bugGNU Octave - Bugs: bug #53151, Clang OSX buildbot failing

 
 

bug #53151: Clang OSX buildbot failing

Submitter:  Rik <rik5>
Submitted:  Tue 13 Feb 2018 11:44:47 PM UTC
   
 
Category:  Configuration and Build System Severity:  3 - Normal
Priority:  5 - Normal Item Group:  Regression
Status:  Fixed Assigned to:  None
Originator Name:  Open/Closed:  * Closed
Release:  * dev Operating System:  * Mac OS
Fixed Release:  None Planned Release:  None
* Mandatory Fields

Add a New Comment Rich Markup
   

Jump to the original submission

Thu 08 Mar 2018 03:30:54 PM UTC, comment #42: 

I have created a bug report here: https://savannah.gnu.org/bugs/index.php?53305. I suggest to skip that test on macOS for now (@Rik or Carlo: can you do that?) and consider it as a known problem to be solved in #53305.

Seb.

Sebastian <sschoeps>
Thu 08 Mar 2018 01:38:57 PM UTC, comment #41: 

usejava() is designed to report whether a specific feature of Java exists.  In this case, the test is for 'awt' or the Abstract Widget Toolkit.

The code in usejava.m is


    ## For each feature, try methods() on a Java class of a feature
    case "awt"
      try
        dum = methods ("java.awt.Frame");
        retval = ! javaMethod ("isHeadless", "java.awt.GraphicsEnvironment");
      end_try_catch


Is the Java interpreter on macOS so "helpful" that it tries to install a missing package when you try and run a function from it, rather than simply returning an error code?

If attempting to use a function isn't the proper way to do feature checks in Java then someone who understands the language needs to tell us how to do it.


Rik <rik5>
Group administrator
Thu 08 Mar 2018 09:35:06 AM UTC, comment #40: 

Sebastain, agreed, if this issue only happens when calling awt it is probably not a configuration problem for our development machine so we can open a new bug and finally really close this thread.

Carlo de Falco <cdf>
Group Member
Thu 08 Mar 2018 09:20:58 AM UTC, comment #39: 

The problem only occurs when calling usejava ("awt"). You can see it from the command line, too. It is only present in this single test. I am not sure where the problem comes from. This can be  macOS or octave issue (Java was previously supplied by Apple (version  <=1.6) and now it comes from Sun (>1.6). Sun's Java can be treated on macOS like on Unix, while Apple's Java needed in former times special treatment during the linking "-framework...". Even today, if using "-framework..." the old dialog pops up to install Apple's Java 1.6. However, I could not find such thing in the current development sources. On the other hand, I have compiled java code on macOS that contains awt without issues.

I suggest the following: I will open a new bug report for usejava("awt") on macOS. Then, somebody marks this test as known to fail on macOS. Agreed?

Seb.

Sebastian <sschoeps>
Thu 08 Mar 2018 08:40:30 AM UTC, comment #38: 

Is that a problem with that specific test only or with any test that uses java?

If it is a general problem I would just re-add the --disable-java flag in the buildbot options, if it is only for that specific test I agree to mark it as a known failure.

In the latter case I think we need a bug report to reference, could you please create one?

Carlo de Falco <cdf>
Group Member
Thu 08 Mar 2018 07:54:42 AM UTC, comment #37: 

Dear Carlo,

I guess this is a macOS bug. The test that contains 

usejava ("awt")

in "ov-java.cc" makes macOS try to install the old and outdated Java 1.6. This triggers a GUI dialog which craches because there is no graphical frontend running.

I suggest to remove that test on macOS and mark it as a known failure.

Best
Sebastian

Sebastian <sschoeps>
Thu 08 Mar 2018 01:39:21 AM UTC, comment #36: 

sorry it appears that


/Library/Java/JavaVirtualMachines/jdk-9.0.4.jdk/Contents/Home


doe exist un the buildbot machine. nonetheless now make check is crashing with the following error


libinterp/octave-value/ov-java.cc-tst .......................No Java runtime present, requesting install.
2018-03-08 02:00:40.903 octave-cli[12874:8335154] JLRequestRuntimeInstall: Error calling: CFMessagePortCreateRemote



Carlo de Falco <cdf>
Group Member
Tue 06 Mar 2018 09:35:32 PM UTC, comment #35: 

I don't know why java was disabled for this build.

