Haskell failures: relocation refers to local symbol "" [1], which is defined in a discarded section

[Elliott Sales de Andrade]

Hello all,

All Haskell packages seem to be broken now, e.g., ghc-optional-args [1] or anything on
koschei [2]. They all fail like this:

+ ghc --make -no-user-package-db -dynamic Setup
[1 of 1] Compiling Main             ( Setup.hs, Setup.o )
Linking Setup ...
/usr/lib/gcc/x86_64-redhat-linux/8/../../../../lib64/crti.o(.gnu.build.attributes+0x14):
error: relocation refers to local symbol "" [1], which is defined in a discarded
section
/usr/lib/gcc/x86_64-redhat-linux/8/../../../../lib64/crti.o(.gnu.build.attributes+0x1c):
error: relocation refers to local symbol "" [1], which is defined in a discarded
section
/usr/lib/gcc/x86_64-redhat-linux/8/../../../../lib64/crtn.o(.gnu.build.attributes+0x14):
error: relocation refers to local symbol "" [1], which is defined in a discarded
section
/usr/lib/gcc/x86_64-redhat-linux/8/../../../../lib64/crtn.o(.gnu.build.attributes+0x1c):
error: relocation refers to local symbol "" [1], which is defined in a discarded
section
collect2: error: ld returned 1 exit status
`gcc' failed in phase `Linker'. (Exit code: 1)

Does anyone know what's going on here? Is is annobin, new binutils, something else?
The earliest failing build I can find is git-annex [3], and the dependency changes show
annobin, binutils, and glibc (among others.)

[1] https://koji.fedoraproject.org/koji/taskinfo?taskID=28143935
[2] https://apps.fedoraproject.org/koschei/search?q=ghc-
[3] https://apps.fedoraproject.org/koschei/build/4976352

--
Elliott

[Matt Milosevic]

On Thu, Jul 12, 2018, 02:05 Elliott Sales de Andrade <
quantum.analyst(a)gmail.com&gt; wrote:

>  Does anyone know what's going on here? Is is annobin, new binutils,
>  something else? The earliest failing build I can find is git-annex [3], and
>  the dependency changes show annobin, binutils, and glibc (among others.)
>  
Annobin has been causing numerous issues with other packages recently, so
it is very likely this is another one.

>

[Florian Weimer]

On 07/12/2018 02:04 AM, Elliott Sales de Andrade wrote:
>  Does anyone know what's going on here? Is is annobin, new
> binutils, something else? The earliest failing build I can find is git-annex [3], and the
> dependency changes show annobin, binutils, and glibc (among others.) 
It's a change in the glibc startup code, which triggers something which 
looks like a bug in the gold linker.  I filed a non-Haskell reproducer here:

   https://bugzilla.redhat.com/show_bug.cgi?id=1600431

Thanks,
Florian

[Florian Weimer]

On 07/12/2018 10:44 AM, Florian Weimer wrote:
>  On 07/12/2018 02:04 AM, Elliott Sales de Andrade wrote:
> > Does anyone know what's going on here? Is is annobin, new binutils, 
> > something else? The earliest failing build I can find is git-annex 
> > [3], and the dependency changes show annobin, binutils, and glibc 
> > (among others.)
>  
>  It's a change in the glibc startup code, which triggers something which 
>  looks like a bug in the gold linker.  I filed a non-Haskell reproducer 
>  here:
>  
>     https://bugzilla.redhat.com/show_bug.cgi?id=1600431 
Nick patched binutils, and I rebuilt ghc-optional-args-1.0.2-1.fc29 
without failure, so this should be fixed now.

Thanks,
Florian

[Miro Hrončok]

On 12.7.2018 18:48, Florian Weimer wrote:
>  On 07/12/2018 10:44 AM, Florian Weimer wrote:
> > On 07/12/2018 02:04 AM, Elliott Sales de Andrade wrote:
> >> Does anyone know what's going on here? Is is annobin, new binutils, 
> >> something else? The earliest failing build I can find is git-annex 
> >> [3], and the dependency changes show annobin, binutils, and glibc 
> >> (among others.)
> >
> > It's a change in the glibc startup code, which triggers something 
> > which looks like a bug in the gold linker.  I filed a non-Haskell 
> > reproducer here:
> >
> >    https://bugzilla.redhat.com/show_bug.cgi?id=1600431
>  
>  Nick patched binutils, and I rebuilt ghc-optional-args-1.0.2-1.fc29 
>  without failure, so this should be fixed now. 
I now get:

