Re: [Boost-build] msvc 64-bit test targets vs. maifest files

Subject: Re: [Boost-build] msvc 64-bit test targets vs. maifest files
From: Mat Marcus (mat-lists_at_[hidden])
Date: 2008-10-12 22:06:47

On Sun, Oct 12, 2008 at 11:51 AM, Vladimir Prus <ghost_at_[hidden]> wrote:
> On Thursday 09 October 2008 00:34:13 Mat Marcus wrote:
>> Second attempt (no answer the first time. if no answer again, will try
>> boost-users, then boost developers)
>
> Hi Mat,
> I was hoping somebody with windows experience would comment, before
> impressing the world with my deep lack of knowledge about it, but alas...
>
>> On Fri, Oct 3, 2008 at 11:50 AM, Mat Marcus <mat-lists_at_[hidden]> wrote:
>> > For several years I have been working with a set of libraries, built
>> > with boost build and dependent upon some boost libraries. I've been
>> > working on a 32-bit machine, but I have occasionally been building
>> > 64-bit targets to try to help support clients who will be using my
>> > library on 64-bit machines. My suite of boost-build test targets
>> > (unit-test, run, etc.) for my lib is a bit inconvenient when on 32 bit
>> > machines, since I get an error dialog each time I build, but sometimes
>> > I want to test as much as possible. The commands I use to build (msvc
>> > 9) are:
>> >
>> > ## 32 bit
>> > bjam msvc debug
>> >
>> > ## 64 bit
>> > bjam msvc debug architecture=x86 address-model=64
>> > define=BOOST_DISABLE_ABI_HEADERS
>> >
>> > Yesterday, I tried to build/run the 64 bit targets for the same files
>> > using the same invocation on a 64-bit Vista machine. I thought that
>> > perhaps my code would exhibit some 64-bit run-time issues, but things
>> > didn't go that far. Instead, I got manifest/DLL-related problems for
>> > each test. Here is an example of the error that I receive each time
>> > boost build tries to run a built artifact:
>> >
>> > ====== BEGIN OUTPUT ======
>> > The application has failed to start because its side-by-side
>> > configuration is incorrect. Please see the application event log for
>> > more detail.
>> >
>> > EXIT STATUS: 14001
>> > ====== END OUTPUT ======
>> >
>> > The corresponding Application Event Log Entry is:
>> >
>> > Activation context generation failed for
>> > "c:\asl\staff\mmarcus\sandbox\built_artifacts\documentation\examples\enum_ops_example.test\msvc-9.0\debug\address-model-64\architecture-x86\link-static\threading-multi\enum_ops_example.exe".
>> > Dependent Assembly
>> > Microsoft.VC90.DebugCRT,processorArchitecture="amd64",publicKeyToken="1fc8b3b9a1e18e3b",type="win32",version="9.0.21022.8"
>> > could not be found. Please use sxstrace.exe for detailed diagnosis.
>
> So, at the link time, reference to the right library is made, but at runtime, the library is not available.

That sounds correct.

> Do you know the location of the CRT library found at link time?

Some research reveals that one copy is located at:

C:\Program Files (x86)\Microsoft Visual Studio
9.0\VC\Debug_NonRedist\amd64\Mcirosoft.VC90.DebugCRT

(For completeness, the path to the 32-bit one replace amd64 with x86)

> It probably has a .manifest file giving
> the names of actual .dll files. Are those .dll founds located in directory in PATH? Maybe
> the VisualStudio install process installs those libraries so that they are available to linker,
> but not at runtime? Does that sxstrace.exe tool tell anything?

The strace didn't seem to shed any new light. The problem seems not
uncommon, e.g., IIRC

http://social.msdn.microsoft.com/Forums/en-US/vcgeneral/thread/b2aacd05-d9e9-4c6f-95f5-d6c005c9a5c6/

Is there other info that I could provide that might help?

Thanks,
Mat

P.S. FWIW, one other issue was that I think I had to add the path to
(the 32-bit) cl.exe to my user-config.jam, since the fact that Dev
Studio gets installed under "Program Files (x86)" by default on vista
64 seemed to take boost build by surprise.