Perl 5.8.5 - OS2

This is a discussion on Perl 5.8.5 - OS2 ; Has anyone attempted to build Perl 5.8.5 - the latest stable release? I've always had a small number of test failures whenever I've built it in the past. It would be nice to achieve a completely faultless build one day... ...

+ Reply to Thread
Results 1 to 11 of 11

Thread: Perl 5.8.5

  1. Perl 5.8.5

    Has anyone attempted to build Perl 5.8.5 - the latest stable release?

    I've always had a small number of test failures whenever I've built it
    in the past. It would be nice to achieve a completely faultless build
    one day...

    Here are my test results for 5.8.5:-

    Failed 7/895 test scripts, 99.22% okay. 51/85671 subtests failed, 99.94%
    okay.
    Failed Test Stat Wstat Total Fail Failed List of Failed
    -------------------------------------------------------------------------------
    .../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    .../lib/Shell.t 1 256 7 1 14.29% 3
    lib/os2_base.t 1 256 19 1 5.26% 8
    lib/os2_ea.t 21 8 38.10% 7-11 14-16
    op/magic.t 54 1 1.85% 6
    op/stat.t 78 3 3.85% 75-77
    op/write.t 87 32 36.78% 55-86
    68 tests and 641 subtests skipped.
    Relative UNIX domain socket name 'sock-45846' unsupported at
    .../lib/Socket.pm line 383.
    # Failed test (t/../lib/Shell.t at line 36)
    # Looks like you failed 1 tests of 7.
    # Failed test (t/lib/os2_base.t at line 44)
    # 'T:/USR/LIB/PERLF12F.DLL'
    # doesn't match
    '(?i-xsm:^T\:\/ux2bs\/workdir\/perl\-5\.8\.5(/t)?\/perlF12F\.dll)'
    # Looks like you failed 1 tests of 19.

    Process terminated by SIGINT

    Process terminated by SIGINT
    # Failed at t/op/stat.t line 445
    # Failed at t/op/stat.t line 446
    # Failed at t/op/stat.t line 447
    Can't open perl script "t/68": No such file or directory.
    Use -S to search $PATH for it.


    Any suggestions on why any of these tests fail?

    It was run on a TVFS filesystem if that makes any difference.

  2. Re: Perl 5.8.5

    [A complimentary Cc of this posting was sent to
    jp
    ], who wrote in article :
    > Here are my test results for 5.8.5:-


    > Failed 7/895 test scripts, 99.22% okay. 51/85671 subtests failed, 99.94% okay.
    > Failed Test Stat Wstat Total Fail Failed List of Failed
    > -------------------------------------------------------------------------------
    > ../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    > ../lib/Shell.t 1 256 7 1 14.29% 3
    > lib/os2_base.t 1 256 19 1 5.26% 8
    > lib/os2_ea.t 21 8 38.10% 7-11 14-16
    > op/magic.t 54 1 1.85% 6
    > op/stat.t 78 3 3.85% 75-77
    > op/write.t 87 32 36.78% 55-86
    > 68 tests and 641 subtests skipped.


    Yesterday I needed to build 5.8.6 (one of my hairiest modules would
    fail with 5.8.6, but not with earlier versions). Here are the
    results:

    Failed 6/922 test scripts, 99.35% okay. 39/87899 subtests failed, 99.96% okay.
    Failed Test Stat Wstat Total Fail Failed List of Failed
    -------------------------------------------------------------------------------
    .../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    .../lib/CPAN/t/signature.t 1 1 100.00% 1
    .../lib/DBM_Filter/t/compress.t 0 11 ?? ?? % ??
    lib/rx_vrexx.t 0 15 ?? ?? % ??
    op/magic.t 54 1 1.85% 6
    op/write.t 87 32 36.78% 55-86
    67 tests and 650 subtests skipped.

    I had no time to test whether my patches were included; to see the history, google for "group:*os2* perl 5.8 op/magic".

