Re: 3.2.4 release candidate 1

[Jaroslav Hajek]

On Wed, Jan 20, 2010 at 11:25 AM, David Grundberg <address@hidden> wrote:
> Jaroslav Hajek wrote:
>>
>> On Wed, Jan 20, 2010 at 11:00 AM, David Grundberg <address@hidden>
>> wrote:
>>
>>>
>>> Jaroslav Hajek wrote:
>>>
>>>>
>>>> On Wed, Jan 20, 2010 at 10:56 AM, David Grundberg <address@hidden>
>>>> wrote:
>>>>
>>>>
>>>>>
>>>>> Jaroslav Hajek wrote:
>>>>>
>>>>>
>>>>>>
>>>>>> On Tue, Jan 19, 2010 at 1:37 PM, Marco Atzeri <address@hidden>
>>>>>> wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>>>
>>>>>>> --- Mar 19/1/10, Jaroslav Hajek <address@hidden> ha scritto:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>>
>>>>>>>> On Tue, Jan 19, 2010 at 7:48 AM, John
>>>>>>>> W. Eaton <address@hidden>
>>>>>>>> wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 19-Jan-2010, Jaroslav Hajek wrote:
>>>>>>>>>
>>>>>>>>> | I don't understand why you should need pthreads. I
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> guess one of the
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> | libraries you use is a static and multithreaded, but
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> I really can't
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> | tell. Adding -lpthread to LIBS should probably help
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> you.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> Look at the octave_mutex class defined in
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> liboctave/oct-mutex.cc and
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> used in the command_editor::event_handler function in
>>>>>>>>> liboctave/cmd-edit.cc.
>>>>>>>>>
>>>>>>>>> This problem was solved in the current development
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> sources by these
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> changes:
>>>>>>>>>
>>>>>>>>> changeset:   9488:ee572cdd4b97
>>>>>>>>> user:        John W. Eaton <address@hidden>
>>>>>>>>> date:        Wed Aug 05 13:27:58 2009 -0400
>>>>>>>>> summary:     add configure checks for pthread
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> library and compiler flags
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> changeset:   9497:54ae52abd4c9
>>>>>>>>> user:        Jaroslav Hajek <address@hidden>
>>>>>>>>> date:        Thu Aug 06 10:57:11 2009 +0200
>>>>>>>>> summary:     add missing file
>>>>>>>>>
>>>>>>>>> These changes to the connfigure script were
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> pre-automake, so they
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> might apply, but I have not tried to test them.  They
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> also could
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> depend on other previous changes to the configure
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> scripts and
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> Makefiles, so I don't know whether it is worth trying
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> to add them.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> Maybe it is better to just tell people who have
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> trouble compiling to
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>> add -lpthread to their LDFLGS.
>>>>>>>>>
>>>>>>>>> jwe
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> I'd like to avoid complex changes in build process in the
>>>>>>>> stable
>>>>>>>> series in general, but these two apply cleanly and are
>>>>>>>> based on code
>>>>>>>> from cryp.to, which should be somewhat proven, so why not.
>>>>>>>> I uploaded
>>>>>>>> RC2:
>>>>>>>> http://artax.karlin.mff.cuni.cz/~hajej2am/ulozna/octave/
>>>>>>>>
>>>>>>>> I don't think it's necessary to rebuild for those who
>>>>>>>> succeeded with RC1.
>>>>>>>> regards
>>>>>>>>
>>>>>>>> --
>>>>>>>> RNDr. Jaroslav Hajek, PhD
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>> I have not yet tested rc1 but on rc2
>>>>>>>
>>>>>>> make -C misc all
>>>>>>> make[3]: Entering directory
>>>>>>> `/pub/cygports/octave/octave-3.2.4-1/build/libcruft/
>>>>>>> misc'
>>>>>>> making quit.d from
>>>>>>> /pub/cygports/octave/octave-3.2.4-1/src/octave-3.2.4/libcruft
>>>>>>>       /misc/quit.cc
>>>>>>> making f77-extern.d from
>>>>>>> /pub/cygports/octave/octave-3.2..4-1/src/octave-3.2.4/li
>>>>>>>  bcruft/misc/f77-extern.cc
>>>>>>> making cquit.d from
>>>>>>> /pub/cygports/octave/octave-3.2.4-1/src/octave-3.2.4/libcruf
>>>>>>>     t/misc/cquit.c
>>>>>>> g++-4: @PTHREAD_CFLAGS@: No such file or directory
>>>>>>> g++-4: @PTHREAD_CFLAGS@: No such file or directory
>>>>>>> gcc-4: @PTHREAD_CFLAGS@: No such file or directory
>>>>>>> making lo-error.d from
>>>>>>> /pub/cygports/octave/octave-3.2.4-1/src/octave-3.2.4/libc
>>>>>>>   ruft/misc/lo-error.c
>>>>>>> ....
>>>>>>>
>>>>>>> gcc-4 -c  -I.
>>>>>>> -I/pub/cygports/octave/octave-3.2.4-1/src/octave-3.2.4/libcruft/mi
>>>>>>>           sc -I.../.. -I../../liboctave -I../../src
>>>>>>> -I../../libcruft/misc
>>>>>>> -I/pub/cygports/o
>>>>>>> ctave/octave-3.2.4-1/src/octave-3.2.4
>>>>>>> -I/pub/cygports/octave/octave-3.2.4-1/src/
>>>>>>> octave-3.2.4/liboctave
>>>>>>> -I/pub/cygports/octave/octave-3.2.4-1/src/octave-3.2.4/sr
>>>>>>>   c
>>>>>>> -I/pub/cygports/octave/octave-3.2.4-1/src/octave-3.2.4/libcruft/misc
>>>>>>>  -DHAVE_C                       ONFIG_H -mieee-fp -Wall -W -Wshadow
>>>>>>> -Wformat
>>>>>>> -O2 -pipe address@hidden@ /pub/cy
>>>>>>> gports/octave/octave-3.2.4-1/src/octave-3.2.4/libcruft/misc/cquit.c
>>>>>>> -o
>>>>>>> cquit.o
>>>>>>> gcc-4: @PTHREAD_CFLAGS@: No such file or directory
>>>>>>> gcc-4: @PTHREAD_CFLAGS@: No such file or directory
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> /pub/cygports/octave/octave-3.2.4-1/src/octave-3.2.4/libcruft/misc/f77-fcn.c:
>>>>>>> In                        function ‘xstopx_’:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> /pub/cygports/octave/octave-3.2.4-1/src/octave-3.2.4/libcruft/misc/f77-fcn.c:64:
>>>>>>>                      warning: function declared ‘noreturn’ has a
>>>>>>> ‘return’
>>>>>>> statement
>>>>>>> make[3]: *** [f77-fcn.o] Error 1
>>>>>>> make[3]: *** Waiting for unfinished jobs....
>>>>>>> make[3]: *** [cquit.o] Error 1
>>>>>>>
>>>>>>>
>>>>>>> Marco
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> This is probably because the acx_pthread source was not included in
>>>>>> distribution. I fixed it and uploaded rc3, so please try that.
>>>>>>
>>>>>> regards
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>> Using rc3, I'm still having problems with pthreads. The problem is that
>>>>> the
>>>>> .so-linking is done without -pthread, since PTHREAD_LIBS is empty. If I
>>>>> manually add -pthread to Makeconf's PTHREAD_LIBS, it seems to build
>>>>> correctly, and make check works (3/5429, the 2 unexpected ones are
>>>>> because
>>>>> of missing qhull).
>>>>>
>>>>> David
>>>>>
>>>>>
>>>>>
>>>>
>>>> Can you show the config.log...?
>>>>
>>>>
>>>>
>>>
>>> See attached file.
>>>
>>>
>>
>> Hmm. Does your configure script have the correct date (2010-01-19
>> 13:47)? Does it contain the ACX_PTHREAD section (grep ACX_PTHREAD
>> configure)?
>>
>
> Sorry. I attached the previous config.log.bz2 by mistake. This is the right
> one.
>
> address@hidden:/scratch/octave-3.2.4-rc3$ ls -al configure
> -rwxr-xr-x 1 davidg tdb 594947 2010-01-19 13:47 configure
> address@hidden:/scratch/octave-3.2.4-rc3$ grep ACX_PTHREAD configure
> #   ACX_PTHREAD([ACTION-IF-FOUND[, ACTION-IF-NOT-FOUND]])
> address@hidden:/scratch/octave-3.2.4-rc3$
>
> David
>

