Custom Query (431 matches)

Filters
 
Or
 
  
 
Columns

Show under each result:


Results (100 - 102 of 431)

Ticket Resolution Summary Owner Reporter
#379 fixed IZ2102: Interpret $TASK_ID also for non-array jobs in -o / -e Dave Love <d.love@…> reuti
Description

[Imported from gridengine issuezilla http://gridengine.sunsource.net/issues/show_bug.cgi?id=2102]

        Issue #:      2102             Platform:     All           Reporter: reuti (reuti)
       Component:     gridengine          OS:        All
     Subcomponent:    kernel           Version:      6.0u8            CC:    None defined
        Status:       NEW              Priority:     P3
      Resolution:                     Issue type:    ENHANCEMENT
                                   Target milestone: ---
      Assigned to:    andreas (andreas)
      QA Contact:     andreas
          URL:
       * Summary:     Interpret $TASK_ID also for non-array jobs in -o / -e
   Status whiteboard:
      Attachments:

     Issue 2102 blocks:
   Votes for issue 2102:


   Opened: Tue Sep 26 02:45:00 -0700 2006 
------------------------


Some automatic processing would be easier to implement, if also non-array jobs would have a task id
set in their filenames (-o / -e), which could be 0. The assembly of the filenames seems to happen in
"execd/get_path.c":

          if (ja_task_id) {
             if (!strncmp(t, "$TASK_ID", sizeof("$TASK_ID") - 1)) {
                sprintf(exp_path, "%s" sge_u32, exp_path, ja_task_id);
                s = t + sizeof("$TASK_ID") - 1;
             }
          }

by removing the ja_task_id test and use just the 0. So the RFE would be to interpret $TASK_ID always in
-o / -e. Maybe keeping the default name without an additional .0 at the end, would allow users to
decide whether non-array jobs should have such an extension or not. If they want it, the simply have to
request: -o '$JOB_NAME.o$JOB_ID.$TASK_ID' (well, the po/pe output will be joint with o/e this way,
which you can't prevent).
#381 wontfix IZ2112: compilation pains and solutions on fedora core 6 dlove hintak
Description

[Imported from gridengine issuezilla http://gridengine.sunsource.net/issues/show_bug.cgi?id=2112]

   Issue #: 2112   Platform: Other   Reporter: hintak (hintak)
   Component: gridengine   OS: Linux
   Subcomponent: install   Version: 6.0u9   CC:
   [_] opoplawski
   [_] Remove selected CCs
   Status: NEW   Priority: P3
   Resolution:   Issue type: DEFECT
     Target milestone: ---
   Assigned to: dom (dom)
   QA Contact: dom
   URL:
   * Summary: compilation pains and solutions on fedora core 6
   Status whiteboard:
   Attachments:
   Date/filename:                                      Description:                                                                                                                  Submitted by:
   Thu Nov 9 13:27:00 -0700 2006: config.sub           config.sub from automake 1.9 ; please replace the config.sub in source/3rdparty/qtcsh and source/3rdparty/qmake. (text/plain) hintak
   Tue Nov 21 09:07:00 -0700 2006: qmon-noWerror.patch patch to remove all -Werror (text/plain)                                                                                      hintak
   Tue May 27 13:52:00 -0700 2008: warnings            Current warnings compiling on Fedora 10 (text/plain)                                                                          opoplawski
     Issue 2112 blocks:
   Votes for issue 2112:

   Opened: Thu Nov 9 12:23:00 -0700 2006 
------------------------


I have encountered some problems building 60u9 (and 60u8, 60u4) on fedora
core 6, and managed to fixed most of them, so I am going to detail it here.

(1) why the *&^%$£ is the default compiler switch in aimk contains both
-Werror and -Wno-error? Earlier gcc seems to think if you specify both, it
means no (or the last one take effect) - gcc 4.1 on fc6 seems to believe it is
the opposite - if you specify both, it means aborts on any warnings. I have to
strip out the -Werror's from aimk.

(2) "config.sub" in  source/3rdparty/qtcsh and source/3rdparty/qmake
are too old to recognise "x86_64-unknown-linux-gnu" on kernel 2.6 as a
valid platform.
You need to copy this file from a more recent automake/autoconf/libtool.
On my box many of these are the same and in any case, more up to date
that V60u9 is carrying:

/usr/lib/rpm/config.sub
/usr/lib/rpm/redhat/config.sub
/usr/lib64/php/build/config.sub
/usr/share/apps/kapptemplate/admin/config.sub
/usr/share/automake-1.4/config.sub
/usr/share/automake-1.5/config.sub
/usr/share/automake-1.6/config.sub
/usr/share/automake-1.7/config.sub
/usr/share/automake-1.9/config.sub
/usr/share/libtool/config.sub
/usr/share/libtool/libltdl/config.sub

(3) There wasn't any instructions about using the platform version
of ssl and crypto (which fedora ships). I had to do something quite ugly like this:

perl -pi -e \
's(^set SECLIBS_STATIC.+)(set SECLIBS_STATIC = " -lssl -lcrypto ")' source/aimk

I don't have a static version of those libraries around. (most linux users don't!).

