[Bug localedata/27320] New: localedata/gen-locale.sh failure is ignored

[Bug localedata/27320] New: localedata/gen-locale.sh failure is ignored

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

            Bug ID: 27320
           Summary: localedata/gen-locale.sh failure is ignored
           Product: glibc
           Version: 2.34
            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: ---

with

diff --git a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
index 50bb1fccb2..ee8651ade7 100644
--- a/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
+++ b/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
@@ -256,6 +256,7 @@ L(movsb):
 # if AVOID_SHORT_DISTANCE_REP_MOVSB
   movq  %rdi, %rcx
   subq  %rsi, %rcx
+  hlt
   jmp   2f
 # endif
 1:

$ make subdirs=string -j8 check

gave me:

../localedata/gen-locale.sh: line 29: 4190351 Segmentation fault      (core
dumped) ${localedef_before_env} ${run_program_env} I18NPATH=../localedata
${localedef_after_env} $flags -f $charmap -i $input
${common_objpfx}localedata/$out
Charmap: "UTF-8" Inputfile: "de_DE" Outputdir: "de_DE.UTF-8" failed
...
Summary of test results:
     78 PASS

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

[Bug localedata/27320] localedata/gen-locale.sh failure is ignored

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |security-
                 CC|                            |fweimer at redhat dot com

--- Comment #1 from Florian Weimer <fweimer at redhat dot com> ---
Are the locale-generation failures reported somewhere?

The ignoring-the-error part is intentional, per:

commit 6f99f280b00a30b8f0a89a4be1adb2bea41e2954
Author: Joseph Myers <joseph@codesourcery.com>
Date:   Fri Mar 7 03:31:41 2014 +0000

    Count miscellaneous files built on host for testing as tests.

    In <https://sourceware.org/ml/libc-alpha/2014-01/msg00198.html> I
    raised the question of counting miscellaneous dependencies of tests,
    built on the host rather than the build system, as tests, so that when
    test failures don't stop "make check" neither do those other builds on
    the host, so that a flaky host doesn't stop "make check" from
    producing a complete summary of test results.  Brooks supported that
    idea in <https://sourceware.org/ml/libc-alpha/2014-02/msg00301.html>.

    This patch implements that change for all the examples I could find:
    one message catalog in catgets/, locales in localedata/ and timezone
    files in timezone/.

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

[Bug localedata/27320] localedata/gen-locale.sh failure is ignored

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to Florian Weimer from comment #1)
> Are the locale-generation failures reported somewhere?

I was checking coverage ofmemmove test.

> The ignoring-the-error part is intentional, per:
> 
> commit 6f99f280b00a30b8f0a89a4be1adb2bea41e2954
> Author: Joseph Myers <joseph@codesourcery.com>
> Date:   Fri Mar 7 03:31:41 2014 +0000
> 
>     Count miscellaneous files built on host for testing as tests.
>     
>     In <https://sourceware.org/ml/libc-alpha/2014-01/msg00198.html> I
>     raised the question of counting miscellaneous dependencies of tests,
>     built on the host rather than the build system, as tests, so that when
>     test failures don't stop "make check" neither do those other builds on
>     the host, so that a flaky host doesn't stop "make check" from
>     producing a complete summary of test results.  Brooks supported that
>     idea in <https://sourceware.org/ml/libc-alpha/2014-02/msg00301.html>.
>     
>     This patch implements that change for all the examples I could find:
>     one message catalog in catgets/, locales in localedata/ and timezone
>     files in timezone/.

In this particular case, localedata/gen-locale.sh failed because of a bug
in memmove.  I don't think these kinds of failures should be ignored.

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

[Bug localedata/27320] localedata/gen-locale.sh failure is ignored

[joseph at codesourcery dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

--- Comment #3 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
On Tue, 2 Feb 2021, hjl.tools at gmail dot com via Glibc-bugs wrote:

> In this particular case, localedata/gen-locale.sh failed because of a bug
> in memmove.  I don't think these kinds of failures should be ignored.