I set JAVA_HOME in the buildbot user's .profile file using the command from your shell script:


export JAVA_HOME="$(/usr/libexec/java_home -v 9)"


Here's what I see on the buildbot system:


mathicsepc32:buildbot jwe$ ls -l $(/usr/libexec/java_home -v 9)
total 16
-r--r--r--   1 root  wheel   158 Dec 20 03:14 README.html
drwxr-xr-x  49 root  wheel  1666 Dec 20 03:17 bin
drwxr-xr-x   7 root  wheel   238 Dec 20 03:14 conf
drwxr-xr-x  11 root  wheel   374 Dec 20 03:14 include
drwxr-xr-x  99 root  wheel  3366 Dec 20 03:14 jmods
drwxr-xr-x  99 root  wheel  3366 Dec 20 03:14 legal
drwxr-xr-x  94 root  wheel  3196 Dec 20 03:15 lib
-rw-r--r--   1 root  wheel  1924 Dec 20 03:14 release


and


mathicsepc32:buildbot jwe$ ls -l $(/usr/libexec/java_home -v 9)/include/jni.h
-rw-r--r--  1 root  wheel  73955 Dec 20 03:14 /Library/Java/JavaVirtualMachines/jdk-9.0.4.jdk/Contents/Home/include/jni.h


I can try removing the --disable-java flag and see whether it actually works.

John W. Eaton <jwe>
Group administrator
Tue 06 Mar 2018 07:48:33 PM UTC, comment #34: 

@Simone, where did you install java 9?

The build currently works but java is
disable because it is not found ...

The value currently set for JAVA_HOME is


JAVA_HOME=/Library/Java/JavaVirtualMachines/jdk-9.0.4.jdk/Contents/Home


but that path does not exist.

@JWE it seems that also configure is done with "--disable-java" is that intentional?



Carlo de Falco <cdf>
Group Member
Tue 06 Mar 2018 05:38:45 PM UTC, comment #33: 

I pushed the following changeset:

http://hg.savannah.gnu.org/hgweb/octave/rev/2ee45d3f6ec1

HAVE_QT_OFFSCREEN implies Qt.  Other tests that require a graphical display also check have_window_system so this should be done anyway and won't need to be changed if we eventually have a way to create a virtual display on the OS X buildbot system.

John W. Eaton <jwe>
Group administrator
Tue 06 Mar 2018 04:46:49 PM UTC, comment #32: 

In the short term, the BIST tests for getframe could be altered to make sure they are running the qt toolkit, rather than gnuplot.

Rik <rik5>
Group administrator
Tue 06 Mar 2018 03:46:56 PM UTC, comment #31: 

Is there any way to achieve a similar result as what we get with xvfb-run on other systems?  To run the graphics tests, we need to be able to create some kind of virtual graphics device that behaves the same as if we had a real display.  Does aqua provide that in some way?

John W. Eaton <jwe>
Group administrator
Tue 06 Mar 2018 03:18:43 PM UTC, comment #30: 

ooops, sorry for the missing text!
I don't think it makes any sense to
use xvfb on the mac, we should (and AFAIK we are)
only use aqua and not X11.

And xvfb is not installed anyway.

Carlo de Falco <cdf>
Group Member
Tue 06 Mar 2018 02:59:59 PM UTC, comment #29: 

A bad verbatim tag caused text to be omitted from comment 27.  Here it is:


!!!!! test failed
getframe: not implemented for gnuplot graphics toolkit
***** testif HAVE_QT_OFFSCREEN
 hf = figure ("visible", "off");
 unwind_protect
   hax = axes ("visible", "off", "position", [0 0 1 1]);
   verts = [0 0; .5 0; 1 0; ...
            0 .5; .5 .5; 1 .5; ...
            0 1; .5 1; 1 1];
   faces = [1 2 5 4; 2 3 6 5; 4 5 8 7; 5 6 9 8];
   fvc = [1 0 0; 0 1 0; 0 0 1; 1 0 1];
   patch ("vertices", verts, "faces", faces, "facevertexcdata", fvc, ...
          "facecolor", "flat");

   kk = 1;
   pos = get (hf, "position");

   for jj = [0.05 0.55]
     for ii = [0.05 0.55]
       rect = [ii jj .4 .4].*[pos(3:4) pos(3:4)];
       frame = getframe (hax, rect).cdata;
       assert (frame(:,:,1) == fvc(kk,1)*255);
       assert (frame(:,:,2) == fvc(kk,2)*255);
       assert (frame(:,:,3) == fvc(kk,3)*255);
       kk++;
     endfor
   endfor
 unwind_protect_cleanup
   close (hf);
 end_unwind_protect
!!!!! test failed
getframe: not implemented for gnuplot graphics toolkit