(4) same thing with autodetecting the location of the openssl and includes.
I have to do:
perl -pi -e 's(/vol.+?\$buildarch)(/usr);' source/aimk.site

(3), (4) can be fixed and routinely done automatically by many
open-source software with autoconf, etc. ok, at least they can be detailed
in README.BUILD for manual editing. (1) and (2) is just plain-broken.

BTW, the dependency checking also does not seem to work reliably. often
e.g. aimk -only-qmon does not works after aimk depend (dies from not being able
to find some files), sometimes it requires a manual "aimk -only-core" or
some others in the middle. The problem seems intermittent.

   ------- Additional comments from hintak Thu Nov 9 13:27:58 -0700 2006 -------
Created an attachment (id=85)
config.sub from automake 1.9 ; please replace the config.sub in source/3rdparty/qtcsh and source/3rdparty/qmake.

   ------- Additional comments from hintak Fri Nov 17 11:34:16 -0700 2006 -------
The first issue apparently is well-known to the oscar project people -
this is in the SRPMS of sge-V60u8 they ship:
(http://svn.oscar.openclustergroup.org/oscar/branches/branch-5-0/packages/sge/SRPMS)

# The following patch (aimk.patch) removes all "-Werror"
# from the aimk script. Otherwise, the SGE build fails
# after warnings (not errors) when built against
# recent versions of glibc - Babu Sundaram, 04/13/06
Patch4: aimk.patch

   ------- Additional comments from hintak Tue Nov 21 09:07:29 -0700 2006 -------
Created an attachment (id=90)
patch to remove all -Werror

   ------- Additional comments from opoplawski Tue May 27 13:51:06 -0700 2008 -------
I have to strip -Werror as well from my Fedora rpm builds.  I'll attach a list
of all of the warnings I currently get during compilation.

   ------- Additional comments from opoplawski Tue May 27 13:52:31 -0700 2008 -------
Created an attachment (id=174)
Current warnings compiling on Fedora 10

   ------- Additional comments from opoplawski Tue May 27 13:53:57 -0700 2008 -------
Add myself to CC
#383 fixed IZ2114: build failure (and solution) on fedora core 3 to 6 (any platform where /usr/bin/java exists) hintak
Description

[Imported from gridengine issuezilla http://gridengine.sunsource.net/issues/show_bug.cgi?id=2114]

        Issue #:      2114             Platform:     All     Reporter: hintak (hintak)
       Component:     gridengine          OS:        Linux
     Subcomponent:    install          Version:      6.0u9      CC:    None defined
        Status:       NEW              Priority:     P3
      Resolution:                     Issue type:    PATCH
                                   Target milestone: ---
      Assigned to:    dom (dom)
      QA Contact:     dom
          URL:
       * Summary:     build failure (and solution) on fedora core 3 to 6 (any platform where /usr/bin/java exists)
   Status whiteboard:
      Attachments:
                      Date/filename:                                   Description:                                    Submitted by:
                      Tue Nov 21 09:04:00 -0700 2006: qmon-nojni.patch patch to exclude /usr as JAVA_HOME (text/plain) hintak

     Issue 2114 blocks:
   Votes for issue 2114:


   Opened: Fri Nov 17 06:43:00 -0700 2006 
------------------------


aimk calls aimk.site, which tries to work out JAVA_HOME from `which java`.
Unfortunately, from fedora core 3 onwards (and mandriva and
enterprise linux 4), and particularly on 64-bit platforms, there are gcj-based
package called java-gcj-compat java-gcj-compat-devel, which provides
a java implementation in /usr/bin/java .

This causes JAVA_HOME to be set to "/usr" from the logic in aimk.site,
which in turn, results in "-I${JAVA_HOME}/${JAVA_INCL}/${JAVA_ARCH}"
inside aimk to resolves to "-I/usr/include/linux". and /usr/include/linux
contains too many includes which are of the same names but different content
to /usr/include, thus resulting in catatrophic failure with
compilation, even on simple ones like "./aimk -only-core" or
"./aimk -only-qmon".

There are a few workarounds, like setting "JAVA_HOME=/nowhere" explicitly,
or explicitly append "-no-jni" in any of aimk's invocation.

suggest fix: test for JAVA_HOME != /usr .

This is the same issue as issue 2035, just with a lot more details:
http://gridengine.sunsource.net/issues/show_bug.cgi?id=2035

feel free to resolve as duplicate of issue 2035

gcj currently seems to be the only JNI implementation on 64-bit linux,
it is a shame that the grid engine gets really confused by it.

(unlike other bugzilla-based site, this site doesn't seem to allow
3rd party to offer advice/workaround/comments on existing bugs. The only option
seems to be attaching a patch - well, it is just going to be a one-liner in
aimk.site to make sure JAVA_HOME is not /usr ).

   ------- Additional comments from hintak Tue Nov 21 09:04:16 -0700 2006 -------
Created an attachment (id=89)
patch to exclude /usr as JAVA_HOME

   ------- Additional comments from hintak Tue Nov 21 09:05:50 -0700 2006 -------
change from defect to patch
Note: See TracQuery for help on using queries.