[Bug localedata/18589] New: sort-test.sh fails at random

[Bug localedata/18589] New: sort-test.sh fails at random

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

            Bug ID: 18589
           Summary: sort-test.sh fails at random
           Product: glibc
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: localedata
          Assignee: unassigned at sourceware dot org
          Reporter: hjl.tools at gmail dot com
                CC: libc-locales at sourceware dot org
  Target Milestone: ---

On master with commit 8b1bab5ffa235bc494c33fdbe4c9994202936cfc,
sort-test.sh fails at random:

[hjl@gnu-6 localedata]$ /bin/sh sort-test.sh
/export/build/gnu/glibc/build-x86_64-linux/ ' env'  
'GCONV_PATH=/export/build/gnu/glibc/build-x86_64-linux/iconvdata
LOCPATH=/export/build/gnu/glibc/build-x86_64-linux/localedata LC_ALL=C' ''  
de_DE.ISO-8859-1 en_US.ISO-8859-1 da_DK.ISO-8859-1 hr_HR.ISO-8859-2
sv_SE.ISO-8859-1 tr_TR.UTF-8 fr_FR.UTF-8 si_LK.UTF-8 uk_UA.UTF-8
de_DE.ISO-8859-1 collate-test OK
de_DE.ISO-8859-1 xfrm-test OK
en_US.ISO-8859-1 collate-test OK
en_US.ISO-8859-1 xfrm-test OK
da_DK.ISO-8859-1 collate-test FAIL
  --- da_DK.in  2014-01-29 06:48:59.283186329 -0800
  +++ /export/build/gnu/glibc/build-x86_64-linux/localedata/da_DK.out  
2015-06-24 03:02:12.813754190 -0700
  @@ -1,21 +1,14 @@
   A
   ANDRE
  -ANDR
   ANDREAS
  -AS
  -A/S
   CA
  -\ufffdA
   CB
   \ufffdC
  -DA
   \ufffdA
   DB
   \ufffdC
  -DSB
   D.S.B.
   DSC
  -EKSTRA-ARBEJDE
   EKSTRABUD
   H\ufffdST
   HAAG
  @@ -58,4 +51,11 @@
   \ufffdBERG
   \ufffdBERG
   AA
  +ANDR
  +AS
  +A/S
  +\ufffdA
  +DA
  +DSB
  +EKSTRA-ARBEJDE
   aa
da_DK.ISO-8859-1 xfrm-test FAIL
hr_HR.ISO-8859-2 collate-test OK
hr_HR.ISO-8859-2 xfrm-test OK
sv_SE.ISO-8859-1 collate-test OK
sv_SE.ISO-8859-1 xfrm-test OK
tr_TR.UTF-8 collate-test OK
tr_TR.UTF-8 xfrm-test OK
fr_FR.UTF-8 collate-test OK
fr_FR.UTF-8 xfrm-test OK
si_LK.UTF-8 collate-test OK
si_LK.UTF-8 xfrm-test OK
uk_UA.UTF-8 collate-test OK
uk_UA.UTF-8 xfrm-test OK
[hjl@gnu-6 localedata]$ 

Sometimes it also passes.

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

[Bug localedata/18589] sort-test.sh fails at random

[jsm28 at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #1 from Joseph Myers <jsm28 at gcc dot gnu.org> ---
This is a regression from 2.21, caused by:

commit f13c2a8dff2329c6692a80176262ceaaf8a6f74e
Author: Leonhard Holz <leonhard.holz@web.de>
Date:   Tue May 12 11:37:52 2015 +0200

    Improve strcoll with strdiff.

    This patch improves strcoll hot case by finding first byte that
    mismatches. That is in likely case enough to determine comparison
    result.

sort-test.sh is randomized.  The following deterministic test reproduces this
reliably.  Note that AA should be treated as the last letter of the Danish
alphabet, hence sorting after AS.