Ah.  On other systems, we run the make check process with a virtual framebuffer:


xvfb-run -a -s '-screen 0 640x480x24' make V=1 check


From a quick check for xvfb-run on OS X it wasn't clear to me whether this is possible on OS X.

John W. Eaton <jwe>
Group administrator
Tue 06 Mar 2018 02:37:20 PM UTC, comment #28: 

The build succeeds now, so closing as fixed.

There are 2 test failures for getframe, but that's a separate issue and probably not unique to Clang or OS X since we are seeing the same tests failing on other systems.

John W. Eaton <jwe>
Group administrator
Tue 06 Mar 2018 02:34:38 PM UTC, comment #27: 

buildbot seems to be running correctly but make check is producing some failures.

All failures that I see are of the following type:


!!!!! test failed
getframe: not implemented for gnuplot graphics toolkit
***** testif HAVE_QT_OFFSCREEN
 hf = figure ("visible", "off");
 unwind_protect
   hax = axes ("visible", "off", "position", [0 0 1 1]);
   verts = [0 0; .5 0; 1 0; ...
            0 .5; .5 .5; 1 .5; ...
            0 1; .5 1; 1 1];
   faces = [1 2 5 4; 2 3 6 5; 4 5 8 7; 5 6 9 8];
   fvc = [1 0 0; 0 1 0; 0 0 1; 1 0 1];
   patch ("vertices", verts, "faces", faces, "facevertexcdata", fvc, ...
          "facecolor", "flat");

   kk = 1;
   pos = get (hf, "position");

   for jj = [0.05 0.55]
     for ii = [0.05 0.55]
       rect = [ii jj .4 .4].*[pos(3:4) pos(3:4)];
       frame = getframe (hax, rect).cdata;
       assert (frame(:,:,1) == fvc(kk,1)*255);
       assert (frame(:,:,2) == fvc(kk,2)*255);
       assert (frame(:,:,3) == fvc(kk,3)*255);
       kk++;
     endfor
   endfor
 unwind_protect_cleanup
   close (hf);
 end_unwind_protect
!!!!! test failed
getframe: not implemented for gnuplot graphics toolkit
-verbaim-

Carlo de Falco <cdf>
Group Member
Mon 05 Mar 2018 10:02:13 PM UTC, comment #26: 

The commands that are executed are stored in the configuration file for the buildbot master.  Normally, buildbot expects the environment to be set up on the worker system so that configure and make will work without a lot of special options.

I set the PATH and JAVA_HOME variables in the $HOME/.profile file for the buildbot user on the OS X worker system.

It seems better to set the CPPFLAGS, LDFLAGS, and PKG_CONFIG_PATH as argument to the configure command, so I set those in the buildbot master config file for the osx target.

The build is running now.

John W. Eaton <jwe>
Group administrator
Mon 05 Mar 2018 08:57:55 PM UTC, comment #25: 

Hi,

Texinfo is installed by homebrew keg-only it is available in


 /usr/local/opt/texinfo/{lib,bin,share}


This looks like it is version 6.5:


$ /usr/local/opt/texinfo/bin/makeinfo --version
texi2any (GNU texinfo) 6.5

Copyright (C) 2017 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.


@Sebastian, the appropriate flags are already set in the configure script I am using (provided by Ben and slightly updated), I attach the full script again.

@JWE where is the script used by the buildbot placed?





(file #43467)

Carlo de Falco <cdf>
Group Member
Mon 05 Mar 2018 08:51:52 PM UTC, comment #24: 

Version 6.5 is installed as "/usr/local/opt/texinfo/bin/makeinfo".

