[PATCH] D26130: [ELF] - Implemented --symbol-ordering-file option.

[Rafael Avila de Espindola via llvm-commits]

I am against reverting it.

The gold feature is clearly incomplete. It is also dependent on how
sections are named, which is a bad design for ELF and incompatible with
-fno-unique-section-names.

We can add the gold feature if we really want to, but using symbol names
is actually far more flexible.

Cheers,
Rafael

On November 19, 2016 1:24:26 PM EST, Rui Ueyama <ruiu at google.com> wrote:
>Well, if I knew it was a completely new feature, I didn't LGTM on this
>patch. It needs a strong justification to add a new feature, and even
>if we
>agree with a need of a feature, we still have to discuss what is the
>best
>way of doing it. I think if gold already has a similar feature, we
>should
>implement that, instead of inventing our own.
>
>We probably should revert this now, and restart from a proposal and
>discussion.
>
>On Fri, Nov 18, 2016 at 11:15 PM, George Rimar
><grimar at accesssoftek.com>
>wrote:
>
>> >Until now, I was thinking that --symbol-ordering-file is a feature
>that
>> ld.gold has, but looks like ld.gold doesn't have one. gold has
>> --section-ordering-file instead. >Where does this
>--symbol-ordering-file
>> feature come from?
>>
>>
>> Yes, that is true. Reason to do that was that symbol ordering file
>allows
>>
>> to map symbols to sections with the same name like .text and reorder
>them.
>> Sections ordering file does not allow to
>>
>> do that.
>>
>>
>> George.
>>

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20161120/d8774f98/attachment.html>