    > Relative UNIX domain socket name 'sock-45846' unsupported at
    > ../lib/Socket.pm line 383.
    > # Failed test (t/../lib/Shell.t at line 36)
    > # Looks like you failed 1 tests of 7.
    > # Failed test (t/lib/os2_base.t at line 44)
    > # 'T:/USR/LIB/PERLF12F.DLL'
    > # doesn't match
    > '(?i-xsm:^T\:\/ux2bs\/workdir\/perl\-5\.8\.5(/t)?\/perlF12F\.dll)'
    > # Looks like you failed 1 tests of 19.


    This is as expected. Absolute path like /foo/sock-45846 would be
    translated to \socket\foo\sock-45846. Relative paths are not
    supported... (should I just prepend the drive and dir?)

    Hope this helps,
    Ilya

  3. Re: Perl 5.8.5

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > jp
    > ], who wrote in article :
    >
    >>Here are my test results for 5.8.5:-

    >
    >
    >>Failed 7/895 test scripts, 99.22% okay. 51/85671 subtests failed, 99.94% okay.
    >>Failed Test Stat Wstat Total Fail Failed List of Failed
    >>-------------------------------------------------------------------------------
    >>../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    >>../lib/Shell.t 1 256 7 1 14.29% 3
    >>lib/os2_base.t 1 256 19 1 5.26% 8
    >>lib/os2_ea.t 21 8 38.10% 7-11 14-16
    >>op/magic.t 54 1 1.85% 6
    >>op/stat.t 78 3 3.85% 75-77
    >>op/write.t 87 32 36.78% 55-86
    >>68 tests and 641 subtests skipped.

    >
    >
    > Yesterday I needed to build 5.8.6 (one of my hairiest modules would
    > fail with 5.8.6, but not with earlier versions). Here are the
    > results:
    >
    > Failed 6/922 test scripts, 99.35% okay. 39/87899 subtests failed, 99.96% okay.
    > Failed Test Stat Wstat Total Fail Failed List of Failed
    > -------------------------------------------------------------------------------
    > ../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    > ../lib/CPAN/t/signature.t 1 1 100.00% 1
    > ../lib/DBM_Filter/t/compress.t 0 11 ?? ?? % ??
    > lib/rx_vrexx.t 0 15 ?? ?? % ??
    > op/magic.t 54 1 1.85% 6
    > op/write.t 87 32 36.78% 55-86
    > 67 tests and 650 subtests skipped.


    Just for comparison, this is what I got for 5.8.6:-

    Failed 8/922 test scripts, 99.13% okay. 49/87677 subtests failed, 99.94%
    okay.
    Failed Test Stat Wstat Total Fail Failed List of Failed
    -------------------------------------------------------------------------------
    .../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    .../ext/Time/HiRes/t/HiRes.t 25 1 4.00% 11
    lib/os2_base.t 1 256 19 1 5.26% 8
    lib/os2_ea.t 21 8 38.10% 7-11 14-16
    lib/os2_process_kid.t ?? ?? % ??
    op/magic.t 54 1 1.85% 6
    op/stat.t 78 1 1.28% 44
    op/write.t 87 32 36.78% 55-86
    72 tests and 641 subtests skipped.


    I have a suspicion that the os2_ea.t test may be related to JFS. How
    would I test that?

    I notice that the failures you have for io_unix.t, magic.t and write.t
    are identical to mine. Do you have any intention of correcting the errors?




    > I had no time to test whether my patches were included; to see the history, google for "group:*os2* perl 5.8 op/magic".
    >
    >
    >>Relative UNIX domain socket name 'sock-45846' unsupported at
    >>../lib/Socket.pm line 383.
    >># Failed test (t/../lib/Shell.t at line 36)
    >># Looks like you failed 1 tests of 7.
    >># Failed test (t/lib/os2_base.t at line 44)
    >># 'T:/USR/LIB/PERLF12F.DLL'
    >># doesn't match
    >>'(?i-xsm:^T\:\/ux2bs\/workdir\/perl\-5\.8\.5(/t)?\/perlF12F\.dll)'
    >># Looks like you failed 1 tests of 19.