@Carlo: can you add the path, please?

Sebastian <sschoeps>
Mon 05 Mar 2018 08:19:31 PM UTC, comment #23: 

The build is now failing with


checking for makeinfo... makeinfo
checking for makeinfo support for @sortas command... no
configure: WARNING:

I wasn't able to find a version of GNU makeinfo that supports the
@sortas command, but it's only a problem if you need to build the
manual, which is the case if you're building from VCS sources.


We need Texinfo 6.1 or later.

http://lists.gnu.org/archive/html/octave-maintainers/2018-01/msg00016.html

John W. Eaton <jwe>
Group administrator
Tue 27 Feb 2018 02:13:48 AM UTC, comment #22: 

Can someone kick the buildbot?  It doesn't seem to have restarted after all the updates.

Rik <rik5>
Group administrator
Mon 26 Feb 2018 04:46:41 PM UTC, comment #21: 

I have installed Java 9 this morning.
Best
Simone

Simone Deparis <deparis>
Mon 26 Feb 2018 09:07:58 AM UTC, comment #20: 

sundials was automatically detected after you installed it.
java 9 needs to be installed via gui so it is still disabled as it is not found.

Carlo de Falco <cdf>
Group Member
Mon 26 Feb 2018 06:25:17 AM UTC, comment #19: 

@Carlo: I thought sundials (and potentially Java9).

Seb.

Sebastian <sschoeps>
Sun 25 Feb 2018 05:47:53 PM UTC, comment #18: 

what is it that needs fixing in the script?
It seems to be working as is ...

Carlo de Falco <cdf>
Group Member
Sat 24 Feb 2018 11:42:00 AM UTC, comment #17: 

Dear all,

all dependencies but Java should be in place. If Carlo fixes the installation script than we should be good to go.

@Simone: can you install Java9, please?

Sebastian

Sebastian <sschoeps>
Fri 23 Feb 2018 11:00:42 AM UTC, comment #16: 

Yes, it works. I am currently (re)installing suite-sparse, arpack, qrupdate and sundials27 compiled with openblas.

Sebastian

Sebastian <sschoeps>
Fri 23 Feb 2018 10:21:20 AM UTC, comment #15: 

@Sebastian I did change permissions for the contents of /usr/local but I'm not sure yet whether you can now install homebrew package, can you please try installing sundials 2?

Apart from sundials and Java the attached modification of Ben's script is working for me on the buildbot mac


#! /bin/sh

## Start with Octave to fulfill most dependencies
# brew install octave
# brew uninstall octave
## Use Sundails27 instead of Sundials
# brew unlink sundials
# brew install sundials27
## Add qscintilla2
# brew install qscintilla2
## Use Homebrew's texinfo
# brew install texinfo
## Use Homebrew's bison
# brew install bison

## Install MacTex if building docs is desired
## https://tug.org/mactex/mactex-download.html

## Apply Daniel Sebald's patch for Qt
# patch -p1 < ../octave-qt_wait_on_worker_thread-DEBUG1-djs2017sep25.patch
# This fix only works with the fprintf (stderr, ...) debug lines.

## Set Java's home directory
## export JAVA_HOME="$(/usr/libexec/java_home -v 9)"

# Setup dependencies that are not symbolically linked by Homebrew
LDFLAGS=""
CPPFLAGS=""
LDFLAGS="$LDFLAGS  -L/usr/local/opt/readline/lib"
CPPFLAGS="$CPPFLAGS -I/usr/local/opt/readline/include"
LDFLAGS="$LDFLAGS -L/usr/local/opt/sqlite/lib"
CPPFLAGS="$CPPFLAGS -I/usr/local/opt/sqlite/include"
LDFLAGS="$LDFLAGS -L/usr/local/opt/openssl/lib"
CPPFLAGS="$CPPFLAGS -I/usr/local/opt/openssl/include"
LDFLAGS="$LDFLAGS -L/usr/local/opt/bison/lib"
LDFLAGS="$LDFLAGS -L/usr/local/opt/gettext/lib"
CPPFLAGS="$CPPFLAGS -I/usr/local/opt/gettext/include"
LDFLAGS="$LDFLAGS -L/usr/local/opt/libffi/lib"
LDFLAGS="$LDFLAGS -L/usr/local/opt/icu4c/lib"
CPPFLAGS="$CPPFLAGS -I/usr/local/opt/icu4c/include"
LDFLAGS="$LDFLAGS  -L/usr/local/opt/texinfo/lib"
LDFLAGS="$LDFLAGS  -L/usr/local/opt/qt/lib"
CPPFLAGS="$CPPFLAGS -I/opt/X11/include"
LDFLAGS="$LDFLAGS  -L/opt/X11/lib"
export LDFLAGS
export CPPFLAGS

