Debian Bug report logs - #718047
ia64, ld segfault on --as-needed

Reply or subscribe to this bug.

Display info messages

Message #5 received at [email protected] (full text, mbox, reply):

From: Mathieu Malaterre <[email protected]>

To: Debian Bug Tracking System <[email protected]>

Subject: binutils: assertion fail elf64-ia64.c:2723

Date: Sun, 28 Jul 2013 09:46:05 +0200

Package: binutils
Severity: grave
Justification: renders package unusable

Package gdcm fails to build on ia64. Errors is:

Linking CXX shared library ../../bin/libvtkgdcm.so
cd /build/buildd-gdcm_2.2.4-1-ia64-JJJW9m/gdcm-2.2.4/obj-ia64-linux-gnu/Utilities/VTK && /usr/bin/cmake -E cmake_link_script CMakeFiles/vtkgdcm.dir/link.txt --verbose=1
/usr/bin/c++  -fPIC -g -O2 -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2  -DNDEBUG  -Wno-deprecated  -Wl,--as-needed  -shared -Wl,-soname,libvtkgdcm.so.2.2 -o ../../bin/libvtkgdcm.so.2.2.4 CMakeFiles/vtkgdcm.dir/vtkGDCMTesting.cxx.o CMakeFiles/vtkgdcm.dir/vtkGDCMImageReader.cxx.o CMakeFiles/vtkgdcm.dir/vtkGDCMImageWriter.cxx.o CMakeFiles/vtkgdcm.dir/vtkGDCMMedicalImageProperties.cxx.o CMakeFiles/vtkgdcm.dir/vtkImageMapToWindowLevelColors2.cxx.o CMakeFiles/vtkgdcm.dir/vtkImageYBRToRGB.cxx.o CMakeFiles/vtkgdcm.dir/vtkImageRGBToYBR.cxx.o CMakeFiles/vtkgdcm.dir/vtkGDCMPolyDataReader.cxx.o CMakeFiles/vtkgdcm.dir/vtkGDCMPolyDataWriter.cxx.o CMakeFiles/vtkgdcm.dir/vtkRTStructSetProperties.cxx.o CMakeFiles/vtkgdcm.dir/vtkLookupTable16.cxx.o CMakeFiles/vtkgdcm.dir/vtkImageMapToColors16.cxx.o CMakeFiles/vtkgdcm.dir/vtkImageColorViewer.cxx.o CMakeFiles/vtkgdcm.dir/vtkGDCMThreadedImageReader2.cxx.o CMakeFiles/vtkgdcm.dir/vtkgdcmInstantiator.cxx.o ../../bin/libgdcmMSFF.so.2.2.4 /us
 r/lib/li
 bvtkCommon.so.5.8.0 /usr/lib/libvtkIO.so.5.8.0 /usr/lib/libvtkImaging.so.5.8.0 /usr/lib/libvtkRendering.so.5.8.0 -lpthread ../../bin/libgdcmDICT.so.2.2.4 ../../bin/libgdcmIOD.so.2.2.4 ../../bin/libgdcmDSED.so.2.2.4 ../../bin/libgdcmCommon.so.2.2.4 /usr/lib/libvtkIO.so.5.8.0 /usr/lib/libvtkImaging.so.5.8.0 /usr/lib/libvtkGraphics.so.5.8.0 /usr/lib/libvtkFiltering.so.5.8.0 /usr/lib/libvtkCommon.so.5.8.0 /usr/lib/libvtksys.so.5.8.0 -ldl -lm -Wl,-rpath-link,/build/buildd-gdcm_2.2.4-1-ia64-JJJW9m/gdcm-2.2.4/obj-ia64-linux-gnu/bin 
collect2: ld terminated with signal 11 [Segmentation fault]
/usr/bin/ld: BFD (GNU Binutils for Debian) 2.23.52.20130722 assertion fail elf64-ia64.c:2723
make[3]: *** [bin/libvtkgdcm.so.2.2.4] Error 1
make[3]: Leaving directory `/build/buildd-gdcm_2.2.4-1-ia64-JJJW9m/gdcm-2.2.4/obj-ia64-linux-gnu'

ref:
https://buildd.debian.org/status/fetch.php?pkg=gdcm&arch=ia64&ver=2.2.4-1&stamp=1374928496