#include <locale.h>
#include <stdio.h>
#include <string.h>

int
main (void)
{
  if (setlocale (LC_ALL, "da_DK.ISO-8859-1") == NULL)
    {
      perror ("setlocale");
      return 1;
    }
  int a = strcoll ("AS", "AA");
  int b = strcoll ("AA", "AS");
  printf ("strcoll (\"AS\", \"AA\") = %d\n", a);
  printf ("strcoll (\"AA\", \"AS\") = %d\n", b);
  if (a < 0 && b > 0)
    {
      puts ("PASS");
      return 0;
    }
  else
    {
      puts ("FAIL");
      return 1;
    }
}

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

[Bug locale/18589] sort-test.sh fails at random

[jsm28 at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

Joseph Myers <jsm28 at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|localedata                  |locale

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

[Bug locale/18589] sort-test.sh fails at random

[egmont at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

Egmont Koblinger <egmont at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |egmont at gmail dot com

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

[Bug locale/18589] sort-test.sh fails at random

[egmont at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

Egmont Koblinger <egmont at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Blocks|                            |18934


Referenced Bugs:

https://sourceware.org/bugzilla/show_bug.cgi?id=18934
[Bug 18934] [PATCH] Hungarian collate: fix multiple bugs and add tests
-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug locale/18589] sort-test.sh fails at random

[egmont at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #2 from Egmont Koblinger <egmont at gmail dot com> ---
This is indeed a bug, effecting multiple locales. Could you please revert it?

In bug 18934 I created an extensive test coverage for a newly fixed version of
hu_HU, which also fails due to this issue.

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

[Bug locale/18589] sort-test.sh fails at random

[egmont at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #4 from Egmont Koblinger <egmont at gmail dot com> ---
Thanks Carlos!

I wanted to investigate a bit but didn't have time, so sorry if I make wrong
assumptions:

Does this bug result in strcoll(a, b) returning different result than
[pseudocode] strcmp(strxfrm(a), strxfrm(b))?

I would have guessed the former is just a thin wrapper performing the latter;
apparently it has its standalone implementation for performance reasons which
is error prone to let the two go out of sync, am I right?

Wouldn't it make sense to extend the unittests to sort the files with both
comparison approaches? (Although that'd still not give full protection against
weird undiscovered corner cases where the two methods produce different
results.)

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

[Bug locale/18589] sort-test.sh fails at random

[carlos at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #5 from Carlos O'Donell <carlos at redhat dot com> ---
(In reply to Egmont Koblinger from comment #4)
> Thanks Carlos!
> 
> I wanted to investigate a bit but didn't have time, so sorry if I make wrong
> assumptions:
> 
> Does this bug result in strcoll(a, b) returning different result than
> [pseudocode] strcmp(strxfrm(a), strxfrm(b))?
> 
> I would have guessed the former is just a thin wrapper performing the
> latter; apparently it has its standalone implementation for performance
> reasons which is error prone to let the two go out of sync, am I right?
> 
> Wouldn't it make sense to extend the unittests to sort the files with both
> comparison approaches? (Although that'd still not give full protection
> against weird undiscovered corner cases where the two methods produce
> different results.)

My current analysis is that the fast-forward is wrong because it should not
fast forward into the middle of a digraph since that results in incorrect
sorting. I'm not sure why Leonhard didn't notice this.

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

[Bug locale/18589] sort-test.sh fails at random

[carlos at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

Carlos O'Donell <carlos at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |2.23

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

[Bug locale/18589] sort-test.sh fails at random

[egmont at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #6 from Egmont Koblinger <egmont at gmail dot com> ---
> I'm not sure why Leonhard didn't notice this.

We're humans, we make mistakes, that's fine. Maybe he didn't run the tests.
Maybe he ran them and they happened to pass. I really wouldn't blame him.

What worries me more: how come that some continuous build/test system didn't
catch this as a blocker bug, how come that it took such a long time to revert
the change, and how come that a stable version was released with this bug.
There's definitely room for improvement in these processes :) Nevermind!

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

[Bug locale/18589] sort-test.sh fails at random

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #7 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to Egmont Koblinger from comment #6)
> What worries me more: how come that some continuous build/test system didn't
> catch this as a blocker bug, how come that it took such a long time to
> revert the change, and how come that a stable version was released with this
> bug. There's definitely room for improvement in these processes :) Nevermind!

I opened the bug well before 2.22 was released.

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

[Bug locale/18589] sort-test.sh fails at random

[joseph at codesourcery dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #8 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
Random failures are hard to investigate.  In addition, while we have a 
buildbot, (a) it has no documentation for how people can add buildslaves 
for new glibc build configurations, (b) it doesn't alert committers to 
regressions and (c) it has "make check" mysteriously fail at random 
without an evident failing command, which requires someone to try to 
reproduce that manually and investigate it (see 
<https://sourceware.org/ml/libc-alpha/2015-09/msg00215.html> and 
<https://sourceware.org/ml/libc-alpha/2015-09/msg00309.html>).  Help with 
all those things would be welcome.