# Add ICU4C to the pkg-config path
PKG_CONFIG_PATH="/usr/local/opt/icu4c/lib/pkgconfig"
# Add Qt5 to the pkg-config path
PKG_CONFIG_PATH="$PKG_CONFIG_PATH:/usr/local/opt/qt/lib/pkgconfig"
export PKG_CONFIG_PATH

# Apple's bison [yacc] is too old, so use Homebrew's
PATH="/usr/local/opt/bison/bin:$PATH"
# Add texinfo to the PATH
PATH="/usr/local/opt/texinfo/bin:$PATH"
# Add Qt to the PATH
PATH="/usr/local/opt/qt/bin:$PATH"
export PATH

PREFIX="/usr/local/$(whoami)/octave"
../octave/configure --prefix=$PREFIX \
      --disable-silent-rules \
      --enable-link-all-dependencies \
      --enable-shared \
      --disable-static \
      --disable-docs \
      --without-OSMesa \
      --with-hdf5-includedir=/usr/local/opt/hdf5/include \
      --with-hdf5-libdir=/usr/local/opt/hdf5/lib \
      --with-x=no \
      --with-blas="-L/usr/local/opt/openblas/lib -lopenblas" \
      --with-portaudio \
      --with-sndfile


Carlo de Falco <cdf>
Group Member
Fri 23 Feb 2018 08:22:43 AM UTC, comment #14: 


> Homebrew tries to have no redundant libraries in public places that is the reason behind the warning. I think for example MacPorts does not care at all.


Macports does not need to care at all as it does not install ANYTHING in public places. In order to use MacPorts provided libraries one has to specify the appropriate compiler/linker flags.

The warning seems to indicate tha homebrew suggests to do the same with readline. I think this is probably better as if we do it this way the buildbot machine will more closely resemble the standard situation of a generic mac user and also we will clearly mark in the configure options that we need to avoid the apple readline and use gnu readline.

I'll try Ben's script.


> Did you compile everything from scratch or did you use the precompiled binaries? (if not using "--build-from-source" homebrew tries to get a binary). I dot see this warning.


I installed all with binaries, not building anything from source. Probably building from source it would be possible to specify what libc to link to?


> @carlo: it seems that I am admin but not allowed to write to "/usr/local/". Can you give the admin-group access to that folder (it's owned by Carlo)?


that was set-up by homebrew automatically, I can change group ownership if it is OK.

c.


Carlo de Falco <cdf>
Group Member
Fri 23 Feb 2018 07:47:08 AM UTC, comment #13: 


> Is it safe to go ahead? Do I need to reinstall other packages that depend on readline?


It should work. I have done the same when building the Octave-App and did not see any problems. Homebrew tries to have no redundant libraries in public places that is the reason behind the warning. I think for example MacPorts does not care at all. However, if you look for a cleaner solution, you may keep readline keg-only and add the paths to the environment variables like Ben proposed here: https://savannah.gnu.org/bugs/download.php?file_id=42932.

> I also got warnings when installing other dependencies such as hdf5 about use of different libc versions, any idea how to avoid this?


Did you compile everything from scratch or did you use the precompiled binaries? (if not using "--build-from-source" homebrew tries to get a binary). I dot see this warning.

@carlo: it seems that I am admin but not allowed to write to "/usr/local/". Can you give the admin-group access to that folder (it's owned by Carlo)?

Best
Sebastian

Sebastian <sschoeps>
Fri 23 Feb 2018 05:20:33 AM UTC, comment #12: 

I installed homebrew without sudo so I think I can do that myself.


I get the following warning:


Warning: readline is keg-only and must be linked with --force
Note that doing so can interfere with building software.


Is it safe to go ahead? Do I need to reinstall other packages that depend on readline?