    >
    >
    > This is as expected. Absolute path like /foo/sock-45846 would be
    > translated to \socket\foo\sock-45846. Relative paths are not
    > supported... (should I just prepend the drive and dir?)


    OS/2 sockets have a different naming convention to Unix sockets

    > Hope this helps,
    > Ilya


  4. Re: Perl 5.8.5

    [A complimentary Cc of this posting was sent to
    jp
    ], who wrote in article :
    > > Yesterday I needed to build 5.8.6 (one of my hairiest modules would
    > > fail with 5.8.6, but not with earlier versions). Here are the
    > > results:
    > >
    > > Failed 6/922 test scripts, 99.35% okay. 39/87899 subtests failed, 99.96% okay.
    > > Failed Test Stat Wstat Total Fail Failed List of Failed
    > > -------------------------------------------------------------------------------
    > > ../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    > > ../lib/CPAN/t/signature.t 1 1 100.00% 1
    > > ../lib/DBM_Filter/t/compress.t 0 11 ?? ?? % ??
    > > lib/rx_vrexx.t 0 15 ?? ?? % ??
    > > op/magic.t 54 1 1.85% 6
    > > op/write.t 87 32 36.78% 55-86
    > > 67 tests and 650 subtests skipped.


    Only op/write.t may be a genuine failure of the port:

    io_unix: relative name for unix domain socket; already explained

    ../lib/*: were picking up installed DLLs for 5.8.2 without
    binary-compatibility DLL in place;

    rx_vrexx: my reorganization of LIBPATH was not completed;

    op/magic: Test::Harness studipidity.

    > Just for comparison, this is what I got for 5.8.6:-
    >
    > Failed 8/922 test scripts, 99.13% okay. 49/87677 subtests failed, 99.94%
    > okay.
    > Failed Test Stat Wstat Total Fail Failed List of Failed
    > -------------------------------------------------------------------------------
    > ../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    > ../ext/Time/HiRes/t/HiRes.t 25 1 4.00% 11
    > lib/os2_base.t 1 256 19 1 5.26% 8
    > lib/os2_ea.t 21 8 38.10% 7-11 14-16
    > lib/os2_process_kid.t ?? ?? % ??
    > op/magic.t 54 1 1.85% 6
    > op/stat.t 78 1 1.28% 44
    > op/write.t 87 32 36.78% 55-86
    > 72 tests and 641 subtests skipped.


    > I have a suspicion that the os2_ea.t test may be related to JFS. How
    > would I test that?


    Copy build dir to HPFS, and rerun

    perl t/harness

    ?

    > I notice that the failures you have for io_unix.t, magic.t and write.t
    > are identical to mine. Do you have any intention of correcting the errors?


    Somebody needs to spend time and analyse what breaks in op/write.t.
    The tests which fail "write with output formats" to "|-" (i.e., to a
    fork()ed out copy of the process).

    > > This is as expected. Absolute path like /foo/sock-45846 would be
    > > translated to \socket\foo\sock-45846. Relative paths are not
    > > supported... (should I just prepend the drive and dir?)

    >
    > OS/2 sockets have a different naming convention to Unix sockets


    Who cares? The question is whether one should/can translate the API
    calls that existing scripts/modules work without user intervention.

    I repeat: if one opens a unix domain socket "/foo/bar", this is
    translated to OS/2 name '\socket\foo\bar'. The questing is whether
    doing a similar thing to a relative name (e.g., to 'bar' if in
    directory k:/foo) will lead to more breakage than benefits.