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

[Bug locale/18589] sort-test.sh fails at random

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #9 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, master has been updated
       via  87701a58e291bd7ac3b407d10a829dac52c9c16e (commit)
      from  fd91891a5091b5f54680180dc0c8e91827f63c70 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=87701a58e291bd7ac3b407d10a829dac52c9c16e

commit 87701a58e291bd7ac3b407d10a829dac52c9c16e
Author: Carlos O'Donell <carlos@systemhalted.org>
Date:   Thu Oct 8 16:34:53 2015 -0400

    strcoll: Remove incorrect STRDIFF-based optimization (Bug 18589).

    The optimization introduced in commit
    f13c2a8dff2329c6692a80176262ceaaf8a6f74e, causes regressions in
    sorting for languages that have digraphs that change sort order, like
    cs_CZ which sorts ch between h and i.

    My analysis shows the fast-forwarding optimization in STRCOLL advances
    through a digraph while possibly stopping in the middle which results
    in a subsequent skipping of the digraph and incorrect sorting. The
    optimization is incorrect as implemented and because of that I'm
    removing it for 2.23, and I will also commit this fix for 2.22 where
    it was originally introduced.

    This patch reverts the optimization, introduces a new bug-strcoll2.c
    regression test that tests both cs_CZ.UTF-8 and da_DK.ISO-8859-1 and
    ensures they sort one digraph each correctly. The optimization can't be
    applied without regressing this test.

    Checked on x86_64, bug-strcoll2.c fails without this patch and passes
    after. This will also get a fix on 2.22 which has the same bug.

-----------------------------------------------------------------------

Summary of changes:
 ChangeLog                    |   14 ++++++
 NEWS                         |   14 +++---
 locale/C-collate.c           |    4 +-
 locale/categories.def        |    1 -
 locale/langinfo.h            |    1 -
 locale/localeinfo.h          |    8 ----
 locale/programs/ld-collate.c |    9 ----
 string/bug-strcoll2.c        |   93 ++++++++++++++++++++++++++++++++++++++++++
 string/strcoll_l.c           |   38 +-----------------
 wcsmbs/wcscoll_l.c           |    1 -
 10 files changed, 116 insertions(+), 67 deletions(-)
 create mode 100644 string/bug-strcoll2.c

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

[Bug locale/18589] sort-test.sh fails at random

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #10 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, master has been updated
       via  02018629a1397d03eccceacaf2ee1c50e3c4001c (commit)
      from  87701a58e291bd7ac3b407d10a829dac52c9c16e (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=02018629a1397d03eccceacaf2ee1c50e3c4001c

commit 02018629a1397d03eccceacaf2ee1c50e3c4001c
Author: Carlos O'Donell <carlos@systemhalted.org>
Date:   Thu Oct 8 16:54:30 2015 -0400

    strcoll: Add bug-strcoll2 to testsuite (Bug 18589).

    Adds bug-strcoll2 to the string tests, along with the
    generation of required locales.

-----------------------------------------------------------------------

Summary of changes:
 ChangeLog             |    6 +++++-
 string/Makefile       |    5 +++--
 string/bug-strcoll2.c |    2 +-
 3 files changed, 9 insertions(+), 4 deletions(-)

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

[Bug locale/18589] sort-test.sh fails at random

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #11 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, master has been updated
       via  facdd9ea29ab94aac2b188ec3cc41f8733d769e0 (commit)
      from  67ac0f74a4a5021f77566fdbe4cda5880eba8ba4 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=facdd9ea29ab94aac2b188ec3cc41f8733d769e0

commit facdd9ea29ab94aac2b188ec3cc41f8733d769e0
Author: Carlos O'Donell <carlos@systemhalted.org>
Date:   Fri Oct 9 16:35:53 2015 -0400

    Fix typo in bug-strcoll2 (Bug 18589)

    Fix the copyright year and remove contributed by in the
    bug-strcoll2 test. In addition add the correct dependency
    on $(gen-locales) to ensure all the test locales are generated.