Message #14 received at [email protected] (full text, mbox, reply):

From: Stephan Schreiber <[email protected]>

To: [email protected]

Cc: Mathieu Malaterre <[email protected]>

Subject: ia64, ld segfault on --as-needed

Date: Sun, 25 Aug 2013 12:18:05 +0200

retitle 718047 ia64, ld segfault on --as-needed
found 718047 2.23.52.20130727-1
tags 718047 - sid jessie
thanks

ia64.

The gdcm package isn't the only one which is affected by that ld bug.
Another example is the iceweasel 17.0.8esr-1 package; ld crashes on  
the test suite.

The bug is triggered by the --as-needed switch.

The problem is in bfd/elflink.c in the function elf_link_add_object_symbols().
When an input file is linked with the ---as-needed in effect, the  
elf_link_add_object_symbols() function
(1) clones the symbol table;
(2) adds the new information
(3) restores the symbol table if it turns out that the new input is  
*not* needed.

The problem is that on bullet (2) things are modified that aren't  
restored on bullet (3).
The main problem is that elf_link_add_object_symbols() calls the
*bed->elf_backend_copy_indirect_symbol()
function both directly and in a nested way on some calls of  
_bfd_elf_merge_symbol().
elf_backend_copy_indirect_symbol is elfNN_ia64_hash_copy_indirect() on  
ia64 (in /bfd/elfnn-ia64.c).
elfNN_ia64_hash_copy_indirect() moves the info about the usage of  
dynamic symbols from one symbol to another; it is an array of struct  
elfNN_ia64_link_hash_entry in separately allocated memory.
It sets the ->info member of the symbol to NULL and adjusts the ->h  
member of all dynamic symbols to the target symbol.
When elf_link_add_object_symbols() restores the symbol table (3), the  
->info member (which was NULL) is set to the former array of dynamic  
symbols, but the ->h members of the dynmaic symbols point to another  
symbol.
The other symbol is deleted (freed memory); when the implementation  
walks through the symbols in a later stage, the crash occurs. (The  
failed assert is just a side effect.)

There is another thing which is modified on (2) but isn't restored on (3):
The bfd_elf_link_record_dynamic_symbol() function is called, which may  
add a new string to the string table via _bfd_elf_strtab_add().
Since the string table isn't restored by elf_link_add_object_symbols  
(3), the result is a failing assert on writing out the string table on  
a later stage:
bfd/elf-strtab.c:273:
      BFD_ASSERT (tab->array[i]->refcount == 0);


I think we should talk with the upstream about the problem because I  
don't have a good idea how to fix it at the moment.
Comments are appreciated.

Stephan

Message #27 received at [email protected] (full text, mbox, reply):

From: Mathieu Malaterre <[email protected]>

To: Control bugs server <[email protected]>, [email protected]

Date: Tue, 3 Sep 2013 08:35:18 +0200

affects 718047  + src:vxl
thanks

I think this also affects VXL:

https://buildd.debian.org/status/fetch.php?pkg=vxl&arch=ia64&ver=1.17.0-6&stamp=1378127208