Linking dist/build/tests-asn1-encoding/tests-asn1-encoding ...
/tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x18): error: 
relocation refers to local symbol "" [11], which is defined in a 
discarded section
   section group signature: "(null)"
/tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x20): error: 
relocation refers to local symbol "" [2], which is defined in a 
discarded section
   section group signature: "(null)"
collect2: error: ld returned 1 exit status
`gcc' failed in phase `Linker'. (Exit code: 1)

For various packages. See
https://bugzilla.redhat.com/show_bug.cgi?id=1607054#c6

Is this related?

-- 
Miro Hrončok
--
Phone: [phone]
IRC: mhroncok

[Florian Weimer]

On 07/23/2018 07:24 PM, Miro Hrončok wrote:
>  
>  
>  On 12.7.2018 18:48, Florian Weimer wrote:
> > On 07/12/2018 10:44 AM, Florian Weimer wrote:
> >> On 07/12/2018 02:04 AM, Elliott Sales de Andrade wrote:
> >>> Does anyone know what's going on here? Is is annobin, new binutils, 
> >>> something else? The earliest failing build I can find is git-annex 
> >>> [3], and the dependency changes show annobin, binutils, and glibc 
> >>> (among others.)
> >>
> >> It's a change in the glibc startup code, which triggers something 
> >> which looks like a bug in the gold linker.  I filed a non-Haskell 
> >> reproducer here:
> >>
> >>    https://bugzilla.redhat.com/show_bug.cgi?id=1600431
> >
> > Nick patched binutils, and I rebuilt ghc-optional-args-1.0.2-1.fc29 
> > without failure, so this should be fixed now.
>  
>  I now get:
>  
>  Linking dist/build/tests-asn1-encoding/tests-asn1-encoding ...
>  /tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x18): error: 
>  relocation refers to local symbol "" [11], which is defined in a 
>  discarded section
>     section group signature: "(null)"
>  /tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x20): error: 
>  relocation refers to local symbol "" [2], which is defined in a 
>  discarded section
>     section group signature: "(null)"
>  collect2: error: ld returned 1 exit status
>  `gcc' failed in phase `Linker'. (Exit code: 1)
>  
>  For various packages. See
>  https://bugzilla.redhat.com/show_bug.cgi?id=1607054#c6
>  
>  Is this related? 
Nick, would you please have a look?

Thanks,
Florian

[Miro Hrončok]

On 23.7.2018 19:28, Florian Weimer wrote:
>  On 07/23/2018 07:24 PM, Miro Hrončok wrote:
> >
> >
> > On 12.7.2018 18:48, Florian Weimer wrote:
> >> On 07/12/2018 10:44 AM, Florian Weimer wrote:
> >>> On 07/12/2018 02:04 AM, Elliott Sales de Andrade wrote:
> >>>> Does anyone know what's going on here? Is is annobin, new binutils, 
> >>>> something else? The earliest failing build I can find is git-annex 
> >>>> [3], and the dependency changes show annobin, binutils, and glibc 
> >>>> (among others.)
> >>>
> >>> It's a change in the glibc startup code, which triggers something 
> >>> which looks like a bug in the gold linker.  I filed a non-Haskell 
> >>> reproducer here:
> >>>
> >>>    https://bugzilla.redhat.com/show_bug.cgi?id=1600431
> >>
> >> Nick patched binutils, and I rebuilt ghc-optional-args-1.0.2-1.fc29 
> >> without failure, so this should be fixed now.
> >
> > I now get:
> >
> > Linking dist/build/tests-asn1-encoding/tests-asn1-encoding ...
> > /tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x18): 
> > error: relocation refers to local symbol "" [11], which is defined in 
> > a discarded section
> >    section group signature: "(null)"
> > /tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x20): 
> > error: relocation refers to local symbol "" [2], which is defined in a 
> > discarded section
> >    section group signature: "(null)"
> > collect2: error: ld returned 1 exit status
> > `gcc' failed in phase `Linker'. (Exit code: 1)
> >
> > For various packages. See
> > https://bugzilla.redhat.com/show_bug.cgi?id=1607054#c6
> >
> > Is this related?
>  
>  Nick, would you please have a look? 
Note that sticking "%undefine _annotated_build" to the top of the 
specfile doesn't solve that.

-- 
Miro Hrončok
--
Phone: [phone]
IRC: mhroncok

[Miro Hrončok]

On 23.7.2018 19:34, Miro Hrončok wrote:
>  On 23.7.2018 19:28, Florian Weimer wrote:
> > On 07/23/2018 07:24 PM, Miro Hrončok wrote:
> >>
> >>
> >> On 12.7.2018 18:48, Florian Weimer wrote:
> >>> On 07/12/2018 10:44 AM, Florian Weimer wrote:
> >>>> On 07/12/2018 02:04 AM, Elliott Sales de Andrade wrote:
> >>>>> Does anyone know what's going on here? Is is annobin, new 
> >>>>> binutils, something else? The earliest failing build I can find is 
> >>>>> git-annex [3], and the dependency changes show annobin, binutils, 
> >>>>> and glibc (among others.)
> >>>>
> >>>> It's a change in the glibc startup code, which triggers something 
> >>>> which looks like a bug in the gold linker.  I filed a non-Haskell 
> >>>> reproducer here:
> >>>>
> >>>>    https://bugzilla.redhat.com/show_bug.cgi?id=1600431
> >>>
> >>> Nick patched binutils, and I rebuilt ghc-optional-args-1.0.2-1.fc29 
> >>> without failure, so this should be fixed now.
> >>
> >> I now get:
> >>
> >> Linking dist/build/tests-asn1-encoding/tests-asn1-encoding ...
> >> /tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x18): 
> >> error: relocation refers to local symbol "" [11], which is defined in 
> >> a discarded section
> >>    section group signature: "(null)"
> >> /tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x20): 
> >> error: relocation refers to local symbol "" [2], which is defined in 
> >> a discarded section
> >>    section group signature: "(null)"
> >> collect2: error: ld returned 1 exit status
> >> `gcc' failed in phase `Linker'. (Exit code: 1)
> >>
> >> For various packages. See
> >> https://bugzilla.redhat.com/show_bug.cgi?id=1607054#c6
> >>
> >> Is this related?
> >
> > Nick, would you please have a look?
>  
>  Note that sticking "%undefine _annotated_build" to the top of the 
>  specfile doesn't solve that. 
Forget that statement, I haven't recreated the SRPM. It actually fixes 
the failure.

-- 
Miro Hrončok
--
Phone: [phone]
IRC: mhroncok

[Nick Clifton]

Hi Florian,

> > Linking dist/build/tests-asn1-encoding/tests-asn1-encoding ...
> > /tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x18): error: relocation
> refers to local symbol "" [11], which is defined in a discarded section
> >    section group signature: "(null)" 
This should now be fixed.  (binutils-2.31.1-3.fc29)

The problem was an extension of the problem originally reported in BZ 1600431.
The gold linker is complaining about relocations from the annobin notes to
discarded code sections.  This particular version of the problem arose because
I patched annobin to create section groups containing both the notes and the
code, but I did not update the patch to gold to allow it to detect these
sections.  Anyway it should all be fixed now.

Cheers
  Nick

[Miro Hrončok]

On 24.7.2018 11:53, Nick Clifton wrote:
>  Hi Florian,
>  
> >> Linking dist/build/tests-asn1-encoding/tests-asn1-encoding ...
> >> /tmp/ghc8eea_0/ghc_2.o(.gnu.build.attributes..text.startup+0x18): error:
> relocation refers to local symbol "" [11], which is defined in a discarded
> section
> >>     section group signature: "(null)"
>  
>  This should now be fixed.  (binutils-2.31.1-3.fc29)
>  
>  The problem was an extension of the problem originally reported in BZ 1600431.
>  The gold linker is complaining about relocations from the annobin notes to
>  discarded code sections.  This particular version of the problem arose because
>  I patched annobin to create section groups containing both the notes and the
>  code, but I did not update the patch to gold to allow it to detect these
>  sections.  Anyway it should all be fixed now.
>   
Thanks.

-- 
Miro Hrončok
--
Phone: [phone]
IRC: mhroncok