The basic principle is: suppose your board falls over (kernel panic etc.) 
half way through a test run.  It should still produce a toplevel tests.sum 
file, with all the tests that couldn't run because the board had fallen 
over having their results reported as FAIL.

The failure shouldn't be *ignored* - but it should result in a FAIL and 
other tests continuing to run, rather than an early exit from "make 
check".

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

[Bug localedata/27320] localedata/gen-locale.sh failure is ignored

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

--- Comment #4 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to joseph@codesourcery.com from comment #3)
> On Tue, 2 Feb 2021, hjl.tools at gmail dot com via Glibc-bugs wrote:
> 
> > In this particular case, localedata/gen-locale.sh failed because of a bug
> > in memmove.  I don't think these kinds of failures should be ignored.
> 
> The basic principle is: suppose your board falls over (kernel panic etc.) 
> half way through a test run.  It should still produce a toplevel tests.sum 
> file, with all the tests that couldn't run because the board had fallen 
> over having their results reported as FAIL.
> 
> The failure shouldn't be *ignored* - but it should result in a FAIL and 
> other tests continuing to run, rather than an early exit from "make 
> check".

When localedata/gen-locale.sh fails, can we trust the test results which
depend on the output of localedata/gen-locale.sh?

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

[Bug localedata/27320] localedata/gen-locale.sh failure is ignored

[schwab@linux-m68k.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

--- Comment #5 from Andreas Schwab <schwab@linux-m68k.org> ---
When a test fails to compile you also get an immediate stop.

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

[Bug localedata/27320] localedata/gen-locale.sh failure is ignored

[joseph at codesourcery dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

--- Comment #6 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
The basic principle is that anything running on the *build* system should 
be reliable (and it's OK for failure to result in the termination of "make 
check"), while anything running on the glibc *host* may not be reliable, 
may run into problems with the kernel, network between build and host 
systems, embedded hardware, etc. (and failure should be reported as a FAIL 
without terminating "make check").

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

[Bug localedata/27320] localedata/gen-locale.sh failure in native build is ignored

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|localedata/gen-locale.sh    |localedata/gen-locale.sh
                   |failure is ignored          |failure in native build is
                   |                            |ignored

--- Comment #7 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to joseph@codesourcery.com from comment #6)
> The basic principle is that anything running on the *build* system should 
> be reliable (and it's OK for failure to result in the termination of "make 
> check"), while anything running on the glibc *host* may not be reliable, 
> may run into problems with the kernel, network between build and host 
> systems, embedded hardware, etc. (and failure should be reported as a FAIL 
> without terminating "make check").

My case is the native build where build == host == target.
localedata/gen-locale.sh uses the newly built glibc for target
and the target failure is ignored.

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

[Bug localedata/27320] localedata/gen-locale.sh failure in native build is ignored

[joseph at codesourcery dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

--- Comment #8 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
An appropriate fix would be to make any failure in 
localedata/gen-locale.sh produce a FAIL for the generated locale.

The first step in that is that gen-locale.sh should exit with error status 
on error from localedef, which it looks to me like it does (an explicit 
"exit 1").  The second step is that the makefile rule should use 
$(evaluate-test), which it also does (in gen-locales.mk).  So there are a 
few other issues to resolve:

* The locales should be listed in tests-special (like how 
timezone/Makefile does

tests-special += $(addprefix $(testdata)/, $(test-zones))

to get files built by zic listed as tests, for example).

* That listing in tests-special probably needs to be in the localedata/ 
directory as that's where the files in question are built.

* Then the FAIL would be in localedata/subdir-tests.sum.

* But you're using subdirs=string which prevents FAILs outside 
string/subdir-tests.sum from being considered in the process of merging 
FAILs from subdirectories.  Fixing the points above should ensure a FAIL 
when gen-locale.sh fails in a full testsuite run, but getting one if 
you're only building a non-localedata subdirectory may be harder.

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

[Bug localedata/27320] localedata/gen-locale.sh failure in native build is ignored

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=27320

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |skpgkp2 at gmail dot com

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