[Bug default/28301] New: libabigail asserts with older versions of gcc / glibc

[Bug default/28301] New: libabigail asserts with older versions of gcc / glibc

[jjardon at gnome dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

            Bug ID: 28301
           Summary: libabigail asserts with older versions of gcc / glibc
           Product: libabigail
           Version: unspecified
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: default
          Assignee: dodji at redhat dot com
          Reporter: jjardon at gnome dot org
                CC: libabigail at sourceware dot org
  Target Milestone: ---

libabigail version:
https://sourceware.org/git/?p=libabigail.git;a=commit;h=ca08bae7422bb5e2a972a2ef88524c67a2811fbc

This version works fine in our current master branch, but It's failing in our
previous stable version (release/20.08):

Package versions on those branches:
- master:
  - GCC: gcc-11.2.0
  - glibc: glibc-2.33-49-g22d37364aeaf5f8e4c5108677baaf084871c3006
- relase/20.08:
  - GCC: gcc-10.2.0
  - glibc: glibc-2.31-77-g0c92f409a7b078a50eb7e63ff57eb1e7ef57f90d

Downstream issue:
https://gitlab.com/freedesktop-sdk/freedesktop-sdk/-/issues/1310

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[gprocida at google dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

gprocida at google dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |gprocida at google dot com

--- Comment #1 from gprocida at google dot com ---
It would be most useful to attach or link directly some libraries that cause
the problem.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[jjardon at gnome dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

--- Comment #2 from Javier Jardón <jjardon at gnome dot org> ---
There is a log in the downstream issue with the libraries that are reporting
problems, copying here:

┌────────────────────────┐
│ ABI Break: libanl.so.1 │
└────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:3728: abigail::ir::function_type_sptr
abigail::dwarf_reader::read_context::lookup_fn_type_from_die_repr_per_tu(const
Dwarf_Die*): Assertion `__abg_cond__' failed.
┌────────────────────────┐
│ ABI Break: libnsl.so.1 │
└────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:3728: abigail::ir::function_type_sptr
abigail::dwarf_reader::read_context::lookup_fn_type_from_die_repr_per_tu(const
Dwarf_Die*): Assertion `__abg_cond__' failed.
┌────────────────────────────┐
│ ABI Break: libpthread.so.0 │
└────────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:10859: bool
abigail::dwarf_reader::get_scope_die(const
abigail::dwarf_reader::read_context&, const Dwarf_Die*, size_t, Dwarf_Die&):
Assertion `__abg_cond__' failed.
┌──────────────────────────┐
│ ABI Break: libcrypt.so.1 │
└──────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:3728: abigail::ir::function_type_sptr
abigail::dwarf_reader::read_context::lookup_fn_type_from_die_repr_per_tu(const
Dwarf_Die*): Assertion `__abg_cond__' failed.
┌─────────────────────────┐
│ ABI Break: libutil.so.1 │
└─────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:3728: abigail::ir::function_type_sptr
abigail::dwarf_reader::read_context::lookup_fn_type_from_die_repr_per_tu(const
Dwarf_Die*): Assertion `__abg_cond__' failed.
┌───────────────────────┐
│ ABI Break: libdl.so.2 │
└───────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:15015:
abigail::ir::type_or_decl_base_sptr
abigail::dwarf_reader::build_ir_node_from_die(abigail::dwarf_reader::read_context&,
Dwarf_Die*, abigail::ir::scope_decl*, bool, size_t, bool, bool): Assertion
`__abg_cond__' failed.
┌───────────────────────────┐
│ ABI Break: libnss_db.so.2 │
└───────────────────────────┘
in build_ir_node_from_die at: ../../src/abg-dwarf-reader.cc:15182: execution
should not have reached this point!
┌───────────────────────────┐
│ ABI Break: libJISX0213.so │
└───────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:9024: std::string
abigail::dwarf_reader::die_qualified_type_name(const
abigail::dwarf_reader::read_context&, const Dwarf_Die*, size_t): Assertion
`__abg_cond__' failed.
┌──────────────────────────────┐
│ ABI Break: libnss_files.so.2 │
└──────────────────────────────┘
in build_ir_node_from_die at: ../../src/abg-dwarf-reader.cc:15182: execution
should not have reached this point!
┌───────────────────────────────┐
│ ABI Break: libnss_hesiod.so.2 │
└───────────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:3728: abigail::ir::function_type_sptr
abigail::dwarf_reader::read_context::lookup_fn_type_from_die_repr_per_tu(const
Dwarf_Die*): Assertion `__abg_cond__' failed.
┌───────────────────────────────┐
│ ABI Break: libnss_compat.so.2 │
└───────────────────────────────┘
in build_ir_node_from_die at: ../../src/abg-dwarf-reader.cc:15182: execution
should not have reached this point!
┌──────────────────────────────┐
│ ABI Break: libthread_db.so.1 │
└──────────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:9024: std::string
abigail::dwarf_reader::die_qualified_type_name(const
abigail::dwarf_reader::read_context&, const Dwarf_Die*, size_t): Assertion
`__abg_cond__' failed.
┌──────────────────────┐
│ ABI Break: libc.so.6 │
└──────────────────────┘
in build_ir_node_from_die at: ../../src/abg-dwarf-reader.cc:15182: execution
should not have reached this point!
┌───────────────────────┐
│ ABI Break: librt.so.1 │
└───────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:3728: abigail::ir::function_type_sptr
abigail::dwarf_reader::read_context::lookup_fn_type_from_die_repr_per_tu(const
Dwarf_Die*): Assertion `__abg_cond__' failed.
┌───────────────────────────┐
│ ABI Break: libresolv.so.2 │
└───────────────────────────┘
abidiff: ../../src/abg-dwarf-reader.cc:3728: abigail::ir::function_type_sptr
abigail::dwarf_reader::read_context::lookup_fn_type_from_die_repr_per_tu(const
Dwarf_Die*): Assertion `__abg_cond__' failed.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[seppo.yliolli at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

