This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: binutils 2.19.92 linker broke with --as-needed flag
- From: Richard Sandiford <rdsandiford at googlemail dot com>
- To: Vincent Wen <wenvincent90 at gmail dot com>
- Cc: Alan Modra <amodra at gmail dot com>, binutils at sourceware dot org
- Date: Tue, 23 Apr 2013 18:27:04 +0100
- Subject: Re: binutils 2.19.92 linker broke with --as-needed flag
- References: <5154E517 dot 4000409 at gmail dot com> <CALehDX77XzF6hGxT82y+Ofe+ypwG64uZmrfj+e4ZZm-nSEhgOA at mail dot gmail dot com> <CAF_=-cAo_+bni4N++ynNcYQ5RKkMBg1i7aNcXZ-aCJPVb3ub8Q at mail dot gmail dot com> <CALehDX5F_WXmKS9K+EipTw5uGkM37cOc2K+Zqxpe8XwBtzaTxg at mail dot gmail dot com> <CAF_=-cDsPpAs8=D3dH1ho1NRDJ+W5VVhug8HU1fJmusDq_aNAQ at mail dot gmail dot com> <CAF_=-cDJDPPzhh7TF2-rcro=PnL1iBkMv8jvV1BU7EO7En5zbg at mail dot gmail dot com> <20130402232743 dot GA23282 at bubble dot grove dot modra dot org> <CAF_=-cC7M0jhvSnsHA7xrPPedbWefQC0B9STpvQATP7nqZJg1g at mail dot gmail dot com> <20130403022010 dot GA27605 at bubble dot grove dot modra dot org> <CAF_=-cAYSgTU67tySGebDntQi1uMyANiuHGs0nWDHOC=GfPtOw at mail dot gmail dot com> <20130405003501 dot GA3208 at bubble dot grove dot modra dot org> <CAF_=-cB_uw9e++UmHwr+XSisgTAKRuAskvTuW0ZvGDd_z=6vkQ at mail dot gmail dot com> <CAF_=-cBM424MyCaPf0F=orbx3P3VaUrDeJ+Xncxjn2X18iO6mQ at mail dot gmail dot com>
Hi Vincent,
Sorry for the slow response. I've been on holiday for the past few weeks.
I agree that _GLOBAL_OFFSET_TABLE_ doesn't need to be dynamic.
The failures you're seeing:
Vincent Wen <wenvincent90@gmail.com> writes:
> FAIL: MIPS16 PIC test 2
> FAIL: MIPS JALX test 2
> FAIL: MIPS rel32 o32
> FAIL: MIPS rel32 n32
> FAIL: MIPS rel64 n64
> FAIL: PIC and non-PIC test 3 (shared library)
> FAIL: PIC and non-PIC test 3 (executable)
> FAIL: PIC and non-PIC test 4 (executable)
> FAIL: PIC and non-PIC test 5 (executable)
> FAIL: PIC and non-PIC test 6 (o32 executable)
> FAIL: PIC and non-PIC test 6 (n32 executable)
> FAIL: PIC and non-PIC test 6 (n64 executable)
> FAIL: MIPS eh-frame 5
> FAIL: Stub for dynsym 0x7fff
> FAIL: Stub for dynsym 0x8000
> FAIL: Stub for dynsym 0xfff0
> FAIL: Stub for dynsym 0x10000
> FAIL: Stub for dynsym 0x2fe80
> FAIL: Shared library with TLS
> FAIL: Dynamic executable with TLS
> FAIL: Shared library with multiple GOTs and TLS
> FAIL: Shared library with TLS and versioning
> FAIL: Dynamic executable with TLS and versioning
> FAIL: Dynamic executable with TLS and versioning (order 2)
> FAIL: Dynamic executable with TLS and versioning (order 3)
are probably just testcases that incidentally (rather than deliberately)
assume the old behaviour. If you can confirm and update the tests
to match the new behaviour, then the patch should be OK.
Thanks,
Richard