    [E.g., nowadays such scripts die() with a descriptive error message.
    If "fixed" as above, the scripts which just open sockets (and maybe
    also change directories inside the same drive) will start to work.
    However, scripts which change drives, and/or check for sockets with
    stat() calls, will break "mysteriously", without a clear error message.]

    Thanks,
    Ilya

  5. Re: Perl 5.8.5

    Ilya Zakharevich wrote:
    >>>
    >>>Failed 6/922 test scripts, 99.35% okay. 39/87899 subtests failed, 99.96% okay.
    >>>Failed Test Stat Wstat Total Fail Failed List of Failed
    >>>-------------------------------------------------------------------------------
    >>>../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    >>>../lib/CPAN/t/signature.t 1 1 100.00% 1
    >>>../lib/DBM_Filter/t/compress.t 0 11 ?? ?? % ??
    >>>lib/rx_vrexx.t 0 15 ?? ?? % ??
    >>>op/magic.t 54 1 1.85% 6
    >>>op/write.t 87 32 36.78% 55-86
    >>>67 tests and 650 subtests skipped.

    >
    >
    > Only op/write.t may be a genuine failure of the port:
    >
    > io_unix: relative name for unix domain socket; already explained
    >
    > ../lib/*: were picking up installed DLLs for 5.8.2 without
    > binary-compatibility DLL in place;
    >
    > rx_vrexx: my reorganization of LIBPATH was not completed;
    >
    > op/magic: Test::Harness studipidity.
    >
    >
    >>Just for comparison, this is what I got for 5.8.6:-
    >>
    >>Failed 8/922 test scripts, 99.13% okay. 49/87677 subtests failed, 99.94%
    >>okay.
    >>Failed Test Stat Wstat Total Fail Failed List of Failed
    >>-------------------------------------------------------------------------------
    >>../ext/IO/t/io_unix.t 29 7424 5 10 200.00% 1-5
    >>../ext/Time/HiRes/t/HiRes.t 25 1 4.00% 11
    >>lib/os2_base.t 1 256 19 1 5.26% 8
    >>lib/os2_ea.t 21 8 38.10% 7-11 14-16
    >>lib/os2_process_kid.t ?? ?? % ??
    >>op/magic.t 54 1 1.85% 6
    >>op/stat.t 78 1 1.28% 44
    >>op/write.t 87 32 36.78% 55-86
    >>72 tests and 641 subtests skipped.


    I just noticed that the number of subtests performed is different which
    suggests that I'm performing 222 fewer subtests. How can I find out
    which they are?


    >
    >>I have a suspicion that the os2_ea.t test may be related to JFS. How
    >>would I test that?

    >
    >
    > Copy build dir to HPFS, and rerun
    >
    > perl t/harness
    >
    > ?


    The os2_ea.t test does not handle the '+' character in EAs properly. I'm
    told it is an illegal character.

    >
    > Thanks,
    > Ilya


  6. Re: Perl 5.8.5

    [A complimentary Cc of this posting was sent to
    jp
    ], who wrote in article <9_idnXkjkNMtN1_cRVnyvg@eclipse.net.uk>:

    > I just noticed that the number of subtests performed is different which
    > suggests that I'm performing 222 fewer subtests. How can I find out
    > which they are?


    For the coarse understanding, I ran

    egrep -B1 "^ +[1-9]" 00h1 >00h1-skipped

    here 00h1 is the file where STDERR/STDOUT of t/harness was redirected
    to. The result is put at the end. Compare with what you got.

    > The os2_ea.t test does not handle the '+' character in EAs properly.


    Can you be more specific, please?

    > I'm told it is an illegal character.


    Do you have a section of cpref for this?

