Re: [Botan-devel] Help with Windows build

[Jack L. <ll...@ac...>]

On Thu, 5 Jun 2003, Jeff B wrote:

> I'm trying to get the 1.2.4 build to compile with Visual Studio .NET 2003
> (ver 7) and I'm having a bit of trouble.  Here's what I've done so far:
>
>
> 1. D/l 1.2.4 source
> (http://prdownloads.sourceforge.net/opencl/Botan-1.2.4.tgz) and unzipped
> 2. Ran configure.pl (I believe I used "perl configure.pl msvc-windows-i486")
> 3. ran VS, created a Visual C++ Win32 console application
> 4. added the contents of Botan's SRC directory to the new project
> 5. Went into the project's settings and
>     a) Turned off using precompiled headers
>     b) Added the Botan\include path to the "include additional directories"
> setting.
> 6. Rebuilt my test project

Hmm, I've never tried this method before, but I imagine it might run into
trouble. Here's the prefered method:

Steps 1/2 OK.
3. run 'nmake' to build libbotan.lib
4. Add Botan's base directory to the library search path in VC
4a. Tell the compiler to link with libbotan.lib
Step 5/6 OK.

> OK, first issue: There are a _LOT_ of compile warnings.  One of the
> warnings, C4290, found in base.h, line 42,
>       void set_key(const byte[], u32bit) throw(Invalid_Key_Length);
>
> is about exception specification being ignored.  Doing a search on C4290 in
> VS, I found "Handling Exceptions, part 6" in Deep C++, which goes on to say
> that VS ignores the exception specification at run time.  Lovely, I thought
> VS .NET 2003 was supposed to bring conformance to ANSI C++.

Well, in this case it's not really a warning - it's just the compiler
spouting off about how much it sucks (exception specfications are well
established, the MSDN page I found about this warning is basically lying).
I have always been under the impression that compilers would warn about
problems in the code, not problems in the compiler. Anyway...

> Is this warning a known issue or is my configure.pl settings wrong, or do I
> need to set any additional parameters in the project?  I can just silence
> the warning, but it seems dangerous.  OK, so much for the warnings.  Moving
> on...

I would say either use a pragma to ignore this specific 'warning', or lower
your warning level down to 2 so it never shows up. I don't see much of a
way out of this in the general case, besides massive code modifications.

I'll have config.h include a pragma disabling this warning in 1.2.5 - I
honestly see no useful reason for it to exist.

> 7. At the end of the build, I get 3 errors pertaining to get_ciph.cpp, line
> 87, which is
>
> Keyed_Filter* get_cipher(const std::string& algo_spec, Cipher_Dir direction)
>    {
>    struct Invalid_Combination : public Exception
>       {
>       Invalid_Combination(const std::string& mode, const std::string& pad) :
>          Exception("Invalid mode/padding pair: " + mode + "/" + pad) {}
>       };
>
> The compiler doesn't like
>         Exception("Invalid mode/padding pair: " + mode + "/" + pad)
>
>
> Are these known issues?

Hmmm, I hadn't heard about this second problem at all. Seems to depends on
the exact version of VC7 used (THANK YOU MICROSOFT). I'll try to figure out
a workaround for the next release.

> Are my configure options correct?

Yes, but you should build the library as a seperate entity, rather than
building all the source as part of your application.

-Jack