[lld] r205566 - Fix ELFFileNode::resetNextIndex()

Rui Ueyama ruiu at google.com
Mon May 12 16:45:54 PDT 2014


http://reviews.llvm.org/D3734 is sent out for review to fix this.


On Mon, May 12, 2014 at 12:56 PM, Rui Ueyama <ruiu at google.com> wrote:

> I guess I can fix it by restoring "_files[0]->kind() ==
> File::kindSharedLibrary" to GnuLdInputGraph.h. I'll try to see if it works.
>
>
> On Sun, May 11, 2014 at 10:36 PM, Simon Atanasyan <simon at atanasyan.com>wrote:
>
>> CC to the mail list and attach the patch directly to the message.
>>
>> On Mon, May 12, 2014 at 9:28 AM, Simon Atanasyan <simon at atanasyan.com>
>> wrote:
>> > It looks like this commit introduced the following regression. Suppose
>> we have:
>> >
>> >   - Object file `main.o` which refers symbol `RT1`
>> >   - Object file `obj.o` defines symbol `RT1` and refers symbol `T1`
>> >   - Shared library libso.so defines symbol `T1`
>> >
>> > If we link these files using the following command we incorrectly get
>> an error:
>> >
>> >   % lld -flavor gnu -target mipsel -e T -o a.out main.o --start-group
>> libso.so obj.o --end-group
>> >   Undefined symbol: obj.o: T1
>> >
>> > That is because we do not scan the `libso.so` once again to find the
>> `T1` definition. The attached patch {F57334} has a reproduction test case.
>> >
>> > http://reviews.llvm.org/rL205566
>>
>> --
>> Simon Atanasyan
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140512/e885f356/attachment.html>


More information about the llvm-commits mailing list