Linking CXX shared library ../../../lib/libvipl.so
cd /«PKGBUILDDIR»/obj-ia64-linux-gnu/contrib/tbl/vipl &&
/usr/bin/cmake -E cmake_link_script CMakeFiles/vipl.dir/link.txt
--verbose=1
/usr/bin/c++  -fPIC -g -O2 -Wformat -Werror=format-security
-D_FORTIFY_SOURCE=2    -Wl,--as-needed  -shared
-Wl,-soname,libvipl.so.1.17 -o ../../../lib/libvipl.so.1.17.0
CMakeFiles/vipl.dir/filter/vipl_filter_abs.o
../../../lib/libvil.so.1.17.0 ../../../lib/libvil1.so.1.17.0
../../../lib/libvnl.so.1.17.0 ../../../lib/libvbl.so.1.17.0 -lgeotiff
../../../lib/libdcmtk.so.1.17.0 ../../../lib/libopenjpeg2.so.2.0.0
-ljpeg -ltiff -lpng -lz ../../../lib/libvcl.so.1.17.0 -lm
-Wl,-rpath,/«PKGBUILDDIR»/obj-ia64-linux-gnu/lib:
collect2: ld terminated with signal 11 [Segmentation fault]
make[3]: *** [lib/libvipl.so.1.17.0] Error 1
make[3]: Leaving directory `/«PKGBUILDDIR»/obj-ia64-linux-gnu'
make[2]: *** [contrib/tbl/vipl/CMakeFiles/vipl.dir/all] Error 2
make[2]: *** Waiting for unfinished jobs....

Message #34 received at [email protected] (full text, mbox, reply):

From: Matthias Klose <[email protected]>

To: [email protected]

Cc: Mathieu Malaterre <[email protected]>, [email protected]

Subject: Re: ia64, ld segfault on --as-needed

Date: Tue, 17 Sep 2013 10:56:15 +0200

Control: severity -1 normal
Control: tags -1 + moreinfo help

filing this report with this severity, for a non-working non-default option
seems to be wrong.

not including the object files (including the shared objects) doesn't help,
these are missing in the upstream report as well.

Message #43 received at [email protected] (full text, mbox, reply):

From: Mathieu Malaterre <[email protected]>

To: Matthias Klose <[email protected]>, [email protected]

Cc: [email protected], [email protected]

Subject: Re: ia64, ld segfault on --as-needed

Date: Wed, 18 Sep 2013 10:53:53 +0200

On Tue, Sep 17, 2013 at 10:56 AM, Matthias Klose <[email protected]> wrote:

> filing this report with this severity, for a non-working non-default option
> seems to be wrong.

This is a default option at least on one derivative: Ubuntu
https://wiki.debian.org/qa.debian.org/FTBFS

This is also a recommended flag AFAIK:
https://wiki.debian.org/HardeningWalkthrough

> not including the object files (including the shared objects) doesn't help,
> these are missing in the upstream report as well.

I am missing your point here. Ideally what should I report upstream ?

Thanks.

Message #54 received at [email protected] (full text, mbox, reply):

From: "Manuel A. Fernandez Montecelo" <[email protected]>

To: [email protected]

Subject: ia64, ld segfault on --as-needed

Date: Fri, 17 Jan 2014 20:21:27 +0000

affects 718047 + src:ogre-1.9 src:openscenegraph
stop

This affects ogre-1.9 (see fix below [1]) and openscenegraph uploaded
recently [2].


[1] http://anonscm.debian.org/gitweb/?p=pkg-games/ogre-1.9.git;a=commitdiff;h=23eb0ca4fac15f5b300c82be130e729deda4cdc9#patch3

[2] https://buildd.debian.org/status/fetch.php?pkg=openscenegraph&arch=ia64&ver=3.2.0%7Erc1-2&stamp=1389850577

-- 
Manuel A. Fernandez Montecelo <[email protected]>

Message #61 received at [email protected] (full text, mbox, reply):

From: Mathieu Malaterre <[email protected]>

To: [email protected]

Date: Fri, 12 Sep 2014 17:59:58 +0200

IA64 has been removed from jessie. No way it is coming back. closing.

ref:
https://lists.debian.org/debian-devel-announce/2014/01/msg00009.html

Send a report that this bug log contains spam.