I also got warnings when installing other dependencies such as hdf5 about use of different libc versions, any idea how to avoid this?


Carlo de Falco <cdf>
Group Member
Thu 22 Feb 2018 10:28:48 PM UTC, comment #11: 

The easiest would be to run "brew unlink readline && brew link readline" and force brew to link it in /usr/local. My account does not seem to have sufficient rights, so couldn't do it myself.

@Simone: can you check if I have admin-rights? If yes, I could install qt, sundials etc...

Sebastian

Sebastian <sschoeps>
Thu 22 Feb 2018 05:56:14 PM UTC, comment #10: 

Hi,

Xcode is now working and command line tools are installed.

Octave configure fails because the macos readline is incompatible
with gnu readline and the one installed by homebrew is keg-only.

Sebastian, maybe you can help add the appropriate flags to access the keg-only readline provided by homebrew?

I am attaching the configure script I tried, some features are disabled (sundials, qt, ...) as I didn't install the respective dependencies


../octave/configure \
    --prefix=/opt/octave/4.3.0+  \
    --without-osmesa --with-framework-carbon \
    --without-x --disable-static --disable-openmp \
    --disable-java \
    --disable-jit --with-sndfile --with-portaudio --with-magick=GraphicsMagick \
    --enable-docs --with-amd="-lamd -lsuitesparseconfig" \
    --with-camd="-lcamd -lsuitesparseconfig" \
    --with-colamd="-lcolamd -lsuitesparseconfig" \
    --with-cholmod="-lcholmod -lsuitesparseconfig" \
    --with-cxsparse="-lcxsparse -lsuitesparseconfig" \
    --with-umfpack="-lumfpack -lsuitesparseconfig" \
    --without-openssl --with-blas="-L/opt/local/lib -lopenblas" \
    --with-lapack="-L/opt/local/lib -lopenblas" \
    CPP="/usr/bin/cpp" \
    CC="/usr/bin/clang"
    SED=gsed \
    CXX="/usr/bin/clang++" \
    F77="gfortran" \
    CXXFLAGS="-std=gnu++11" \
    --with-fltk --with-framework-opengl \
    --without-qt --enable-link-all-dependencies


Carlo de Falco <cdf>
Group Member
Thu 22 Feb 2018 02:34:48 PM UTC, comment #9: 


> Simone,
>
> I am getting the following error:
> $
> xcode-select: error: no developer tools were found, and no install could be
> requested (perhaps no UI is present), please install manually from
> 'developer.apple.com'.
>  Any idea why is this happening if XCode is correctly installed?


Sorry, I had to run Xcode. I also ran  xcode-select --install, should be fine now.

> c.
> P.S. did you get this thread off the tracker on purpose?


I did lose my passwd for Savannah. ...

Ciao
Simone


Simone Deparis <deparis>
Wed 21 Feb 2018 10:55:37 AM UTC, comment #8: 

I have been having problems maintaining a stable ssh connection with that machine, but from what I have seen it needs XCode and homebrew to be reinstalled. it seems installing XCode from the command line is very complicated.

Simone, could you please update XCode from consolle?

c.




Carlo de Falco <cdf>
Group Member
Tue 20 Feb 2018 07:11:08 PM UTC, comment #7: 

The poor Mac buildbot now seems to be failing bootstrap because of  a lack of readline library.

Rik <rik5>
Group administrator
Sat 17 Feb 2018 04:19:34 AM UTC, comment #6: 

I restarted it again.  I'm not sure why the buildbot worker process keeps failing there.  Maybe we should create some script that can run as a cron job to check its status and restart it if necessary.

John W. Eaton <jwe>
Group administrator
Sat 17 Feb 2018 02:05:10 AM UTC, comment #5: 

The Mac buildbot has gone offline again, and 16 builds pending.  Can someone reboot it again?

Rik <rik5>
Group administrator
Wed 14 Feb 2018 06:56:52 AM UTC, comment #4: 

Yes,

We should keep the machine updated.

It is to be expected that some things may brake after an upgrade,
fixing them is the purpose of the testing machine.

I'm not sure we actually need the X server at all,
but if we do it's better to have the latest.

c.


