Bug in gcc and/or binutils?

[Charles Wilson]

Danny Smith wrote:
> Charles Wilson <cygwin at cwilson dot fastmail dot fm>  wrote:
>>Using .def files turns off the auto-EXport logic (which it should,
>>because if you specify a specific set of exports you don't want binutils
>>adding a few more on its own).
> 
> 
> There seems to be a common misconception that auto-export logic and
> explicit designation of exports are incompatible.
> You can still use --export-all  (to get the auto-export of all defined
> symbols) with a .def file or with __declspec(dllexport).
> 
> eg
> gcc -shared -ofoo.dll foo.def foo.c -Wl,--export-all

Right.  Using a .def file or decorating code with __declspec(dllexport) 
DOES turn off auto-EXport, unless you explicitly turn it back on with 
--export-all.

So, libtool could "solve" the problem by adding --export-all; but *I 
think* libtool only wants to export certain symbols, not all of them. 
So this solution is not the proper one for this particular issue.

> The use of a def file (with --export-all) is useful when you want to
> export all symbols but want special handling of some, say an alias or
> marking a symbol as PRIVATE (exclude from import lib) or NONAME (exclude
> the name of the symbol from the dll's export table), (almost like
> __attribute__((hidden)) Or you may want to add a symbol that is just
> forwarded to another dll. Or you have a function foo() but only want
> the indirect ref __imp__foo visible in the import lib, and not the label:
> 
> foo:
>   jmp * __imp__foo
> 
> so you only export the pointer  by marking as DATA in the def file

Right -- you'd mark it DATA because it IS data:

EXPORTS
   __imp__foo  @1  DATA

exports a pointer.

--
Chuck



--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/