Geometry Boost List: Re: [geometry] access violation in update_discarded(...)

Subject: Re: [geometry] access violation in update_discarded(...)
From: Barend Gehrels (barend_at_[hidden])
Date: 2014-11-06 09:52:05

Hi Volker,

> *From:*Geometry [mailto:geometry-bounces_at_[hidden]] *On Behalf 
> Of *Volker Schöch
> *Sent:* Donnerstag, 30. Oktober 2014 11:34
> *To:* geometry_at_[hidden] <mailto:geometry_at_[hidden]>
> *Subject:* Re: [geometry] access violation in update_discarded(...)
>
> Done.
>
> https://svn.boost.org/trac/boost/ticket/10719
>

Volker Schöch wrote On 6-11-2014 9:55:
>
> Hi,
>
> Can you give any perspective on when or how this problem may be solved?
>
> It's mission critical to our software, so we have to do something 
> about it soon.
>

1) I think I have to write: it is volunteer work - the ticket is opened 
one week ago. We appreciate your usage of Boost.Geometry and the 
feedback too, but take care not to transfer the pressure you feel to us 
volunteers... At least not to soon.

But I looked.

2) Menelaos send a sample program which works for him, and works for me. 
Did you try that program? He also asked

> Do you think you can reproduce the problem in a standalone program?

but you did not answer that question. Actually the whole mail was 
unanswered.

3) you already mention that your input is invalid - so you should be 
able to workaround the problem probably easily.

4) you include a stack-trace (thanks for that) but things have been 
changed, get_situation_map is gone from enrich_intersection_points, and 
so might be the whole problem. This is by chance released this week

So my new question is: does it work for you in 1.57?

5) Valentin reports:
> If seg_id.source_index == -1, copy_segment_points does not copy any 
> data, thus the computation continues with random junk values from the 
> stack :-(

but that may indeed never be the case in that sort functor. source_index 
should be either 0 or 1. If it indeed has another value, this indicates 
an error. But I cannot find or reproduce this error (I used MSVC on the 
old version, and the new version with clang and valgrind).

Regards, Barend