Bug 29934

Summary: abidiff coredump at abg-comparison.cc:ensure_lookup_tables_populated when read libdovecot-storage.so and libdovecot-sieve.so from dovecot
Product: libabigail Reporter: Xiaole He <hexiaole1994>
Component: defaultAssignee: Dodji Seketeli <dodji>
Status: RESOLVED FIXED    
Severity: critical CC: hexiaole1994, libabigail
Priority: P2    
Version: unspecified   
Target Milestone: ---   
Host: Target:
Build: Last reconfirmed: 2022-12-24 00:00:00

Description Xiaole He 2022-12-22 11:49:14 UTC
[Overview]
Executing 'abidiff' against to 'libc.so' from centos7.6's 'glibc' will cause coredump, the output from 'bt' command of 'gdb' shows below:
/* gdb output for 'libdovecot-storage.so' begin */
(gdb) bt
#0  0x0000fffc0df966e0 in raise () from /lib64/libc.so.6
#1  0x0000fffc0df97a8c in abort () from /lib64/libc.so.6
#2  0x0000fffc0df8fb8c in ?? () from /lib64/libc.so.6
#3  0x0000fffc0df8fc0c in __assert_fail () from /lib64/libc.so.6
#4  0x0000fffc0e80ae88 in abigail::comparison::class_or_union_diff::ensure_lookup_tables_populated (this=this@entry=0x15602410) at abg-comparison.cc:4520
#5  0x0000fffc0e8027dc in abigail::comparison::class_diff::ensure_lookup_tables_populated (this=0x15602410) at abg-comparison.cc:5175
#6  0x0000fffc0e80416c in abigail::comparison::compute_diff (first=std::shared_ptr (count 17, weak 15) 0x1372e400, second=std::shared_ptr (count 29, weak 35) 0x13a61eb0,
    ctxt=std::shared_ptr (count 17, weak 3407) 0x12f52be0) at abg-comparison.cc:5756
...
#15 0x0000fffc0e807430 in abigail::comparison::compute_diff (first=std::shared_ptr (count 2, weak 0) 0x1400c2e0, second=std::shared_ptr (count 2, weak 0) 0x15249250,
    ctxt=std::shared_ptr (count 17, weak 3407) 0x12f52be0) at abg-comparison.cc:7118
#16 0x0000fffc0e80cdf4 in abigail::comparison::corpus_diff::priv::ensure_lookup_tables_populated (this=0x13bc1e50) at abg-comparison.cc:8811
#17 0x0000fffc0e80e9dc in abigail::comparison::compute_diff (f=std::shared_ptr (count 3, weak 0) 0x12f523c0, s=std::shared_ptr (count 3, weak 0) 0x1491a920,
    ctxt=std::shared_ptr (count 17, weak 3407) 0x12f52be0) at abg-comparison.cc:11019
#18 0x0000000000406148 in main (argc=<optimized out>, argv=<optimized out>) at abidiff.cc:1409
/* gdb output end */
(the backtrace of 'libdovecot-sieve.so' is similar as 'libdovecot-storage.so', thus it was skipped here.)

[Steps to Reproduce]
(1)confirm that we at the latest master branch
# git remote -v
origin  git://sourceware.org/git/libabigail.git (fetch)
origin  git://sourceware.org/git/libabigail.git (push)
# git branch
* master
# git log -5 --oneline
83bbc679 (HEAD -> master, origin/master, origin/HEAD) elf-reader: reclaim fd and mem before break
e1aadfa5 Bug 29901 - abidiff hangs when comparing libgs.so.10 with itself
3c6a461b Update website documentation for 2.2
73d8782c configure: Bump version number to 2.3
7a7a00c3 ChangeLog: Update for 2.2 release
# ./configure
# make
(2)assure the newly compiled 'abidiff' is linked to the newly compiled 'libabigail.so'
# pwd
/root/tmp/libabigail-git
# ldd tools/.libs/abidiff
        ...
        libabigail.so.1 => /root/tmp/libabigail-git/src/.libs/libabigail.so.1 (0x0000fffbca4b0000)
        ...
(3)download 'libdovecot-storage.so' and 'libdovecot-sieve.so'
# mkdir dovecot/ && cd dovecot/
# wget https://vault.centos.org/7.6.1810/os/x86_64/Packages/dovecot-2.2.36-3.el7.x86_64.rpm
# wget http://debuginfo.centos.org/7/x86_64/dovecot-debuginfo-2.2.36-3.el7.x86_64.rpm
# wget https://repo.huaweicloud.com/openeuler/openEuler-20.03-LTS-SP2/everything/x86_64/Packages/dovecot-2.3.10.1-6.oe1.x86_64.rpm
# wget https://repo.huaweicloud.com/openeuler/openEuler-20.03-LTS-SP2/debuginfo/x86_64/Packages/dovecot-debuginfo-2.3.10.1-6.oe1.x86_64.rpm
# mkdir 2.2.36-with-debug/ 2.2.36-without-debug/ 2.3.10-with-debug/ 2.3.10-without-debug/
# rpm2cpio dovecot-2.2.36-3.el7.x86_64.rpm | cpio -idum -D 2.2.36-without-debug/
# rpm2cpio dovecot-debuginfo-2.2.36-3.el7.x86_64.rpm | cpio -idum -D 2.2.36-with-debug/
# rpm2cpio dovecot-2.3.10.1-6.oe1.x86_64.rpm | cpio -idum -D 2.3.10-without-debug/
# rpm2cpio dovecot-debuginfo-2.3.10.1-6.oe1.x86_64.rpm | cpio -idum -D 2.3.10-with-debug/
# cd ../
(4)use newly compiled 'abidiff' to reproducing the problem that causes coredump
# ./tools/.libs/abidiff --d1 dovecot/2.2.36-with-debug/usr/lib/debug/ --d2 dovecot/2.3.10-with-debug/usr/lib/debug/ dovecot/2.2.36-without-debug/usr/lib64/dovecot/libdovecot-storage.so.0.0.0 dovecot/2.3.10-without-debug/usr/lib64/dovecot/libdovecot-storage.so.0.0.0
# ./tools/.libs/abidiff --d1 dovecot/2.2.36-with-debug/usr/lib/debug/ --d2 dovecot/2.3.10-with-debug/usr/lib/debug/ dovecot/2.2.36-without-debug/usr/lib64/dovecot/libdovecot-sieve.so.0.0.0 dovecot/2.3.10-without-debug/usr/lib64/dovecot/libdovecot-sieve.so.0.0.0

