Re: [Boost-users] Strange issue with boost::date_time

Subject: Re: [Boost-users] Strange issue with boost::date_time
From: Jared Lee Richardson (jaredr26_at_[hidden])
Date: 2009-04-15 12:47:47

On Wed, Apr 15, 2009 at 1:29 AM, Neil Groves <neil_at_[hidden]> wrote:
> Jared,
>
> On Wed, Apr 15, 2009 at 1:04 AM, Jared Lee Richardson <jaredr26_at_[hidden]>
> wrote:
>>
>> Hmm, ok then.  I have a few more things I'll try tonight(go through
>> the code by hand checking every pointer, etc.)
>
> Even though the Debug build appears to be symptom free you might want to try
> the _CrtSetDbgFlag(_CRTDBG_CHECK_
> ALWAYS_DF); approach. If your code is single threaded it should find any
> heap corruption within your code.

I tried this, and unfortunately no errors came up in debug.

While thinking through how my code works, I did realize that I am
probably using a bad pointer somewhere- This is my first project with
a very heavy use of the STL, and I was maintaining pointers to objects
inside vectors.  Bad if the vector resizes and moves them elsewhere.
However, none of that code is being hit or touched before this crash
happens.

> If this is truly a Release build only issue then you might want to check the
> myriad of debug iterator preprocessor defines to ensure they are consistent
> between your project and the Boost.DateTime version you are using.
> Have you ensured that _SECURE_SCL=1 in your project for the Release
> configuration?

As far as I know, I'm not using any iterators at all before this is
hit.  I added _SECURE_SCL 1 to my global defines, but no help.

> If your project is fairly small and you are allowed, I would be happy for
> you to send the Visual Studio solution to my personal email address so that
> I can take a quick look at it. This might be the fastest way to proceed.

Sure- I tried everything I knew how this morning, and nothing fixed
it.  Sending you a separate email with that.  Thanks for that Neil,
much appreciated. :)

>> I'm not able to use the crtdbg.h to solve this it looks like- crtdbg
>> is only useful in debug, and the error does not seem to happen in
>> debug(at least, not the same).  I tried forcing it to run in debug by
>> changing only the settings required for it - #undef _RELEASE, #define
>> _DEBUG, changed runtime library from Multi-Threaded(/MT) to
>> Multi-Threaded Debug(/MTd), however the program ran fine, no crash(or
>> at least, not in that spot- my assertions triggered because I've
>> removed so much code).
>
> Did you mean Multi-Threaded DLL (/MD) and Multi-Threaded Debug DLL (/MDd)?
> IIRC the Boost shared libraries build against the shared C runtime by
> default and these should not be mixed.

I had used the BoostPro downloader to specifically download the static
link libraries.  However, on the chance it would help, I re-downloaded
the DLL's and changed it to /MD and /MDd.  The crash simply jumped to
a later spot in code.

Thanks, Jared