-----------------------------------------------------------------------

Summary of changes:
 ChangeLog             |    8 ++++++++
 string/Makefile       |    2 ++
 string/bug-strcoll2.c |    3 +--
 3 files changed, 11 insertions(+), 2 deletions(-)

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

[Bug locale/18589] sort-test.sh fails at random

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

--- Comment #12 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, release/2.22/master has been updated
       via  2b529a82afc15a76185fca16ba7bcc8fbd82d15d (commit)
       via  c0769d60c1d107d9fb622e80c9d1110deb4fb3b4 (commit)
       via  ed7f3a2e27e816488d5a8ef570052a5a2c3c515b (commit)
       via  6c84109cfa26f35c3dfed3acb97d347361bd5849 (commit)
      from  666bb4dc55f5a93538d020024c34abad641f50ae (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=2b529a82afc15a76185fca16ba7bcc8fbd82d15d

commit 2b529a82afc15a76185fca16ba7bcc8fbd82d15d
Author: Carlos O'Donell <carlos@systemhalted.org>
Date:   Fri Oct 9 16:35:53 2015 -0400

    Fix typo in bug-strcoll2 (Bug 18589)

    Fix the copyright year and remove contributed by in the
    bug-strcoll2 test. In addition add the correct dependency
    on $(gen-locales) to ensure all the test locales are generated.

    (cherry picked from commit facdd9ea29ab94aac2b188ec3cc41f8733d769e0)

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=c0769d60c1d107d9fb622e80c9d1110deb4fb3b4

commit c0769d60c1d107d9fb622e80c9d1110deb4fb3b4
Author: Carlos O'Donell <carlos@systemhalted.org>
Date:   Thu Oct 8 16:54:30 2015 -0400

    strcoll: Add bug-strcoll2 to testsuite (Bug 18589).

    Adds bug-strcoll2 to the string tests, along with the
    generation of required locales.

    (cherry picked from commit 02018629a1397d03eccceacaf2ee1c50e3c4001c)

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=ed7f3a2e27e816488d5a8ef570052a5a2c3c515b

commit ed7f3a2e27e816488d5a8ef570052a5a2c3c515b
Author: Martin Sebor <msebor@gcc.gnu.org>
Date:   Mon Sep 28 16:55:57 2015 -0400

    Let 'make check subdirs=string' succeed even when it's invoked
    immediately after glibc has been built and before 'make check'
    (or after 'make clean').

    (cherry picked from commit 60cf80f09d029257caedc0c8abe7e3e09c64e6c7)

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=6c84109cfa26f35c3dfed3acb97d347361bd5849

commit 6c84109cfa26f35c3dfed3acb97d347361bd5849
Author: Carlos O'Donell <carlos@systemhalted.org>
Date:   Thu Oct 8 16:34:53 2015 -0400

    strcoll: Remove incorrect STRDIFF-based optimization (Bug 18589).

    The optimization introduced in commit
    f13c2a8dff2329c6692a80176262ceaaf8a6f74e, causes regressions in
    sorting for languages that have digraphs that change sort order, like
    cs_CZ which sorts ch between h and i.

    My analysis shows the fast-forwarding optimization in STRCOLL advances
    through a digraph while possibly stopping in the middle which results
    in a subsequent skipping of the digraph and incorrect sorting. The
    optimization is incorrect as implemented and because of that I'm
    removing it for 2.23, and I will also commit this fix for 2.22 where
    it was originally introduced.

    This patch reverts the optimization, introduces a new bug-strcoll2.c
    regression test that tests both cs_CZ.UTF-8 and da_DK.ISO-8859-1 and
    ensures they sort one digraph each correctly. The optimization can't be
    applied without regressing this test.

    Checked on x86_64, bug-strcoll2.c fails without this patch and passes
    after. This will also get a fix on 2.22 which has the same bug.

    (cherry picked from commit 87701a58e291bd7ac3b407d10a829dac52c9c16e)

-----------------------------------------------------------------------

Summary of changes:
 ChangeLog                    |   37 +++++++++++++++++
 NEWS                         |    2 +-
 locale/C-collate.c           |    4 +-
 locale/categories.def        |    1 -
 locale/langinfo.h            |    1 -
 locale/localeinfo.h          |    8 ----
 locale/programs/ld-collate.c |    9 ----
 string/Makefile              |   15 ++++++-
 string/bug-strcoll2.c        |   92 ++++++++++++++++++++++++++++++++++++++++++
 string/strcoll_l.c           |   38 +-----------------
 string/tst-strxfrm2.c        |    6 ++-
 wcsmbs/wcscoll_l.c           |    1 -
 12 files changed, 150 insertions(+), 64 deletions(-)
 create mode 100644 string/bug-strcoll2.c

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

[Bug locale/18589] sort-test.sh fails at random

[carlos at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=18589

Carlos O'Donell <carlos at redhat dot com> changed:

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

--- Comment #13 from Carlos O'Donell <carlos at redhat dot com> ---
Fixed in 2.23, and 2.22. That fixes this issue on all released branches and
active development branches.

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