Seppo Yli-Olli <seppo.yliolli at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |seppo.yliolli at gmail dot com

--- Comment #3 from Seppo Yli-Olli <seppo.yliolli at gmail dot com> ---
Okay. So I just reproduced same asserts with by staging our ABI image twice
(identical files) in and running ABI checker between them. Since bunch of these
files seem to be present in glibc, added
https://nanonyme.kapsi.fi/public/glibc.tar.xz here. I'm not sure if I should be
uploading this large packages into the bugzilla.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[dodji at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

dodji at redhat dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2021-10-04
     Ever confirmed|0                           |1

--- Comment #4 from dodji at redhat dot com ---
(In reply to Seppo Yli-Olli from comment #3)
> Okay. So I just reproduced same asserts with by staging our ABI image twice
> (identical files) in and running ABI checker between them. Since bunch of
> these files seem to be present in glibc, added
> https://nanonyme.kapsi.fi/public/glibc.tar.xz here. I'm not sure if I should
> be uploading this large packages into the bugzilla.

Hello.

For me to reproduce the issue and be able to debug it, it's helpful to provide
a command line to reproduce the issue, like 'abidiff binary1 binary2'.

I have downloaded the tarball at https://nanonyme.kapsi.fi/public/glibc.tar.xz.
 Thank you for having provided it.  But then, how am I supposed to reproduce
the problem exactly?

Also, have you tried to reproduce the issue using libabigail from the master
repository?

The libabigail version from master could handle the binaries from the tarball
that your version of libabigail was failing from, if I believe one of the
comments of the bug.  Here is what I did with the current libabigail from
master to test for that:

$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libanl.so.1
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libnsl.so.1
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libpthread.so.0
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libpthread.so.0
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libcrypt.so.1
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libutil.so.1
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libdl.so.2
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libnss_db.so.2
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/gconv/libJISX0213.so
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libnss_files.so.2
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libnss_hesiod.so.2
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libnss_compat.so.2
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libthread_db.so.1
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libc.so.6
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/librt.so.1
$ abidw --abidiff glibc/usr/lib/x86_64-linux-gnu/libresolv.so.2
$

I think that I really need to release what we have in master now so that you
can update to that one instead as there are many many fixes stacked there.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[seppo.yliolli at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

--- Comment #5 from Seppo Yli-Olli <seppo.yliolli at gmail dot com> ---
We already switched to master a while back to facilitate this being fixed
faster. We have docker image at
registry.gitlab.com/freedesktop-sdk/infrastructure/freedesktop-sdk-docker-images/bst16
that reproduces the issue.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[seppo.yliolli at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

--- Comment #6 from Seppo Yli-Olli <seppo.yliolli at gmail dot com> ---
No, you understood comments wrong.
1. Our master and 21.08 branches use newer GCC and glibc
2. We build docker image with libabigail master based on our branch 21.08
3. Our 20.08 use older GCC and glibc
4. Libagail works correctly in our newer branches, asserts with older
5. Older libabigail seemed to similarly assert on 20.08 when we tried to
downgrade

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[dodji at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

--- Comment #7 from dodji at redhat dot com ---
okay, can I have a reproducer that is not a docker image, please?
Just the binaries to run abidiff on, please please.  I can't integrate a docker
image in libabigail testsuite, for instance.  This would have been much faster
if you could just give me the two binaries that abidiff is failing on.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[seppo.yliolli at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

--- Comment #8 from Seppo Yli-Olli <seppo.yliolli at gmail dot com> ---
Yes, I cannot do that. But I did give you libabigail in docker image that
reproduces the problem for us with the binaries in the tarball.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[seppo.yliolli at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

--- Comment #9 from Seppo Yli-Olli <seppo.yliolli at gmail dot com> ---
It's still this same case I was talking on IRC about except as said, it turns
out this it's a lot more subtle and can be triggered with smaller differences
of glibc. Same asserts going on. It appears libabigail itself (at least in our
case) is affected by the glibc it's built against which makes it impossible to
for abidw'ing another glibc. Something perhaps worth mentioning is we're using
the separate debugedit project, not RPM debugedit.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[ben at demerara dot io]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

--- Comment #10 from Ben Brown <ben at demerara dot io> ---
Here's a link to various tarballs with libraries built with two different
versions of glibc that cause abidiff to break (the versions of glibc are from
the same stable branch, 4 commits apart):
https://gitlab.com/freedesktop-sdk/freedesktop-sdk/-/jobs/4162744139/artifacts/browse/libabigail-tars/

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug default/28301] libabigail asserts with older versions of gcc / glibc

[dodji at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28301

dodji at redhat dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|WAITING                     |RESOLVED

--- Comment #11 from dodji at redhat dot com ---
Thank you for providing me with the tarballs so that I can reproduce the issue.

With the current state of the master branch, it seems abidiff is now properly
comparing the binaries provided in the six tarballs at
https://gitlab.com/freedesktop-sdk/freedesktop-sdk/-/jobs/4162744139/artifacts/browse/libabigail-tars/.

The hash of the tree that I used for testing is this one:

6ba26ed6 Bug 30309 - Support absolute path to alt debug info file in DWARF

The fix will be present in the 2.3 libabigail release.

Thanks again for taking the time to report this problem and sorry

-- 
You are receiving this mail because:
You are on the CC list for the bug.