[PATCH] D29324: [ELF] - Allow to combine sections of type SHT_NOTE with different attributes.

Evgeny Leviant via llvm-commits llvm-commits at lists.llvm.org
Wed Feb 1 02:30:52 PST 2017


Like George and Dave, I also think that having some way to bypass these flag/type checks is not a bad idea. For instance I had to patch some our assembly files, because code sections missed "ax" attributes.?

If we already can merge nobits and progbits sections, why can't we do the same with alloc / non-alloc sections then? May be emit warning instead of error in such cases?

________________________________
От: Rui Ueyama <ruiu at google.com>
Отправлено: 1 февраля 2017 г. 2:28
Кому: Rafael Avila de Espindola
Копия: George Rimar; Evgeny Leviant; llvm-commits at lists.llvm.org; reviews+D29324+public+1af718958cd66f3e at reviews.llvm.org
Тема: Re: [PATCH] D29324: [ELF] - Allow to combine sections of type SHT_NOTE with different attributes.

On Tue, Jan 31, 2017 at 12:41 PM, Rafael Avila de Espindola <rafael.espindola at gmail.com<mailto:rafael.espindola at gmail.com>> wrote:
George Rimar <grimar at accesssoftek.com<mailto:grimar at accesssoftek.com>> writes:

> Sure, I'll check that tomorrow and return with info.
>
> By the way, may be we want to implement some simple option like -z permissive-flags, which
> will ignore flags checks ? It looks it can be not a unique case for software that have such issues.

It is not clear it is worth it. If it is common, we should just make lld
permissive. If it is one or two cases, we should try to fix those cases.

I agree. Adding a new flag might be easy but removing it is extremely hard or even impossible, so naturally I'm conservative when doing it.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170201/7a603802/attachment.html>


More information about the llvm-commits mailing list