    Thanks,
    Ilya

    ================================================== =====
    base/rs.....................................ok
    4/16 skipped: on non-VMS system
    --
    io/fs.......................................ok
    15/34 skipped: various reasons
    --
    io/open.....................................ok
    1/108 skipped: This perl uses perlio
    --
    op/chdir....................................ok
    2/31 skipped: Already in t/
    --
    op/closure..................................ok
    3/187 skipped: tests not in 5.8.
    --
    op/exec.....................................ok
    2/21 skipped: bug/feature of pdksh
    --
    op/filetest.................................ok
    2/10 skipped: no seteuid
    --
    op/pack.....................................ok
    206/5852 skipped: various reasons
    op/pat......................................ok
    2/1065 skipped: only in EBCDIC
    --
    op/regexp...................................ok
    4/945 skipped:
    op/regexp_noamp.............................ok
    6/945 skipped: skipping $&
    --
    op/stat.....................................ok
    25/78 skipped: various reasons
    --
    op/sub_lval.................................ok
    3/67 skipped: removed test
    --
    op/taint....................................ok
    23/236 skipped: various reasons
    --
    op/tr.......................................ok
    4/99 skipped: not EBCDIC
    --
    op/ver......................................ok
    3/53 skipped: No quads
    --
    lib/os2_process.............................ok
    14/232 skipped: various reasons
    lib/os2_process_kid.........................ok
    2/232 skipped: if defaultVIO=MAXIMIZED, new windows are shifted, but maximize to UL corner
    --
    .../ext/B/t/optree_check.....................ok
    4/23 skipped: various reasons
    --
    .../ext/Cwd/t/cwd............................ok
    3/27 skipped: no symlinks on this platform
    --
    .../ext/Digest/MD5/t/files...................ok
    1/5 skipped: PERL_CORE
    --
    .../ext/File/Glob/t/basic....................ok
    1/12 skipped:
    --
    .../ext/SDBM_File/t/sdbm.....................ok
    1/80 skipped: different file permission semantics
    .../ext/Socket/t/socketpair..................ok
    24/45 skipped: No usable SOCK_DGRAM for socketpair
    --
    .../ext/Sys/Syslog/t/syslog..................ok
    6/6 skipped: various reasons
    --
    ok
    9/25 skipped: various reasons
    --
    .../lib/AnyDBM_File..........................ok
    1/12 skipped: different file permission semantics
    --
    .../lib/bignum/t/bn_lite.....................ok
    1/1 skipped: no Math::BigInt::Lite
    --
    .../lib/bignum/t/br_lite.....................ok
    1/1 skipped: no Math::BigInt::Lite
    --
    .../lib/CGI/t/fast...........................ok
    7/7 skipped: FCGI not installed, cannot continue
    --
    .../lib/CGI/t/push...........................ok
    1/12 skipped: do_sleep() test may take a while
    --
    .../lib/ExtUtils/t/00compile.................ok
    30/60 skipped: Test::Pod not installed
    --
    .../lib/ExtUtils/t/Command...................ok
    3/26 skipped: different file permission semantics on os2
    --
    .../lib/ExtUtils/t/MM_OS2....................ok
    4/32 skipped: various reasons
    --
    .../lib/File/Copy............................ok
    4/28 skipped: various reasons
    --
    .../lib/File/Find/t/taint....................ok
    18/45 skipped: various reasons
    --
    .../lib/File/Spec/t/crossplatform............ok
    7/50 skipped: Can't load File::Spec::VMS
    --
    .../lib/File/Spec/t/Spec.....................ok
    83/465 skipped: various reasons
    .../lib/File/stat............................ok
    2/19 skipped: various reasons
    --
    .../lib/File/Temp/t/security.................ok
    9/13 skipped: various reasons
    --
    .../lib/filetest.............................ok
    4/15 skipped: No /usr/bin/chflags available
    --
    .../lib/Net/t/netrc..........................ok
    4/20 skipped: incompatible stat() handling for OS
    --
    .../lib/Term/Complete........................ok
    7/8 skipped: No stty, Term::Complete will not run here
    --
    .../lib/Test/Harness/t/test-harness..........ok
    56/208 skipped: various reasons
    --
    .../lib/Test/Simple/t/skip...................ok
    7/15 skipped: Just testing the skip interface.
    --
    .../lib/Test/t/success.......................ok
    1/11 skipped: just testing skip()
    --
    .../lib/vmsish...............................ok
    24/25 skipped: tests for VMS only

