[PATCH] Add a fallback mechanism for undefined atom.

[Rui Ueyama]

Guys,

I want to submit this patch. This patch is one of the last few patches that
I need to link against the standard C++ library on Windows. Symbol aliasing
does not seems to be needed for ELF. For Mach-O we need the feature, but
how to implement can be discussed later. This patch is short and not
intrusive, so it won't hurt anyone.


On Tue, Sep 3, 2013 at 1:07 PM, Shankar Easwaran <shankare at codeaurora.org>wrote:

> On 9/3/2013 2:59 PM, Nick Kledzik wrote:
>
>> On Sep 3, 2013, at 12:51 PM, Shankar Easwaran <shankare at codeaurora.org>
>> wrote:
>>
>>  On 9/3/2013 2:36 PM, Nick Kledzik wrote:
>>>
>>>> On Sep 3, 2013, at 11:06 AM, Rui Ueyama <ruiu at google.com> wrote:
>>>>
>>>>> Nick, does Mach-O need symbol aliasing to be handled in the linker?
>>>>>
>>>> Yes.  The darwin linker has command line options for adding aliases.
>>>>
>>>> Back when I first started working on the darwin linker, a big complaint
>>>> from the compiler team was that mach-o did not support aliases like ELF
>>>> did, so the compiler team kept having to special case lots of test cases
>>>> and functionality that used __attribute__((alias(“xx”))) on darwin.
>>>>
>>>>
>>>>  looks like ELF also needs the same functionality for implementing
>>> __wrap.
>>>
>>> But in the case of the linker dealing with compiled .o's, the alias
>>> information is lost.
>>>
>> Really?  I thought ELF had a way to encode this.   I’ve received bug
>> reports saying that mach-o should support it too.
>>
>
> Example :-
>
> $cat x.c
> void __f () { /* Do something. */; }
>           void f () __attribute__ ((weak, alias ("__f")));
> $gcc -c x.c
> $readelf -s x.o
>
> Symbol table '.symtab' contains 10 entries:
>    Num:    Value  Size Type    Bind   Vis      Ndx Name
>      0: 00000000     0 NOTYPE  LOCAL  DEFAULT  UND
>      1: 00000000     0 FILE    LOCAL  DEFAULT  ABS x.c
>      2: 00000000     0 SECTION LOCAL  DEFAULT    1
>      3: 00000000     0 SECTION LOCAL  DEFAULT    2
>      4: 00000000     0 SECTION LOCAL  DEFAULT    3
>      5: 00000000     0 SECTION LOCAL  DEFAULT    5
>      6: 00000000     0 SECTION LOCAL  DEFAULT    6
>      7: 00000000     0 SECTION LOCAL  DEFAULT    4
>      8: 00000000     5 FUNC    GLOBAL DEFAULT    1 __f
>      9: 00000000     5 FUNC    WEAK   DEFAULT    1 f
>
>
>
>>  Does Darwin keep the alias information when the
>>> __attribute__((alias(“xx”)) is used.
>>>
>> What the current darwin linker does is if parses a .o file and there are
>> two symbols at the same address (and same section), it picks one to be the
>> atom name (call it “foo”) and the other (call it “bar”) is turned into an
>> alias atom (zero size and atom foo must layout after atom bar)
>>
> We do something very similiar in lld in the ELF reader, except that we
> dont alias it(as we need the content of the original atom, for the linker
> to abide the section property).
>
> Thanks
>
> Shankar Easwaran
>
> --
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted
> by the Linux Foundation
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130906/dcaee506/attachment.html>