[PATCH,PE-COFF,PR11603] Allow ld to export symbols containing dots from DLLs.

[Doug Semler]

On Monday, May 17, 2010, Dave Korn <dave.korn.cygwin@gmail.com> wrote:
> On 17/05/2010 17:30, Doug Semler wrote:
>> On Mon, May 17, 2010 at 12:46 PM, Dave Korn <dave.korn.cygwin@gmail.com> wrote:
>>> On 17/05/2010 17:15, Doug Semler wrote:
>>>> Dave I did a test on x86_64-w64 and it's not exporting the data
>>>> symbols from the DLLs.  I'm not sure whether it's another underscoring
>>>> issue or a PEBKAC error but I'm investigating...i'll know more in a
>>>> little while...
>>>  Simple test case, by any chance?
>>
>> Yeah.  Looks like gcc (4.5) isn't generating a correct .drective if
>> there are both thread local data and exported functions in the same
>> module.  If I take the __thread off, it exports both.  If I get rid of
>> the function definition it exports i properly.
>
>   Urgh.  Yes, this looks like a real GCC bug; it's not generating any export
> directive for the tls var at all.  When there's no other export directives
> (i.e. without the function) auto-export kicks in, but that behaviour gets
> suppressed in the presence of any explicit exports.
>
>   Confirmed for both 4.6.0 and 4.3.4.  Sigh, I'll go debug GCC and figure out
> what's up.
>
>     cheers,
>       DaveK

Dave

I'm not going to get a chance to do it until tomorrow but I'll check
the combo binutils patch with the gcc declspec patch against the code
I have (it exercises this pretty well) and let you know how it goes...

Doug