[Actual Results]
(1)'abidiff' for 'libdovecot-storage.so' crashes with coredump:
/* crash output begin */
abidiff: abg-comparison.cc:4521: void abigail::comparison::class_or_union_diff::ensure_lookup_tables_populated() const: Assertion `__abg_cond__' failed.
Aborted (core dumped)
/* crash output end */

(2)'abidiff' for 'libdovecot-sieve.so' crashes with coredump:
/* crash output begin */
abidiff: abg-comparison.cc:4521: void abigail::comparison::class_or_union_diff::ensure_lookup_tables_populated() const: Assertion `__abg_cond__' failed.
Aborted (core dumped)
/* crash output end */

[Expected Results]
'abidiff' finishes normally with differences between two libraries.
Comment 1 Xiaole He 2022-12-22 11:55:01 UTC
Sorry, the 'Overview' was wrong above, corrected one is below:

[Overview]
Executing 'abidiff' against to 'libdovecot-storage.so' from rpm 'dovecot' will cause coredump, the output from 'bt' command of 'gdb' shows below:
Comment 2 Dodji Seketeli 2022-12-24 14:12:14 UTC
Hello,

This issue should hopefully be fixed in the master branch of the git repository at https://sourceware.org/git/?p=libabigail.git;a=commit;h=e3725f1b96ad2e935290bea8be7925f2a8cc3c89.

The fix should be available in the coming 2.3 version of libabigail.

Thank you for reporting this issue and sorry for the inconvenience.
Comment 3 Xiaole He 2022-12-26 03:40:51 UTC
(In reply to dodji from comment #2)
> Hello,
> 
> This issue should hopefully be fixed in the master branch of the git
> repository at
> https://sourceware.org/git/?p=libabigail.git;a=commit;
> h=e3725f1b96ad2e935290bea8be7925f2a8cc3c89.
> 
> The fix should be available in the coming 2.3 version of libabigail.
> 
> Thank you for reporting this issue and sorry for the inconvenience.

Thank you, dodji, for using your precious time in resolving this problem. 
I using the updated master branch of git, e3725f1b96ad2e935290bea8be7925f2a8cc3c89 (Bug 29934 - Handle buggy data members with empty names), to verify that the problem of 'libdovecot-storage.so' has been disappeared, but the similar problem of 'libdovecot-sieve.so' that caused coredump still exist, sorry for disturbing you again and the reproducing steps of 'libdovecot-storage.so' was wroten at the first comment of this page.
Comment 4 Dodji Seketeli 2022-12-26 14:13:13 UTC
(In reply to Xiaole He from comment #3)

> [...] the similar problem of
> 'libdovecot-sieve.so' that caused coredump still exist, sorry for disturbing
> you again and the reproducing steps of 'libdovecot-storage.so' was wroten at
> the first comment of this page.

Ooops, I totally missed the second issue.  That one is different from the one occurring on libdovecot-storage.so.

I'll be looking at the second issue.

Sorry for the noise.
Comment 5 Dodji Seketeli 2022-12-29 10:35:45 UTC
The second issue of the problem reported here (assertion violation when analysing the binary libdovecot-sieve.so) should be fixed by commit https://sourceware.org/git/?p=libabigail.git;a=commit;h=4f4ab2b1d66770f6e3cc56ec844e47a5ceec1edb.

Thanks for taking the time to report this problem and sorry for the inconvenience.
Comment 6 Xiaole He 2022-12-29 14:54:01 UTC
(In reply to dodji from comment #5)
> The second issue of the problem reported here (assertion violation when
> analysing the binary libdovecot-sieve.so) should be fixed by commit
> https://sourceware.org/git/?p=libabigail.git;a=commit;
> h=4f4ab2b1d66770f6e3cc56ec844e47a5ceec1edb.
> 
> Thanks for taking the time to report this problem and sorry for the
> inconvenience.

Thank you very much, dodji. I already confirmed that this problem has been fixed. professional man, great job!