A workaround for the “Template Haskell + C” bug? - haskell

I've got the following situation:
Library X is a wrapper over some code in C.
Library A depends on library X.
Library B uses Template Haskell and depends on library A.
GHC bug #9010 makes it impossible to install library B using GHC 7.6. When TH is processed, GHCi fires up and tries to load library X, which fails with a message like
Loading package charsetdetect-ae-1.0 ... linking ... ghc:
~/.cabal/lib/x86_64-linux-ghc-7.6.3/charsetdetect-ae-1.0/
libHScharsetdetect-ae-1.0.a: unknown symbol `_ZTV15nsCharSetProber'
(the actual name of the “unknown symbol” differs from machine to machine).
Are there any workarounds for this problem (apart from “don't use Template Haskell”, of course)? Maybe library X has to be compiled differently, or there's some way to stop it from loading (as it shouldn't be called during code generation anyway)?

This is really one of the main reasons that 7.8 switched to dynamic GHCi by default. Rather than try to support every feature of every object file format, it builds dynamic libraries and lets the system dynamic loader handle them.
Try building with the g++ option -fno-weak. From the g++ man page:
-fno-weak
Do not use weak symbol support, even if it is provided by the linker. By default, G++ will use weak symbols if they are available. This option exists only for testing, and should not be used by end-users; it will result in inferior code and has no benefits. This option may be removed in a future release of G++.
There is another issue with __dso_handle. I found that you can at least get the library to load and apparently work by linking in a file which defines that symbol. I don't know whether this hack will cause anything to go wrong.
So in X.cabal add
if impl(ghc < 7.8)
cc-option: -fno-weak
c-sources: cbits/dso_handle.c
where cbits/dso_handle.c contains
void *__dso_handle;

Related

stack statically link external library on mac

I would really like to create a single binary for my application however it depends on libzmq and I am building on OSX, has anyone managed to get an external library linking correctly on a mac?
I did something similar with lua. My aproach is to pass full path to the static library to ghc. In cabal file:
if os(darwin)
ghc-options: /usr/local/lib/liblua5.1.a
According to documentation ghc will pass it to linker:
Files with other suffixes (or without suffixes) are passed straight to
the linker.
However it doesn't work if the library is used in other package which you depend on. So if you are using e.g. zeromq4-haskell package, then you need to modify it locally.

Weak symbols, shared libraries and dlopen

I have a binary with a weak symbol that I want to be able to link at runtime with a run dependent shared library.
$nm testrun
...
w basic2.test
...
My first test was using a .o file at static linktime, that worked, but I need it to be shared.
So, my second test was getting a shared library with that symbol defined and link it at compile time with -lmy (libmy.so), and this, actually worked as well.
Third step tried not linking at compile time and use ld_preload trick and this did not work.
nm libmy.so
...
00000550 T basic2.test
...
I have really no idea why this particular one does not work, looks like dynamic loader should have enough information to set testruns weak symbol with the one in libmy.so.
My final objective, which I guess will require more work is to load at start a small function that does check for the appropiate symbol with dlsym and sets it there.
Any hint?
It seems that you may need to use LD_DYNAMIC_WEAK along with LD_PRELOAD from the man page:
LD_DYNAMIC_WEAK (glibc since 2.1.91) Allow weak symbols to be overridden (reverting to old glibc behavior). For security reasons, since glibc 2.3.4, LD_DYNAMIC_WEAK is ignored for set-user-ID/set-group-ID binaries.
Note: it could be a typo, but you should use -lmylib.so and not -Lmylib.so

How to restrict access to symbols in shared object?

I have a plug-in in the form of a shared library (bar.so) that links into a larger program (foo). Both foo and bar.so depend on the same third party library (baz) but they need to keep their implementations of baz completely separate. So when I link foo (using the supplied object files and archives) I need it to ignore any use of baz in bar.so and vice versa.
Right now if I link foo with --trace-symbol=baz_fun where baz_fun is one of the offending symbols I get the following output:
bar.so: definition of baz_fun
foo/src.a(baz.o): reference to baz_fun
I believe this is telling me that foo is referencing baz_fun from bar.so (and execution of foo confirms this).
Solutions that I have tried:
Using objcopy to "localize" the symbols of interest: objcopy --localize-symbols=local.syms bar.so where local.syms contains all of the symbols of interest. I think I might just be confused here and maybe "local" doesn't mean what I think it means. Regardless, I get the same output from the link above. I should note that if I run the nm tool on bar.so prior to using objcopy all of the symbols in question have the T flag (upper-case indicating global) and after objcopy they have a t indicating they are local now. So it appears I am using objcopy correctly.
Compiling with -fvisibility=hidden however due to some other constraints I need to use GCC 3.3 which doesn't appear to support that feature. I might be able to upgrade to a newer version of GCC but would like confirmation that compiling with this flag will help me before heading down that road.
Other things to note:
I do not have access to the source code of either foo or baz
I would prefer to keep all of my plug-in in one shared object (bar.so). baz is actually a licensing library so I don't want it separated
Use dlopen to load your plugin with RTLD_DEEPBIND flag.
(edit)
Please note that RTLD_DEEPBIND is Linux-specific and need glibc 2.3.4 or newer.

