using windows)



 I've never heard of a problem like this. Is the DustMite crash
 reproducible?



 Yes, it never crashes at the extact same reduction step, but it always
 crashes somewhere in between. I assume that it is a segfault because
 it does not print a stack trace. I could however only test it on 1
 machine so it could be machine specific.



 I'd appreciate it if you could gather more information about the crash
 (for example, running in in a debugger so you can get the backtrace of
 the segfault, if there is one).



rmdirRecurse which gets called in dustmite.d line 457 throws a exception 
that causes the D exception handler to go into inifnite recursion which 
leads to a stack overflow and crashes the program.
Might be a issue with my file system. Going to do a file system check.

Kind Regards
Benjamin Thaut
Feb 19 2012
prev sibling parent reply =?UTF-8?B?U8O2bmtlIEx1ZHdpZw==?= <ludwig informatik.uni-luebeck.de> writes:
Am 19.02.2012 17:06, schrieb Vladimir Panteleev:

 On Sunday, 19 February 2012 at 15:31:32 UTC, Benjamin Thaut wrote:

 Which dmd version should I use to build DustMite?
 I tried 2.055 and 2.058 but DustMite crashes half way through the
 reductino process without a message (I'm using windows)



 I've never heard of a problem like this. Is the DustMite crash
 reproducible?



I've made the experience that it is more-or-less impossible to run 
dustmite on windows for longer time spans. The cause is most probably 
Windows Explorer, which sometimes accesses and locks files.

In a project we wrote a limited retry-loop around the critical file 
system calls to get stable runs...
Feb 19 2012
parent Benjamin Thaut <code benjamin-thaut.de> writes:
Am 20.02.2012 01:40, schrieb Vladimir Panteleev:

 On Sunday, 19 February 2012 at 22:46:29 UTC, Sönke Ludwig wrote:

 Am 19.02.2012 17:06, schrieb Vladimir Panteleev:

 On Sunday, 19 February 2012 at 15:31:32 UTC, Benjamin Thaut wrote:

 Which dmd version should I use to build DustMite?
 I tried 2.055 and 2.058 but DustMite crashes half way through the
 reductino process without a message (I'm using windows)



 I've never heard of a problem like this. Is the DustMite crash
 reproducible?



 I've made the experience that it is more-or-less impossible to run
 dustmite on windows for longer time spans. The cause is most probably
 Windows Explorer, which sometimes accesses and locks files.

 In a project we wrote a limited retry-loop around the critical file
 system calls to get stable runs...



 I've never encountered such problems. DustMite can run for days on end
 without issues.



If I run it inside my C:\Users\<my username>\Documents directory it does 
not wrong very long. 10 mins max before it crashes. Now I put my 
sourcecode directly in the root of a partition F:\ and it works just 
fine without crashing once in hours.

Kind Regards
Benjamin Thaut
Feb 19 2012
prev sibling next sibling parent "Vladimir Panteleev" <vladimir thecybershadow.net> writes:
On Sunday, 19 February 2012 at 15:31:32 UTC, Benjamin Thaut wrote:

 Which dmd version should I use to build DustMite?
 I tried 2.055 and 2.058 but DustMite crashes half way through 
 the reductino process without a message (I'm using windows)



I've never heard of a problem like this. Is the DustMite crash 
reproducible?
Feb 19 2012
prev sibling next sibling parent "Vladimir Panteleev" <vladimir thecybershadow.net> writes:
On Sunday, 19 February 2012 at 16:11:35 UTC, Benjamin Thaut wrote:

 Am 19.02.2012 17:06, schrieb Vladimir Panteleev:

 On Sunday, 19 February 2012 at 15:31:32 UTC, Benjamin Thaut 
 wrote:

 Which dmd version should I use to build DustMite?
 I tried 2.055 and 2.058 but DustMite crashes half way through 
 the
 reductino process without a message (I'm using windows)



 I've never heard of a problem like this. Is the DustMite crash
 reproducible?



 Yes, it never crashes at the extact same reduction step, but it 
 always crashes somewhere in between. I assume that it is a 
 segfault because it does not print a stack trace. I could 
 however only test it on 1 machine so it could be machine 
 specific.