Carlo de Falco <cdf>
Group Member
Wed 14 Feb 2018 06:49:42 AM UTC, comment #3: 

Dear Carlo,

I try to keep the epfl server machine uptodate.

Is that what you usually prefer ?

It is possible that the X server is not the latest. Do you want me to do the upgrade ?

Best
Simone
PS: I answered you privately about the passwd.

Simone Deparis <deparis>
Wed 14 Feb 2018 06:38:19 AM UTC, comment #2: 

Hi,

I am currently able to build on my own mac and get 0 FAIL
(but quite a few XFAIL) in make check, so this should be easily fixed.


IIRC since I installed those packages there were a couple of system upgrades on mathicsepc32, that may be related to this issue.

BTW I cannot log into mathicsepc32 as I lost my private key when
my previous macbook broke down, can someone with administrator privileges give me access?

CC-ing Simone and Sebastian.

c.

Carlo de Falco <cdf>
Group Member
Wed 14 Feb 2018 06:04:58 AM UTC, comment #1: 

It looks like bison is failing:


mathicsepc32:build buildbot$ bison
Illegal instruction: 4
mathicsepc32:build buildbot$ bison -v
Illegal instruction: 4
mathicsepc32:build buildbot$ which bison
/usr/local/bin/bison
mathicsepc32:build buildbot$ ls -l /usr/local/bin/bison
lrwxr-xr-x  1 carlo  admin  31 Sep 14  2016 /usr/local/bin/bison -> ../Cellar/bison/3.0.4/bin/bison


Adding carlo to the list since it looks like he's the one that installed bison on that system.

I'm not sure what happened here.  Maybe the tools need to be updated?  Did an OS upgrade break them?

John W. Eaton <jwe>
Group administrator
Tue 13 Feb 2018 11:44:47 PM UTC, original submission:  

The Mac buildbot seems to be failing because the machine lacks a package and is calling the build-aux/missing script in some weird manner.  The latest log shows:


uic  -o libgui/src/ui-settings-dialog.h ../src/libgui/src/settings-dialog.ui
uic  -o libgui/graphics/ui-annotation-dialog.h ../src/libgui/graphics/annotation-dialog.ui
rm -f test/conv.tst-t test/conv.tst && \
        /bin/sh ../src/test/mk-conv-tst.sh > test/conv.tst-t && \
        mv test/conv.tst-t test/conv.tst
rm -f test/sparse.tst-t test/sparse.tst && \
        /bin/sh ../src/test/mk-sparse-tst.sh > test/sparse.tst-t && \
        mv test/sparse.tst-t test/sparse.tst
/Users/buildbot/epfl-elcapitan-x86_64/clang-osx/build/../src/build-aux/missing: line 81: 22089 Illegal instruction: 4  "$@"
/Users/buildbot/epfl-elcapitan-x86_64/clang-osx/build/../src/build-aux/missing: line 81: 22087 Illegal instruction: 4  "$@"
make: *** [libinterp/parse-tree/oct-parse.h] Error 132
make: *** Waiting for unfinished jobs....
make: *** [libinterp/corefcn/oct-tex-parser.h] Error 132


Rik <rik5>
Group administrator

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach Files:
   
   
Comment:
   

Attached Files
file #43467:  buildcommand.sh added by cdf (2KiB - application/x-sh)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by sschoeps (Posted a comment)
  • -email is unavailable- added by deparis (Posted a comment)
  • -email is unavailable- added by cdf (Posted a comment)
  • -email is unavailable- added by cdf
  • -email is unavailable- added by cdf
  • -email is unavailable- added by jwe (Posted a comment)
  • -email is unavailable- added by jwe
  • -email is unavailable- added by rik5 (Submitted the item)
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

    Only group members can vote.

     

    Follow 7 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2018-03-06 jwe StatusIn Progress Fixed
        Open/ClosedOpen Closed
    2018-03-05 cdf Attached File- Added buildcommand.sh, #43467
    2018-02-14 rik5 StatusNone In Progress
    2018-02-14 cdf Carbon-Copy- Added -email is unavailable-
        Carbon-Copy- Added -email is unavailable-
    2018-02-14 jwe Carbon-Copy- Added cdf

    Back to the top

    Powered by Savane 3.13-4448.
    Corresponding source code