VC6 linking conflict

I have an old c++ project compiled with VC6.
I need to statically link a new library to implement a new functionality.
Unfortunately the new library define a symbol (i.e. _inflate) that is already defined in a previously linked static library.
Of course I cannot get rid of either library, and of course I have no access to the library's source code.
Is there a way to avoid the linker error (LNK2005)?
If you know that the two versions of _inflate are identical, or at least "compatible", then you can use the /FORCE:Multiple linker option to force it to ignore name clashes.
Of course, if it links to a version of that code that is "incompatible" in any way it could cause undefined behaviour.

Link libraries with dependencies in Visual C++ without getting LNK4006

I have a set of statically-compiled libraries, with fairly deep-running dependencies between the libraries. For example, the executable X uses libraries A and B, A uses library C, and B uses libraries C and D:
X -> A
A -> C
X -> B
B -> C
B -> D
When I link X with A and B, I don't want to get errors if C and D were not also added to the list of libraries—the fact that A and B use these libraries internally is an implementation detail that X should not need to know about. Also, when new dependencies are added anywhere in the dependency tree, the project file of any program that uses A or B would have to be reconfigured. For a deep dependency tree, the list of required libraries can become really long and hard to maintain.
So, I am using the "Additional Dependencies" setting of the Librarian section in the A project, adding C.lib. And in the same section of B's project, I add C.lib and D.lib. The effect of this is that the librarian bundles C.lib into A.lib, and C.lib and D.lib into B.lib.
When I link X, however, both A.lib and B.lib contain their own copy of C.lib. This leads to tons of warnings along the lines of
A.lib(c.obj) : warning LNK4006 "symbol" (_symbol) already defined in B.lib(c.obj); second definition ignored.
How can I accomplish this without getting warnings? Is there a way to simply disable the warning, or is there a better way?
EDIT: I have seen more than one answer suggesting that, for the lack of a better alternative, I simply disable the warning. Well, this is part of the problem: I don't even know how to disable it!
As far as I know you can't disable linker warnings.
However, you can ignore some of them, using command line parameter of linker eg. /ignore:4006
Put it in your project properties under linker->command line setting (don't remember exact location).
Also read this:
Link /ignore
MSDN Forum - hiding LNK warnings
Wacek
Update If you can build all involved project in single solution, try this:
Put all project in one sln.
Remove all references to static libraries from projects' linker or librarian properties.
There is "Project Dependencies..." option in context menu for each project in Solution Explorer. Use it to define dependencies between project.
It should work. It doesn't invalidate anything I said before, the basic model of building C/C++ programs stays the same. VS (at least 2005 and newer) is simply smart enough to add all needed static libraries to linker command line. You can see it in project properties.
Of course this method won't help if you need to use already compiled static libraries. Then you need to add them all to exe or dll project that directly or indirectly uses them.
I don't think you can do anything about that. You should remove references to other static libs from static libs projects and add all needed static libs projects as dependences of exe or dll projects. You will just have to live with fact that any project that includes A.lib or B.lib also needs to include C.lib.
As an alternative you can turn your libraries into dlls which provide a richer model.
Statically compiled libraries simply aren't real libraries with dependency information, etc, like dlls. See how, when you build them, you don't really need to provide libraries they depend on? Headers are all that's needed. See? You can't even really say static libraries depend on something.
Static library is just an archive of compiled and not yet linked object code. It's not consistent whole. Each object file is compiled separately and remains separate entity inside the library. Linking happens when you build exe or dll. That's when you need to provide all object code. That's when all the symbol and dependency resolving happens.
If you add other static libraries to static library dependencies, librarian will simply copy all code together. Then, when building exe, linker will give you lots of warnings about duplicate symbols. You might be able to block those warnings (I don't know how) but be careful. It may conceal real problems like real duplicate symbols with differing definitions. And if you have static data defined in libraries, it probably won't work anyway.
Microsoft (R) Incremental Linker Version 9.00.x (link.exe) knows argument /ignore:4006
You could create one library which contains A, B, C & D and then link X against that.
Since it's a library, only object modules which are actually referenced will get linked into the final executable.
Note that one way of getting this warning is to define a member function in a header without the inline statement:
// Foo.h
class Foo
{
void someFunction();
};
void Foo:someFunction() // Warning! - should be "inline void Foo::someFunction()"
{
// do stuff
}
The problem is you are not localizing library C's symbols. So you have a ODR violation when you link in A and B. You need to have a way to make these private. By default all symbols are exported. One way to do this is to have a special linker definition file for both A and B that explicitly mention which files need to be exported.
[1] ODR = One Definition Rule.
I think the best course of action here will be to ignore/disable the linker warnings(LNK4006) since C.lib needs to be part of both A.Lib and B.lib and A.Lib does not need to know that B.lib itself uses C.Lib.
This may not fix your link error, but it might help with your dependency tree issue.
What I do, is just use a #pragma to include a lib in the .cpp file that needs it. For example:
#pragma comment(lib:"wsock32")
Like I said, I'm not sure it would keep the symbols in that object file, I'd have to whip up an example to try it out.
Poor flodin seems frustrated that nobody will explain how to disable the linker warnings. Well, I've had a similar problem, and for years I have simply lived with the fact that several hundred warnings were displayed. Now, however, thanks to the info from Link /ignore, I figured out how to disable the linker warnings.
I'm using Visual Studio 2008. In Project -> Settings -> Configuration Properties -> Librarian -> Command Line -> Additional Options, I added "/ignore:4006" (without the quotes). Now my warnings are gone!

Resources