I'd appreciate it if you could gather more information about the 
crash (for example, running in in a debugger so you can get the 
backtrace of the segfault, if there is one).
Feb 19 2012
prev sibling next sibling parent "Vladimir Panteleev" <vladimir thecybershadow.net> writes:
On Sunday, 19 February 2012 at 17:56:42 UTC, Benjamin Thaut wrote:

 rmdirRecurse which gets called in dustmite.d line 457 throws a 
 exception that causes the D exception handler to go into 
 inifnite recursion which leads to a stack overflow and crashes 
 the program.
 Might be a issue with my file system. Going to do a file system 
 check.



One of the reasons that rmdirRecurse might fail is if a program 
has opened the directory or one of the files. This may be caused 
by trying to peek at Dustmite's results while it's running 
without pausing it, or perhaps antivirus/antimalware software.
Feb 19 2012
prev sibling next sibling parent "Vladimir Panteleev" <vladimir thecybershadow.net> writes:
On Sunday, 19 February 2012 at 17:56:42 UTC, Benjamin Thaut wrote:

 rmdirRecurse which gets called in dustmite.d line 457 throws a 
 exception that causes the D exception handler to go into 
 inifnite recursion which leads to a stack overflow and crashes 
 the program.



Thanks for taking the time to figure this out. I'll make DustMite 
catch rmdirRecurse failures and retry the operation (as the 
causes are usually temporary).
Feb 19 2012
prev sibling next sibling parent "Vladimir Panteleev" <vladimir thecybershadow.net> writes:
On Sunday, 19 February 2012 at 22:46:29 UTC, Sönke Ludwig wrote:

 Am 19.02.2012 17:06, schrieb Vladimir Panteleev:

 On Sunday, 19 February 2012 at 15:31:32 UTC, Benjamin Thaut 
 wrote:

 Which dmd version should I use to build DustMite?
 I tried 2.055 and 2.058 but DustMite crashes half way through 
 the
 reductino process without a message (I'm using windows)



 I've never heard of a problem like this. Is the DustMite crash
 reproducible?



 I've made the experience that it is more-or-less impossible to 
 run dustmite on windows for longer time spans. The cause is 
 most probably Windows Explorer, which sometimes accesses and 
 locks files.

 In a project we wrote a limited retry-loop around the critical 
 file system calls to get stable runs...



I've never encountered such problems. DustMite can run for days 
on end without issues.
Feb 19 2012
prev sibling parent "Vladimir Panteleev" <vladimir thecybershadow.net> writes:
On Sunday, 19 February 2012 at 21:42:17 UTC, Vladimir Panteleev 
wrote:

 On Sunday, 19 February 2012 at 17:56:42 UTC, Benjamin Thaut 
 wrote:

 rmdirRecurse which gets called in dustmite.d line 457 throws a 
 exception that causes the D exception handler to go into 
 inifnite recursion which leads to a stack overflow and crashes 
 the program.



 Thanks for taking the time to figure this out. I'll make 
 DustMite catch rmdirRecurse failures and retry the operation 
 (as the causes are usually temporary).



https://github.com/CyberShadow/DustMite/commit/e2d91dab03b7c98dfbacbbc41f7cf49d3935fd35
Feb 21 2012
prev sibling parent reply Don <nospam nospam.com> writes:
On 19.02.2012 14:24, Benjamin Thaut wrote:

 The last dmd version I could use to actually build my project was dmd
 2.056. Now I want to help finding those bugs that keep me from compiling
 and I'm wondering how I can build repro cases for the ICEs I encounter.
 I tried compiling with -v but that does not really give much information
 about where the crash actually happens. So I wonder what can actually be
 done to find the part of the code or even only the module it crashes in?

 The current ICE I'm encountering is:
 Assertion failure: 't1->ty == t2->ty' on line 2082 in file 'cast.c'



If you don't mind building the compiler,
add this line just before 2082:
if (t1->ty != t2->ty) e1->error("ICE");

and then build the compiler.

Sorry for the newbie-hostile comment. But it would save you a lot of time.
Feb 19 2012
parent reply Benjamin Thaut <code benjamin-thaut.de> writes:
Am 19.02.2012 15:41, schrieb Don:

 On 19.02.2012 14:24, Benjamin Thaut wrote:

 The last dmd version I could use to actually build my project was dmd
 2.056. Now I want to help finding those bugs that keep me from compiling
 and I'm wondering how I can build repro cases for the ICEs I encounter.
 I tried compiling with -v but that does not really give much information
 about where the crash actually happens. So I wonder what can actually be
 done to find the part of the code or even only the module it crashes in?

 The current ICE I'm encountering is:
 Assertion failure: 't1->ty == t2->ty' on line 2082 in file 'cast.c'



 If you don't mind building the compiler,
 add this line just before 2082:
 if (t1->ty != t2->ty) e1->error("ICE");

 and then build the compiler.

 Sorry for the newbie-hostile comment. But it would save you a lot of time.



Thank you very much that helped a lot. I was able to reduce it to the 
following code:

class Foo
{

    property EntityId entityId()
   {
     return EntityId(3);
   }

}

struct EntityId
{
   uint id;

   alias id this;

   this(uint id)
   {
     this.id = id;
   }
}

int main(string[] argv)
{
   Foo foo = null;
   auto id = (foo) ? foo.entityId : -1;

   return 0;
}
Feb 19 2012
next sibling parent reply "Daniel Murphy" <yebblies nospamgmail.com> writes:
"Benjamin Thaut" <code benjamin-thaut.de> wrote in message 
news:jhr584$2d1v$1 digitalmars.com...

 Thank you very much that helped a lot. I was able to reduce it to the 
 following code:

 class Foo
 {

    property EntityId entityId()
   {
     return EntityId(3);
   }

 }

 struct EntityId
 {
   uint id;

   alias id this;

   this(uint id)
   {
     this.id = id;
   }
 }

 int main(string[] argv)
 {
   Foo foo = null;
   auto id = (foo) ? foo.entityId : -1;

   return 0;
 }



"Daniel Murphy" <yebblies nospamgmail.com> wrote in message 
news:jhqua0$20c0$1 digitalmars.com...

  It shouldn't be happening, but I'd guess it's something to do with alias 
 this with integral types (huge guess).  How big is the project?



Damn I'm good.
Feb 19 2012
parent Benjamin Thaut <code benjamin-thaut.de> writes:
Am 19.02.2012 16:44, schrieb Daniel Murphy:

 "Benjamin Thaut"<code benjamin-thaut.de>  wrote in message
 news:jhr584$2d1v$1 digitalmars.com...

 Thank you very much that helped a lot. I was able to reduce it to the
 following code:

 class Foo
 {

     property EntityId entityId()
    {
      return EntityId(3);
    }

 }

 struct EntityId
 {
    uint id;

    alias id this;

    this(uint id)
    {
      this.id = id;
    }
 }

 int main(string[] argv)
 {
    Foo foo = null;
    auto id = (foo) ? foo.entityId : -1;

    return 0;
 }



 "Daniel Murphy"<yebblies nospamgmail.com>  wrote in message
 news:jhqua0$20c0$1 digitalmars.com...

   It shouldn't be happening, but I'd guess it's something to do with alias
 this with integral types (huge guess).  How big is the project?



 Damn I'm good.



Create a bug ticket for it: 
http://d.puremagic.com/issues/show_bug.cgi?id=7545

The bug was introduced because I had to replace a typedef with a wrapper 
struct.

Kind Regards
Benjamin Thaut
Feb 19 2012
prev sibling parent bearophile <bearophileHUGS lycos.com> writes:
Benjamin Thaut:


 Thank you very much that helped a lot. I was able to reduce it to the 
 following code:
 
 class Foo
 {
 
     property EntityId entityId()
    {
      return EntityId(3);
    }
 
 }
 
 struct EntityId
 {
    uint id;
 
    alias id this;
 
    this(uint id)
    {
      this.id = id;
    }
 }
 
 int main(string[] argv)
 {
    Foo foo = null;
    auto id = (foo) ? foo.entityId : -1;
 
    return 0;
 }



Reduced:

struct Foo {
   uint bar;
   alias bar this;
   this(int) {}
}
void main() {
   auto spam = true ? Foo(1) : 1;
}


Bye,
bearophile
Feb 19 2012