  7. Re: Perl 5.8.5

    Ilya Zakharevich wrote:

    >[A complimentary Cc of this posting was sent to
    >jp
    >], who wrote in article <9_idnXkjkNMtN1_cRVnyvg@eclipse.net.uk>:
    >
    >
    >
    >>The os2_ea.t test does not handle the '+' character in EAs properly.
    >>
    >>

    >
    >Can you be more specific, please?
    >
    >
    >
    >>I'm told it is an illegal character.
    >>
    >>

    >
    >Do you have a section of cpref for this?
    >
    >
    >


    Control Programming Guide and Reference that comes with VA and the Dev
    Toolkit, section on EAs, page headed "EA Naming Conventions".



    >Thanks,
    >Ilya
    >
    >
    >


  8. Re: Perl 5.8.5

    [A complimentary Cc of this posting was sent to
    jp
    ], who wrote in article :
    > >>The os2_ea.t test does not handle the '+' character in EAs properly.


    > >Can you be more specific, please?


    Still no answer on this...

    > >>I'm told it is an illegal character.


    > Control Programming Guide and Reference that comes with VA and the
    > Dev Toolkit, section on EAs, page headed "EA Naming Conventions".


    It says nothing about "legality". Only about name separation between
    different applicatioins and system.

    Hope this helps,
    Ilya



  9. Re: Perl 5.8.5

    Ilya Zakharevich wrote:

    >[A complimentary Cc of this posting was sent to
    >jp
    >], who wrote in article :
    >
    >
    >>>>The os2_ea.t test does not handle the '+' character in EAs properly.
    >>>>
    >>>>

    >
    >
    >
    >>>>I'm told it is an illegal character.
    >>>>
    >>>>

    >
    >
    >
    >>Control Programming Guide and Reference that comes with VA and the
    >>
    >>
    > > Dev Toolkit, section on EAs, page headed "EA Naming Conventions".

    >
    >It says nothing about "legality". Only about name separation between
    >different applicatioins and system.
    >
    >
    >

    I don't have the guide but am told that '+' is one of four characters
    not to be used in naming EAs ("reserved for system use"). In effect the
    test should fail on HPFS, AIUI, but doesn't.


    >Hope this helps,
    >Ilya
    >
    >
    >
    >


  10. Re: Perl 5.8.5

    [A complimentary Cc of this posting was sent to
    jp
    ], who wrote in article :
    > >It says nothing about "legality". Only about name separation between
    > >different applicatioins and system.


    > I don't have the guide but am told that '+' is one of four characters
    > not to be used in naming EAs ("reserved for system use"). In effect the
    > test should fail on HPFS, AIUI, but doesn't.


    The words "reserved for system use" do appear in the documentation.
    Given that these words have no meaning, I do not see how you conclude
    that "the test should fail on HPFS".

    In regards to JFS, I assume that these words are just IBM-speak, and
    in reality "Creating or modification of EAs with '+' etc does not has
    defined behaviour".

    Thanks,
    Ilya

  11. Re: Perl 5.8.5

    On Tue, 21 Dec 2004 20:29:01 UTC, Ilya Zakharevich wrote:

    > The words "reserved for system use" do appear in the documentation.
    > Given that these words have no meaning


    If you do a web search e.g. like this
    http://www.google.com/search?q=%22re...+system+use%22
    you will see that this term is used very often and on many different OSs
    and systems and in most of the cases it means it is not available to
    users or don't use it/them.
    --
    Greetings, Please reply in newsgroup, I rarely
    Peter. read emails to pweilba@gwdg.de...

+ Reply to Thread