www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Assertion failure: '!vthis->csym' on line 703 in file 'glue.c'

reply nrgyzer <nrgyzer gmail.com> writes:
Hey guys,

I got "Assertion failure: '!vthis->csym' on line 703 in file 'glue.c'"
after I add "LinkList!(uint) myList;" to my source file. I figured out
that the same bug was already reported on http://lists.puremagic.com/
pipermail/digitalmars-d-bugs/2010-October/019237.html
Ticket 4129 describes a bug in glue.c but for DMD 1.x and I'm using DMD
2.052. I can't also reproduce the error in other projects - it seems
that it only exists in this project. In this case I think it makes no
sense to post thousand lines of code. Thus I hope anyone know what can
be wrong => thanks!
Mar 26 2011
next sibling parent reply bearophile <bearophileHUGS lycos.com> writes:
nrgyzer:


 In this case I think it makes no sense to post thousand lines of code.


I suggest to copy your code, and then keep removing lines from your copy,
making sure it keep showing the same compiler error. In some time you will
probably be able to produce a small program, fit for this newsgroup or even for
Bugzilla. (There are tools to do this reduction automatically in Python, I
think).

Bye,
bearophile
Mar 26 2011
parent reply David Nadlinger <see klickverbot.at> writes:
On 3/26/11 11:08 PM, bearophile wrote:

 I suggest to copy your code, and then keep removing lines from your copy,
making sure it keep showing the same compiler error. In some time you will
probably be able to produce a small program, fit for this newsgroup or even for
Bugzilla. (There are tools to do this reduction automatically in Python, I
think).


You don't necessarily need fancy tools to do that – as I recently 
pointed out in another thread, the Tigris »delta« tool worked fine for 
me on several occasions (http://delta.tigris.org/). Just write a simple 
shell script looking for the error message in the compiler output, and 
delta will try to automatically reduce the file for you.

There is some room for improvement though when you have an actual 
lexer/parser available for heuristically determining what pieces of code 
to delete – this might be a nice demonstration project for a D parsing 
library.

David
Mar 26 2011
parent reply nrgyzer <nrgyzer gmail.com> writes:
== Auszug aus David Nadlinger (see klickverbot.at)'s Artikel

 On 3/26/11 11:08 PM, bearophile wrote:

 I suggest to copy your code, and then keep removing lines from




your copy, making sure it keep showing the same compiler error. In
some time you will probably be able to produce a small program, fit
for this newsgroup or even for Bugzilla. (There are tools to do this
reduction automatically in Python, I think).

 You don't necessarily need fancy tools to do that – as I recently
 pointed out in another thread, the Tigris »delta« tool worked


fine for

 me on several occasions (http://delta.tigris.org/). Just write a


simple

 shell script looking for the error message in the compiler output,


and

 delta will try to automatically reduce the file for you.
 There is some room for improvement though when you have an actual
 lexer/parser available for heuristically determining what pieces of


code

 to delete – this might be a nice demonstration project for a D


parsing

 library.
 David


I figured out what produces the error - "import
dcollections.LinkList;" (declared as private) in one source file. But
I'm still unable to reproduce that error in a sample-application.
What I currently can say, that I have two source files like:

module firstModule;

private {
   import dcollections.LinkList;
   LinkList!(uint) myPrivateList;
}

static this() {
   myPrivateList = new LinkList!(uint);
}

module secondModule;

import firstModule;
import dcollections.LinkList;

LinkList!(uint) list;

void main(string[] args) {
   list = new LinkList!(uint);
}

In my tests, that code above doesn't produces the assertion failure,
but my original source code looks very similar.
Mar 27 2011
parent reply bearophile <bearophileHUGS lycos.com> writes:
nrgyzer:


 In my tests, that code above doesn't produces the assertion failure,
 but my original source code looks very similar.


You may have to start over then. I suggest to duplicate the whole code, and
then start on the copied version removing modules and lines of code, compiling
it every time you remove something, making sure you keep seeing the error
message every time, and undoing the change when you see the error vanish or
change.

Bye,
bearophile
Mar 27 2011
parent nrgyzer <nrgyzer gmail.com> writes:
== Auszug aus bearophile (bearophileHUGS lycos.com)'s Artikel

 nrgyzer:

 In my tests, that code above doesn't produces the assertion




failure,

 but my original source code looks very similar.


 You may have to start over then. I suggest to duplicate the whole


code, and then start on the copied version removing modules and lines
of code, compiling it every time you remove something, making sure
you keep seeing the error message every time, and undoing the change
when you see the error vanish or change.

 Bye,
 bearophile


That's a bit mysterious... I replaced firstModule with the source
which produces the error. Now... when I run my test application, I
get the same failure - but... that's not all - when I restore
firstModule to the version which didn't produce the failure, I just
get the same error. When I copy the source files to a new location
and compile it, I didn't get the error.
It seems that there is anything like a cache, but I removed all
produced (obj, rsp...) files, expect the exe- and d-files. Thus I
think there should be something like a cache.
I currently haven't any idea what I can do to reproduce the failure
efficently.
Mar 27 2011
prev sibling parent Don <nospam nospam.com> writes:
nrgyzer wrote:

 Hey guys,
 
 I got "Assertion failure: '!vthis->csym' on line 703 in file 'glue.c'"
 after I add "LinkList!(uint) myList;" to my source file. I figured out
 that the same bug was already reported on http://lists.puremagic.com/
 pipermail/digitalmars-d-bugs/2010-October/019237.html
 Ticket 4129 describes a bug in glue.c but for DMD 1.x and I'm using DMD
 2.052. 


It's still the same bug. The source code for the backend is the same for 
  all versions of DMD. The reported line number might change slightly 
(eg, might be line 691 on an old DMD version).
Mar 28 2011