I see. The weird thing is that I actually have the same configuration,
but don't see the problem.
What happens with this patch (+reconfiguring)?

diff --git a/liboctave/Makefile.in b/liboctave/Makefile.in
--- a/liboctave/Makefile.in
+++ b/liboctave/Makefile.in
@@ -39,7 +39,7 @@
   $(LIBCRUFT) $(CHOLMOD_LIBS) $(UMFPACK_LIBS) $(AMD_LIBS) $(CAMD_LIBS) \
   $(COLAMD_LIBS) $(CCOLAMD_LIBS) $(CXSPARSE_LIBS) $(ARPACK_LIBS) \
   $(QRUPDATE_LIBS) $(BLAS_LIBS) $(FFTW_LIBS) $(LIBREADLINE) $(LIBGLOB) \
-  $(REGEX_LIBS) $(LIBS) $(FLIBS) $(PTHREAD_LIBS)
+  $(REGEX_LIBS) $(LIBS) $(FLIBS) $(PTHREAD_CFLAGS) $(PTHREAD_LIBS)

 MATRIX_INC := Array.h Array2.h Array3.h ArrayN.h DiagArray2.h \
        Array-util.h MArray-decl.h MArray-defs.h \


-- 
RNDr. Jaroslav Hajek, PhD
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz