FreeFem++-cs
an integrated environment for FreeFem++
Antoine Le Hyaric, Jacques-Louis Lions Laboratory, Pierre and Marie Curie University.

 

Bugs

Known bugs

FreeFem++-server should run FreeFem++-lang as user nobody

Border keyword highlight problem in ffcs editor (eg with svg2edp)

There is still a problem with autosave when the default directory is readonly

Run native FreeFem++ regression tests for ffvtk

Closed TCP server connections stay indefinitely in CLOSE_WAIT state after completion

check autosave feature again (problem when initial directory is not writable)

FreeFem++-cs crashes when resizing graphical windows (reported by JR)

ff/upstream/download: scotch has problems compiling on Ubuntu (-lm library order problem), pastix and hips depend on scotch

ff/upstream/download: hypre/Makefile explicitely refuses to work on Win32

ff/upstream/download: mmg3d does not patch correctly any more (missing file build/mmg3d.c ???)

ff/upstream/download: yams download is broken

Improvements for graphical windows (request from MK)

- add user-defined messages to the FreeFem++-cs "cin" input window - enable setting 2D mesh borders to black - colorize the inside of meshes (to visualize distinct regions) - blinking effect when resizing graphical windows under MacOS - numerical values are too small (and there are not enough of them) - force all pictures to use the same scale - options for comments: location, character size, line returns - because of shadows, colors do not correspond to the reference any more when the light source moves - make interpolated colors (eg when drawing isosurface triangles) crispier

1D (x,y) plots (request from ID)

ability to run regression test suite from read-only directory (necessary e.g. on Debian or on Windows)

convert FreeFem++ doc as wiki pages for [http://webs.um.es/eliseo/ff++]

list all command-line options in the doc

test that the same DLL can work with FreeFem++ and FreeFem++-cs

undo does not always work in the editor

Choose starting points for streamlines

in FreeFem++-vtk, add a message when FreeFem++ is done

explain the fact that the graphical window pauses when the mouse is over it

unstable color scale in examples++-chapt3/convects.edp

examples++-main/wafer-heating-laser-axi.edp pictures are blank

  • it looks like VTK is clipping objects of size 1000 or over.
  • ctrl-z does not remove modified flag for an edited file

    export to HTML for color printing of syntax colors

    dgeev syntax error not detected by regression tests?

    remove examples that do not work from precompiled packages

    add a typing test in the regression tests

    mmg3d is automatically downloaded and compiled now?

    tests should be able to run from a read-only directory (eg for Debian and Windows setups)

    check FAQ questions about FreeFem++-cs batch mode

    FreeFem++-cs should complain with a clear explanation when MPI is not installed

    avoid escaping basic quote characters in FreeFem++-cs comments

  • [file:~/LeHyaric/ffcs/comments.php::All_quotes_are_escaped_by_default]
  • Platform-specific bugs

    Linux: precompiled version for Ubuntu 10.04 LTS (request from Shyam)

    Linux: there have been reports that "install_on_gnome_desktop" bugs on the latest Ubuntu

    MacOS: some trouble with MPI?

  • regression tests need to be run again
  • MacOS: "pointer being freed was not allocated" in dist version

  • only when compiling with optimisation options -O{1,2,3}
  • this happens since the last automatic Xcode update (mid-december 2010)
  • keep compilation option -D_GLIBCXX_FULLY_DYNAMIC_STRING' to solve it?
  • even VTK crashes: problem in VTK/Utilities/kwsys/SystemTools.cxx and Utilities/MaterialProperties/ProcessShader.cxx
  • just wait until the next Xcode update?
  • distribute an unoptimized MacOS version for now?
  • MacOS: FreeFem++-cs does not work with Lion (OP)

  • Message: dyld: Library not loaded: /usr/X11/lib/libX11.6.dylib Referenced from: /usr/X11/lib/libXrender.1.dylib Reason: Incompatible library version: libXrender.1.dylib requires version 10.0.0 or later, but libX11.6.dylib provides version 9.0.0
  • MacOS: Changing the size of the graphical output window (FH)

  • makes it blink when adding new pictures
  • MacOS: some users still use MacOS 10.5

    Windows: heuristic virus scanners

  • These virus scanners produce positive reports on FreeFem++-cs without seeing any actual virus signature
  • One could try and find the FreeFem++-cs (clean) code area that activates the heuristic via dichotomy and the virustotal.com website
  • Windows: problem with make MPICG.dll and many other DLL

    version 12.5

    Complete regression tests for all platforms

    version 12.4 (10/5/12)

    DONE load "tetgen" now works again

    Configuration option "--enable-debian" is now automatically activated on Debian systems

    Developed a new permanent server architecture named FreeFem++-server

  • The previous non-permanent server process has been renamed as FreeFem++-lang. This is transparent to the user.
  • A text-based client named FreeFem++-cli has been developed as a testing procedure for FreeFem++-server. It can also be used as a command-line version of FreeFem++-cs,
  • All the functions of the previous ssh-based server architecture have been replaced.
  • Access to this permanent server should always be restricted to identified users (via the password file provided in FreeFem++-cs/passwd for instance).
  • DONE renamed source archives as ffcs-...-src.tgz to simplify automatic web scans (request for Debian)

    Only displays a message about saving files in verbose mode

    Upgraded to FreeFem++ version 3.19

    Added configuration options to specifically disable any FreeFem++ extra tool that fails to download or compile

    version 12.3 (7/3/12)

    Re-added "--disable-mpi" as a configure option on user request

    DONE problem with links during make in doc subdirectory solved

  • mainly used for the FreeFem++ reference card on the web
  • Compiled successfully on Windows 7 (32 and 64 bits)

    version 12.2 (14/2/12)

    DONE Regression test suite for FreeFem++-vtk (script located in tests/checkffg) reactivated

    Rebuilt on MacOS 10.7

    version 12.1 (1/2/12)

    DONE bug: when pictures were saved in quick succession, some of the recorded files were corrupt

  • This was due to the fact that VTK (rightly) requires the screen representation of a scene to be complete before saving it to a file. But FLTK does not draw windows (and the scene that they contain) on the screen if they are covered by other ones quickly before the FLTK idle loop gains control. The fix was to wait for any picture be saved before creating new VTK scenes.
  • All FreeFem++-cs command-line options are documented

  • To see the full description of all options, just call FreeFem++-cs from the command line with option "-h" ("--h", "-help" and "--help" also work)
  • Included patches by CTrophime to use the default FLTK package under Linux Debian

    Upgraded to FreeFem++ version 3.18

    version 11.14 (5/12/11)

    Upgraded to FreeFem++ version 3.17

    DONE ARGV can now be set from the FreeFem++-cs command line with the option -FreeFemArguments "___"

    The "-Batch" option allows calling FreeFem++-cs from the command line, run a script and return to the command line

    version 11.13 (22/11/11)

    DONE Compiled with MPI on Ubuntu 10.04 LTS, Windows 7 and MacOS Lion

  • Some MPI-enabled shared libraries are still not available in the precompiled Windows version because of unsolved "undefined reference" linking problems.
  • Version 11.12 (14/11/11)

    Upgraded to FreeFem++ version 3.16

    DONE The autosave file (default name autosave.edp) is now saved only when new changes occur in the editor

    Compiled with the Goto BLAS on Windows (http://www.tacc.utexas.edu/tacc-projects/gotoblas2)

    DONE Solved freeze problem when typing macros

    DONE Enabled C comments inside macro definitions

    DONE The color-coded editor now recognises the new FreeFem++ macro syntax ("NewMacro/EndMacro")

    DONE ARGV now available in FreeFem++-cs

  • The ARGV array now contains arguments specified as strings in "Options/Other Options.../FreeFem++ arguments"
  • This version has been validated on Linux Debian Testing and Ubuntu 10.04 64 LTS

  • Windows and MacOS versions are scheduled
  • DONE On Debian Linux, the Debian VTK library is used when specifying "--enable-debian" in the FreeFem++-cs configuration step

    Version 11.11 (5/10/11)

    Upgraded FreeFem++ to version Pi

  • FreeFem++ 3.14 picks up the latest version of mshmet (the old version of mshmet is not available anymore so mshmet is not compiled by FreeFem++ 3.12). Examples++-3d/Laplace-Adapt-3d.edp works again.
  • Compiled with gcc/g++ version 4.6

    Some internal executable names have been changed to avoid ambiguïties with other packages

  • The python-ffc Debian package already contains a file named "ffc". Therefore, our ffc has be renamed to FreeFem++-client, and for coherency ffs -> FreeFem++-server, ffsmpi -> FreeFem++-mpiserver. This has no impact on FreeFem++-cs users.
  • NB: Compiling VTK is problematic with this version

  • so it should only be used on Debian-compatible Linux versions (eg Ubuntu), where a precompiled VTK lib can be used with the "--enable-debian" configuration option.
  • Version 11.10 (19/5/11)

    DONE works on 32-bit Linux Ubuntu 11.04

  • the problem with Ubuntu was that X11 libraries are not located in /usr/lib anymore but in /usr/lib/{i386,x86_64}-linux-gnu from Ubuntu version 11.04
  • Version 11.9 (6/4/11)

    DONE removed executable bit from non-executable files in source files.

    DONE Created two very basic man pages for use in Linux packages

  • they are doc/FreeFem++-cs.1 and doc/ffg.1
  • Version 11.8 (2/4/11)

    Menu option "Graphics/Save Image As..." renamed to "Save Scene As..."

  • because some file formats (vtk, vtu, povray) do not save an image but the whole 3D scene.
  • Cut planes are now deactivated by default

  • This is to avoid cluttering the picture, but it can always be re-activated by hand (in "Graphics/Options...") or by an option in plot(...,CutPlane=true,...)
  • The preferences file freefem++.pref has been renamed to freefem++-cs.pref

  • this is to make sure that FreeFem++ and FreeFem++-cs can coexist without any interference on the same machine.
  • New configure option "--enable-debian"

  • to use Debian-provided VTK and cmake packages instead of downloading them when creating a Debian package (request from C. Trophime)
  • Some patches for Debian packaging provided by C. Trophime

    Version 11.7 (8/3/11)

    DONE accented utf-8 characters do not crash FreeFem++-cs anymore

  • this long-standing problem has been solved by calling yyrestart() when a highlighting scan ends in an error because of an utf-8 escape character.
  • The FreeFem++-cs window is now reduced when used on smaller screens

  • because some window managers may accept a big window but make it difficult to move such a window if its control points are out of the display area
  • DONE Windows: tetgen.dll loading problem solved

  • the visible error message was a _gxx_personality_v0 problem in libstdc++-6.dll
  • The window where a plot appears can now be chosen

  • thanks to a new plot() parameter: WindowIndex=n, where n goes from -1 (use current window), 0 (use main window), 1 or more (create a detached window of that label)
  • All VTK plot() parameters have a textual description

  • obtained by clicking on "Graphics/Parameters for plot()"
  • DONE A race condition has been resolved at the end of each FreeFem++ run

  • the freeze happened randomly. The locks involved were the FLTK lock and 'runserverend'.
  • DONE Windows bug: Saving a file could fail but say "completed successfully"

  • this was a bug in the error checking mechanism!
  • DONE Windows bug: Remote processes are now properly killed

  • sometimes, killing a remote process did not work because of an internal memory allocation problem
  • Version 11.6 (28/2/11)

    Default for "Help/Show Debug Messages" is now off

    New "Graphics" menu summarising all commands for the graphical window

    3D surfaces are now scaled to the size of the graphical window

  • this can be changed with "Adjust Z scale" on the graphical options window
  • plot(ps="") can now save pictures in multiple formats

  • depending on the file name extension: .eps (pixel-based), .png, .jpg, .pov, .vtk, .vtu
  • the FreeFem++ 2D vector-based postcript output can still be obtained by choosing the extension .ps
  • plot() now accepts VTK options

  • to get the list of available options, click "Graphics/Show Plot Parameters". All these options can be copied back into the original .edp script.
  • DONE bug: python not called anymore during Linux installation

  • calling python is a MacOS-only replacement for "readlink -f", which is missing on this platform, during the "install_as_root" procedure. Python should never get called on any Linux version.
  • New FreeFem++-cs configuration option "--enable-tools-failure"

  • By default, FreeFem++-cs does not compile without all the FreeFem++ tools, to make sure that the FreeFem++-cs environment is coherent on all platforms. This options allows some tools downloaded by FreeFem++ not to compile to let users run FreeFem++-cs even on platforms where some of the tools do not work.
  • DONE .vtu and .vtk files numbers are now placed before the extension

    DONE autosave message now specifies the autosave file name

  • before it wrongly specified the original name as saved when only the autosave copy was updated.
  • DONE cut plane now cuts through all 3D vector fields

    Version 11.5 (24/2/11)

    DONE MacOS bug: ffg now works again with FreeFem++

    DONE bug: solved some indexing errors in the reference card

    DONE bug: disabling editor highlighting now removes all colors right away

    DONE bug: "Edit/Hide Editor Panel" works again

    DONE bug: "Edit/Show All Panels" sometimes moved the outside window borders

    DONE (development only) "make tags" creates a project-wide TAGS file

    DONE bug: typing special characters does not crash FreeFem++-cs anymore

  • although I have not solved this bug explicitely, I am not able to reproduce it now. It looks like this bug disappeared when the program was changed for a different reason. If anyone is able to crash FreeFem++-cs again by typing random characters into the editor window, please let me know how to do it myself.
  • DONE Ubuntu bug: load "metis" now works again

    Version 11.4 (10/2/11)

    New button for setting default plot parameters

    DONE bug: MPI is now activated by default on MacOS

    DONE bug: Remote Server option with ssh works again on MacOS

    The remote ffs loader needs the Carbon API on Mac

  • The Carbon API is required to call _NSGetExecutablePath(). Unfortunately, this causes a useless warning message: ": kCGErrorFailure: Set a breakpoint @ CGErrorBreakpoint() to catch errors as they are logged. _RegisterApplication(), FAILED TO establish the default connection to the WindowServer, _CGSDefaultConnection() is NULL."
  • DONE bug: on MacOS, the ffs loader now exits properly when the server ends

    New light indicator showing when the mouse is over a picture

  • All new pictures are paused while this light is on, to let the user interact with the current picture before it is replaced with the next one.
  • Version 11.3 (3/2/11)

    Much improved stability thanks to the new regression tests suite

    based on FreeFem++ 3.12

    It is not possible to use a different version of FLTK or VTK anymore

  • because there are too many differences between versions and too many different locations on different systems. The configuration script has become too complex to keep all the possible options in a stable state.
  • The default for plotting 3D meshes has changed

  • Plotting a 3D mesh with 'plot(mesh)' now displays only its 2D contour by default for clarity. The 3D body of the mesh is quite often too complex to be of use, but click on "Graphical Options/Show Meshes" to see it nevertheless.
  • Cut planes now cut through 3D meshes and 3D surfaces as well

    DONE bug: borders were sometimes colored in white (ie invisible)

  • this happened when the number of colors on the color bar was lower thant the highest label number. When this happens now, high label numbers are all colored with the top color instead of white.
  • Cut planes through 3D data are now optional (and on by default)

    The graphical options window now always points at the current graphical tab

    Cut planes also cut through isovalues and arrows

    DONE bug: Graphical options now work properly on detached windows

    DONE bug: 2d surfaces are now located in the middle of 3D scenes

  • previously, such data could happen to be far from the center of the scene if it contained values far from zero, because the initial position of some other objects (eg the cutplane) would still be zero.
  • DONE bug: SuperLu now compiles on Win32

  • SuperLu.edp needs the BLAS, and requests them through the WHERE mechanism (with header comments in examples++-load/SuperLu.cpp). But when the BLAS were downloaded (which is the case on Windows), the WHERE-library file was not updated.
  • DONE bug: unnecessary warning messages during regression tests (Windows)

  • ffs generated segfaults on purpose (to check for the stability of ffc) but Windows produced message boxes every time. So the segfaults have been replaced by regular exits with code 0.
  • Configuration options for Slackware 13.1 32 bits and 13.0 64 bits

  • Contributed by fred 01/10/10 13:28
  • ./configure --prefix=/opt/ffcs --localstatedir=/var --sysconfdir=/etc --libdir=/usr/lib64 --with-freefem-options="--enable-download --enable-optim --enable-opengl --with-flib=/usr/lib64/libgfortran.so.3.0.0 --with-mpi=yes"
  • added to the INSTALL notes
  • DONE bug: clipping problem when resetting a 3D view

  • 3D objects very close to the camera or very far from it used to vanish when clicking on the "Reset view" button because the camera clipping plane was not properly reset. This is fixed.
  • DONE bug: FreeFem++ reference card links have been updated

    Domain boundaries can now be displayed as meshes or as transparent surfaces

    DONE bug: graphical splash screen stayed on top on 64-bit Unbuntu 10.10

  • This version of FreeFem++-cs has been checked on the same system (but maybe not the same video card). It runs fine.
  • Displays several different graphical windows

  • FreeFem++-cs is now able to redirect FreeFem++ graphical output to detached graphical windows, as soon as FreeFem++ adds a corresponding parameter (eg the index of the window to use for display) to plot().
  • More options added to visualize complex scalars and vectors

  • real part, imaginary part, modulus and phase are available for display.
  • DONE bug: auto-reload option (for files modified externally) works again

    Some of FreeFem++ graphical window keybindings have been reimplemented in FreeFem++-cs

  • the ones that have not been reimplemented do not correspond well the the FreeFem++-cs environment. They will be added on request.
  • DONE bug: new pictures sometimes appear in separate windows

  • There is still no definitive explanation to this. Probably VTK keeps some resources locked between timer calls, which prevents these resources from being used if another timer call happens in between the first two. And VTK is able to create new "RenderWindows" when existing ones are not usable.
  • The workaround has been to introduce an automatic pause: no new window will appear while the user interacts with an existing one. The good side of this is that FreeFem++-cs does not hide a picture that is being worked on anymore.
  • Version 11.2 (18/1/11)

    based on FreeFem++ 3.11

  • Upgrading FreeFem++ has become necessary because tetgen examples from the FreeFem++ documentation do not work without it. Backporting the tetgen DLL alone is not enough because of incompatibilities between C++ classes in FreeFem++ 3.10 and 3.11.
  • Version 11.1 (17/1/11)

    MPI programming on Linux

  • Activated by clicking on "Options/Parallel Server".
  • The number of processes can be changed in "Options/Other Options.../Number of MPI Processes".
  • Only the output from process 0 is displayed in the FreeFem++-cs output window. The output from other processes can be seen in the terminal window if FreeFem++-cs was started from the command line.
  • DONE bug: the FreeFem++-cs icon now appears correctly under Gnome

    DONE bug: MinGW FreeFem++ BLACS compilation now works

  • symbolic links in ff/local/download/blacs/BLACS/SRC/MPI/INTERNAL have been replaced with real files
  • DONE bug: the FreeFem++ load examples are now shipped in the FreeFem++-cs archive

  • because of a typing error, the load examples were missing from previous versions
  • Regression tests now run most of the existing FreeFem++ examples

  • All the numerical results of these examples are checked to be identical to the values computed by FreeFem++ itself.
  • It also means that if the value given by FreeFem++ is wrong, the value given by FreeFem++-cs is wrong as well.
  • FreeFem++ is now automatically connected to FreeFem++-vtk in the FreeFem++-cs package

    Improved display rate for quick sequences of images

    DONE bug: re-added a missing tetgen DLL in the Windows version

    DONE bug: re-activated the stop button on Windows

  • the stop button sometimes did not respond to user requests because the socket flag SO_RCVTIMEO does not apply to the accept() call under Win32. This has been solved by adding a call to select() before accept().
  • DONE bug: the file dialog window sometimes opened a wrong default directory

    Version 10.18 (13/12/10)

    DONE bug: solved a random crash at start-up under MacOS 10.6

    Version 10.17 (3/12/10)

    Some labels in "Options/Other Options..." have been shortened

    All FreeFem++-cs web comments are now included in the FreeFem++-cs RSS feed

  • at http://www.ljll.math.upmc.fr/lehyaric/ffcs/newsrss.xml
  • DONE bug: the "install_as_root" script failed

  • it did not create the necessary scripts when the default shell was sh and not bash (because of a difference in the way functions are defined between the two shells). This is now solved.
  • DONE bug: potential security risk on Ubuntu 10.4

  • Gnome considered the FreeFem++-cs desktop icon as a potential security risk and refused to run it.
  • The corresponding ".desktop" file has been made "executable". This removes the security message and renders the icon usable.
  • DONE bug: yet another change in the FreeFem++-cs loader

  • Sometimes command-line arguments (e.g. the first EDP file to open) were not taken into account yet. This is solved. Sorry for the repeated problem.
  • Ported to 64-bit Windows

    DONE bug: VTK messages in a separate text window now removed

    Version 10.16 (19/10/10)

    DONE bug: "Failed to initialize drm buffer pools - possibly out of graphics memory"

  • This happened for scripts that generated many complex VTK pictures, even when the pictures history was limited, as if previous VTK pictures did not free the graphics memory completely.
  • VTK data structures have been more thoroughly cleaned up. Tests show that small machines (even small laptops) can now handle complex graphics.
  • DONE bug: FreeFem++-cs could not load some files via drag'n'drop

  • This was because the corresponding file names were not converted to the right set of characters. All occurrences of '^M' are now deleted. And all occurrences of '%hex' are replaced with the corresponding character.
  • DONE bug: default VTK arrows are too big for most cases

  • Quite often, VTK arrows are too big both visually and in terms of computing resources. 'hedgehog' (the VTK name for arrows made of a thin line and without tip) is now the default arrow option.
  • DONE Automatic compiling of CMake and VTK under Cygwin

  • Up until now, compiling FreeFem++-cs under Cygwin required VTK to be preinstalled. Now the compilation process is able to compile VTK as well.
  • The Cygwin version of CMake (and any version of CMake compiled under Cygwin) always defines the keyword "UNIX" (right from the beginning in "bootstrap"), which forces VTK to use Unix-style process control. To avoid this, the FreeFem++-cs compilation process downloads a precompiled version of CMake from the official CMake website.
  • Version 10.15 (27/9/10)

    DONE Help/Documentation/Tutorial now points to the right web link

    DONE Function parameters are properly colorized

    DONE File sizes and md5sums are now listed on the download page

    DONE Commands summary page does not display properly

  • all dots that appeared as question marks when this web page was displayed with the FLTK internal browser on MacOS have been removed
  • Improving stability thanks to a new testing procedure in src/regression.cpp

  • This procedure is activated at development time with the option "-runregtests" from the command line
  • There are now two testing procedures: tests/regress checks that FreeFem++-cs gives the same numerical answers as FreeFem++ on an extensive list of examples. src/regression.cpp checks the behaviour of the graphical interface of FreeFem++-cs.
  • "src/regression.cpp" is still very basic at the moment: it will continue to grow as more FreeFem++-cs features are checked.
  • The default FreeFem++-cs behaviour is now to avoid doing any pauses.

  • This can be changed by unchecking the "Compute/Dot Not Pause" menu item. The reason for the change is that new users may not know that the pause button needs to be clicked for the computation to resume after a wait. And FreeFem++ scripts very often containt "wait=1" in plots because FreeFem++ graphics cannot be called back through tabs (although they can be called back with keystrokes).
  • This change only will affect new installations of FreeFem++-cs. Existing installations already have a preferences file (".ffcsrc") which overrides any default options.
  • DONE bug: the client got stuck waiting for the server

  • This was due to an unprotected access to FLTK data structures (in the call to serverended()) from two different threads.
  • DONE bug: .edp file names specified the command line

  • They were not loaded into the editor. This is now checked as part of the regression tests in the tests subdirectory.
  • Removed Options/Remote port tunneling

  • because it is still unclear how best to use it. The remote server functionality is still in the development stage. The main problem still to solve is how to use ssh under Windows.
  • DONE bug: <2010-09-02 jeu.> unnecessary warning "zero-sized chunk"

  • This is a perfectly valid internal message between the FreeFem++-cs client and server and it is not displayed as a warning to the user anymore.
  • VTK picture labels can now be longer (up to 30-ish caracters)

    Stopping a remote server process

  • At the moment, the client does not use "kill". It only closes the socket to force the server to produce an error. If this is not enough, a new option will be added to FreeFem++-cs to describe how to run "kill" (or an equivalent) remotely (e.g. through ssh).
  • Version 10.14 (6/9/10)

    Based on FreeFem++ 3.9-2

    Real-time update of all graphical options widgets

    Added default extension (e.g. .ps or .png) to saved images

    Editor: added "select all" (Control+A) and "undo" (Control+Z)

    Pressing ESC does not close the application anymore (use Ctrl+Q instead)

    DONE bug: linking error messages back to the corresponding source line

  • this did not work for periodic conditions errors
  • solved by adding "current line = NN" to the list of scanned error messages
  • DONE Precompiled 32-bit FreeFem++-cs tested on:

  • Linux Ubuntu 10.10 beta
  • Linux Slackware 13.0
  • Version 10.13 (2/6/10)

    DONE FreeFem++ scripts stopped without a proper explanation if verbosity was set

  • the visible error is "ffc: error: readpipe(vector &v,const int mult): vector size = -708581302"
  • this was due to STDOUT et FreeFem++-vtk streams mixing up. The solution was to create a separate buffer to split the CMD_FFG flux from CMD_STDOUT
  • DONE problem with ffct and regression tests

  • only the VTK-enabled socket code knows how much graphical data is read from the pipe. But now that the CMD_FFG flux is separated from the other fluxes, it can be safely discarded in ffct.
  • DONE The file name is now displayed when there is an error reading the file

    Added a "save" button on the editor window

    DONE Choosing the values of isosurfaces

  • when specific values for isosurfaces are provided in the .edp file, they are used only when the number of isovalues is not changed manually through the GUI.
  • During installation, create a link to ffs in the default bin directory

  • this is useful to have a remote server path without spaces
  • DONE removed message "EXEC of the plot : ffglut"

  • this message is not relevant when FreeFem++-cs is used because it replaces ffglut with its own graphical primitives
  • DONE iceweasel did not work as an html viewer for FreeFem++-cs under Linux?

  • the error message is "/usr/lib/iceweasel/firefox-bin: symbol lookup error: /usr/lib/libXi.so.6: undefined symbol: XESetWireToEventCookie"
  • this bug has not been corrected, but somehow it does not show up any more. Please let me know if it appears again in the future.
  • Version 10.12 (31/5/10)

    DONE "Recent files" menu items are shortened if filenames are too long

    DONE "Options/Reset Panel Sizes" sometimes did not show the output window

  • There is some kind of heuristics involved to grab the Fl_Tile intersection properly. Implemented in ffcs.cpp/requestresetpanelsizes()
  • The menu option is now called "Edit/Show All Panels"
  • DONE crashed without warning during a script run

  • this was due to an uncaught standard C++ library exception provoked by a bad integer value sent from the server.
  • Replaced vtkCubeAxesActor with vtkCubeAxesActor2D

  • because vtkCubeAxesActor2D is always visible by design (it draws both white and black shadows).
  • Version 10.11 (26/5/10)

    DONE "Received unknown command from server: "v"

  • removed useless message
  • DONE Removed Tex ligatures from the HTML version of the FreeFem++ documentation

  • Tex translated character sequences "ff", "fl", "fi", ... into ligatures which were unreadable in HTML.
  • New option to show axes in 2D and 3D plots

    DONE The output window sometimes disappeared without trace

  • this happened when the window size became null. Now that this is fixed, the output window still goes out of sight, but its size is never smaller than 1. It can always be widened again with the mouse.
  • DONE The stop button did not stop a paused FreeFem++ program

  • now the stop button also kills any current pause.
  • DONE crashed when closing a detached graphical panel

  • (problem with a NULL pointer in quitdetached())
  • DONE crashed when resetting the view in a detached graphical panel

  • ("Plot::firstcam" pointer was NULL)
  • Version 10.10 (18/5/10)

    DONE in plot(), the option "viso=" now sets min and max color values.

    Version 10.9 (17/5/10)

    Ability to run a remote server through an ssh tunnel

    Version 10.8 (12/5/10)

    Ability to run a FreeFem++ server on a remote machine

    External commands from the Options menu can have parameters like "%1"

    Vector scaling widget is now logarithmic

    DONE mingwm10.dll problem solved in Windows package

    Version 10.7 (10/5/10)

    Added file index before .vtk or .vtu extension when saving file

    FreeFem++-cs now displays all internal errors

  • For a better understanding of what causes the errors, even in the distributed version.
  • DONE the server process sometimes crashed without any warning

  • This was due to a mismatch in data types between the server and the client. A C++ template has been modified to detect all such type mismatches at compile time.
  • DONE 2D pictures not always centered by default

  • An error in camera coordinates has been corrected
  • DONE plot(mesh) did not show the mesh unless "show meshes" was on

  • Now plot(mesh) always displays the mesh, even if the "show meshes" option is off.
  • DONE EDP files saved without the ".edp" extension

    DONE In the list of recent files, names with relative paths are not reliable

  • The default path changes everytime a new program is opened or run, and names with relative paths cannot be opened any more.
  • The recent files list now only contains files with a full path (and an explicit drive letter under Windows)
  • DONE saving a file when exiting FreeFem++-cs does not work

    DONE Clicking on a edp-file icon did not work on Mac

  • This is a tricky one. Clicking on an icon on Mac does not produce a usable command line. It only produces an event that needs to be caught with Carbon or Cocoa. But FreeFem++-cs does not use Carbon or Cocoa because of other reasons (related to FLTK and VTK).
  • The chosen solution is to link the "loader" process to Carbon and to spawn X11 FreeFem++-cs processes when an event arrives.
  • Version 10.6 (1/3/10)

    DONE bug: MacOS application icon not working

  • pack/info.plist.m4 corrected
  • Start tcp port search from 10000

  • to avoid collision with some well-known user-space services that use fixed ports in the range 1024-9999 (e.g. VNC)
  • Install script checked

    Found generic compilation platform for Linux

  • One should compile with the oldest possible glibc, to avoid problems with version checking (a compiled program checks for a minimum version of the glibc before starting)
  • Try and compile with a Debian Stable (and avoid any automatic upgrade because at the time of the upgrade the version is current)?
  • Debian Stable (Lenny) is too recent for CentOS 5.4 (GLIBCXX version problem)
  • Ubuntu LTS 8.04 is too recent for CentOS 5.4 (GLIBCXX version problem)
  • Compilation platform chosen to be CentOS 5.4
  • Version 10.5 (28/2/10)

    DONE MacOS problem with VTK+FLTK

  • vtkFlRWI example does not work
  • VTK takes the whole FTLK main window space instead of its own vtkFlRenderWindowInteractor subwindow
  • tested MacPorts vtk5/-x11: vtkFlRWI problem
  • tested MacPorts vtk5/+x11: Mesa library problem
  • tested MacPorts vtk-devel/default: vtkFlRWI problem
  • tested MacPorts vtk-devel/+x11:Mesa library problem
  • tested MacPorts vtk-devel/+cocoa: vtkFlRWI problem
  • tested MacPorts with build_arch=i386 (instead of x86_64): MacPorts g++ does not compile (!)
  • probable explanation for the vtkFlRWI problem: RenderWindow::SetWindowId(NSWindow) and RenderWindow::SetDisplayId(NSView) need to be called early on, but the NSView object created by FLTK is not valid for VTK, which uses its own subclass of NSView. FLTK and VTK have different sub-classes for NSView (FLView and vtkCocoaGLView), so I don't know what to pass to SetDisplayId(). A set of potentially useful patches is in the fltk dir.
  • MacPorts vtk/+x11 needs MacPorts Mesa. Mesa crashes in two different ways: "xp_import_surface: assertion failed: s == ((void *)0) error: xp_import_surface returned: 2" and "_gll_noop symbol not found" in libGL.so. Probably a Mesa bug.
  • workaround: compile VTK for MacOS X11 with MacOS OpenGL headers, _without_ MacPorts Mesa.
  • DONE bug: MacOS application icon not working

  • bug corrected in pack/info.plist.m4
  • Checked all VTK keys in the Howto

    Version 10.4 (25/2/10)

    Checked default options for "configure"

  • the basic compilation process is configure/make/make install. It should work by default without any other option.
  • DONE bug: icons are bigger than buttons on MacOS

    New tag for all potentially distinct binary builds

  • add "--with-ostag" option to ./configure
  • DONE bug: FreeFem++-cs does not run and does not compile on CentOS 5.4

  • problem with versions of system libraries (glibc and stdc++)
  • ax_mpi.m4 not found
  • autoreconf problem: AC_COMPUTE_INT not found
  • autoreconf has to be reapplied after patching the FreeFem++ source
  • the chosen solution is to rebuild FreeFem++/configure before building the source package.
  • DONE bug: ffmedit not compiled when Glut is not present

  • configure now checks that GL/glut.h is installed
  • Version 10.3 (22/2/10)

    Based on FreeFem++ 3.8

    Visualizing 3D vectors works

    Access to an external editor (e.g. emacs) works:

  • "Edit/External Editor" calls an external editor on the current file
  • "Options/Other Options/External editor command" selects which command to use to call an external editor
  • "File/Auto Reload" automatically reloads a file that has been modified externally
  • "Run/Auto Run" automatically runs a file that has been modified externally
  • "Edit/Hide Editor Panel" moves the internal editor window out of the way (and "Edit/Show Editor Panel" or "Options/Reset Panel Sizes" recover the editor window after it was hidden)
  • DONE bug: The maximum number of VTK tabs was not properly limited

  • Error corrected in Visucontrol::limitopentabs() calling children() instead of tabs->children()
  • Version 10.2 (16/2/10)

    User input (FreeFem++ "cin" keyword) is implemented.

    Plotting curves and borders is implemented.

    Compiled on Linux and Windows.

    DONE bug: FreeFem++ PATH environment poses problems to FreeFem++-cs under Windows

  • OK (checked with FreeFem++ 3.7-1)
  • DONE bug: plot(...,wait=1) sometimes stops too early (before plot is shown)

  • OK now works even with big meshes that take time to plot
  • DONE bug: FreeFem++-vtk VTK windows not displayed properly

  • The same windows work in FreeFem++-cs!
  • OK after replacing Fl::awake() with Fl::add_timeout(0,...) in ffg.cpp
  • make sourcecheck OK.

    DONE bug: How to run FreeFem++-vtk with FreeFem++ from a path containing spaces?

  • One can use an escape character ("^" under Windows)
  • Version 10.1 (3/2/10)

    New 3D visualization interface for FreeFem++-cs, based on VTK.

    New FreeFem++-vtk executable:

  • Same 3D visualisation interface as FreeFem++-cs. FreeFem++-vtk acts as a drop-in replacement to FreeFem++'s ffglut.
  • Compiled and tested on Linux.

    Other planned improvements

    Editor

    Automatic indents with TAB in the editor window
    Improve print command : there is one planned in FLTK?
    MacOS: add a Carbon menu option for opening new files?
    Add line numbers in editor (request from JR) : this is planned in FLTK 1.4

    Graphics

    use vtkBoxWidget to isolate a 3D region to display
    visualization from parallel programs
    Use vtkRenderLargeImage to get larger image files
    FreeFem++ seems to define a separate plot() parameter for arrow colors
  • but would these colors be used to define a separate look-up table just for arrows?
  • FreeFem++-vtk is slow for 3d scalars on big meshes
  • because it also displays internal triangles, which is not necessary.
  • Increase size of axes values and color bar values
    choose view angles for remaining 3d FreeFem++ examples

    FreeFem++ computational server

    color coding in output window
    remove --enable-generic_blas from FreeFem++ configuration
    copy files (meshes, includes, ...) alongside script on remote server

    Other

    add code coverage check in regression tests
    Keep subwindow proportions when resizing the main window
    Ability to recall previous option values in Options menu
    Add code coverage check in regression tests
    MacOS: make FreeFem++-cs X11 windows appear when clicking on the Carbon icon

    Known issues that cannot be corrected at the moment

    MacOS: the default control key should be "Cmd" instead of "Control"

  • By default, FLTK uses the "Cmd" key on MacOS. But we are not using the FLTK MacOS stuff because of an incompatibility between FLTK and VTK on Cocoa. Moreover, using the "Cmd" key causes some keybinding conflicts with the MacOS X11 app that FreeFem++-cs relies on.
  • MacOS: using the Cmd+Tab hotkey

  • Since FreeFem++-cs is built upon X11 and not Carbon, FreeFem++-cs windows are raised only when selecting the X11 icon, and not the FreeFem++-cs icon. The FreeFem++-cs icon is the visible part of the FreeFem++-cs Carbon-based loader, which does not have a window. So it does not receive kEventWindowFocusAcquired events.
  • But one can still drag'n'drop EDP files onto the FreeFem++-cs icon.
  • Testing policy

    The FreeFem++-cs testing procedure checks that FreeFem++-cs gives the same result as FreeFem++ on my development machine for all the example EDP scripts that I can run. It also means that if FreeFem++ fails to run a test, FreeFem++-cs fails in the same way.

    Unfortunately, even if the testing procedure contains many different test cases, it cannot guaranty that FreeFem++-cs will perform as expected. Even test cases that pass here may fail elsewhere because machines are never identical. So I rely on all users to keep me informed about the bugs that they find. You can use the comment form below or contact me.



    Add your comments

    Name :

    Comments :


    Please copy the following numbers (sorry, this is for computer-generated spam prevention) :





    Previous comments (newest first)
    antoine 15/11/11 17:47
    Hi xingyi,

    FFCS for Ubuntu 11.10 is now available from the download page. Could you let me know if it works for you?

    Best regards,

    Antoine.
    antoine 22/10/11 08:34
    Hi xingyi,

    New Ubuntu version very often require some updates to the FFCS source. I have just installed Ubuntu 11.10 and I am updating FFCS now. It should be working and available for download as a precompiled archive by the end of next week.

    Best regards,

    Antoine.
    xingyi 22/10/11 00:07
    I installed freeFem++ on my Ubuntu 11.10 32-bit, but when I compile it, one error message occurs"
    configure: fatal error: sizeof pointer !
    configure: error: Sorry sizeof c++ pointer are not 4 or 8"

    What shall I do? Anyone can help?

    Best,
    Xingyi
    antoine 11/03/11 22:13
    Dear Houari,

    Some FreeFem examples use software packages which are not freely available for download from the internet. mmg3d is one of them. This is why it is not included in precompiled FFCS versions.

    If you really need to use mmg3d with FreeFem, you should get in contact with the authors (see http://www.math.u-bordeaux1.fr/~dobj/logiciels/mmg3d.php) to obtain mmg3dlib.tar.gz, untar the FFCS source archive, copy mmg3dlib.tar.gz into ff/upstream/download/pkg and compile FFCS.

    Sorry for the inconvenience.

    Best regards,

    Antoine.
    houari 11/03/11 20:53
    Dear Antoine, thank you for solving the below problem:
    DONE Windows: tetgen.dll loading problem solved
    # the visible error message was a _gxx_personality_v0 problem in libstdc++-6.dll

    but no I have the the same error message but for mmg3d.dll . try for example:
    FreeFem++-cs-11.7\Contents\Resources\examples\3d\fallingspheres.edp
    you will have this error. Can you try to see if you have it please?
    Best regards
    antoine 15/02/11 16:11
    Hi Marlon,

    Thank you for the netstat output. I have removed all lines but the two that prove that ffc is properly opening a TCP communication (on ports 10000 and up) for ffs. I can see the same lines on my own machine so this part is ok.

    Could you now check if the following screenshots also match those on your machine? (no need to create screenshots yourself if the french text on the screenshots corresponds to what you see)

    The first set of screenshots deal with authorizing ffc through the firewall. The small windows are opened by clicking on the previous ones:

    http://www.ann.jussieu.fr/~lehyaric/tmp/index.php?dir=&file=ProgrammesAutorises.png

    http://www.ann.jussieu.fr/~lehyaric/tmp/index.php?dir=&file=ModifierUnProgramme.png

    http://www.ann.jussieu.fr/~lehyaric/tmp/index.php?dir=&file=TypeEmplacementReseau.png

    The second set deals with defining the current network zone:

    http://www.ann.jussieu.fr/~lehyaric/tmp/index.php?dir=&file=PareFeuActive.png

    Thank you for your patience. I hope that we will eventually find where the difference lies between your machine and mine.
    Marlon 15/02/11 15:19

    Active Connections

    Proto Local Address Foreign Address
    ...
    TCP 0.0.0.0:10000 marlonrecarte:0
    TCP 0.0.0.0:10001 marlonrecarte:0
    ...
    antoine 15/02/11 10:05
    Hi Marlon,

    Thank you for this test. This is more and more puzzling as my own machine also installs all the Windows updates. Could you open a command prompt and run the command "netstat -a" when the error message appears? (this is to make sure that ffc has indeed opened a TCP port when ffs complains that it cannot connect to it) In the meantime, I will try and play with all the Windows options to try and reproduce the same behaviour over here.
    Marlon 14/02/11 19:20
    I uninstall the antivirus and the error still appears , the problem began with a windows update.
    Marlon 14/02/11 16:19
    http://img407.imageshack.us/i/firewall.png/
    antoine 14/02/11 16:17
    Hi Marlon,

    Thank you for the screenshot and sorry for the trouble in submitting comments (I am working on the website concurrently). I got your comment anyway. Your firewall seems fine as you have the "ffc" entries. Do you have another security software installed (maybe an antivirus) that could stop FFCS from using TCP?
    Marlon 14/02/11 16:11
    http://img407.imageshack.us/i/firewall.png/
    antoine 14/02/11 10:17
    Hi Marlon,

    Getting back to you after checking my own Win7 machine. Could you check if your Win7 machine's firewall has specific rules for letting ffc trafic through? Here is how these rules look like on my machine:

    http://www.ljll.math.upmc.fr/lehyaric/tmp/parefeu.png

    Thank you,

    Antoine.
    antoine 11/02/11 23:06
    Thank you for the listing. ServerType="sequential" and RemoteServerClientName="localhost" so the client and server are doing the right thing and trying to communicate on localhost. But Windows is rejecting these communications with "Connection refused". So the answer is probably somewhere in the Windows networking configuration (tcp options or firewall maybe). Unfortunately, I won't have access to Windows until monday. I'll get back to you as soon as I have been able to check that machine.
    Marlon 11/02/11 22:25
    # -*-mode:conf;-*-
    # Configuration options for FreeFem++-cs and FFG
    PathToDoc="C:\Program Files (x86)\FreeFem++-cs-10.17\Contents\Resources\doc";
    PathToExe="C:\Program Files (x86)\FreeFem++-cs-10.17\Contents\Windows";
    PathToExamples="C:\Program Files (x86)\FreeFem++-cs-10.17\Contents\Windows";
    Browser="";
    MainWindowX=0;
    MainWindowY=14;
    MainWindowHeight=706;
    MainWindowWidth=1362;
    FFGWindowX=0;
    FFGWindowY=0;
    FFGWindowHeight=600;
    FFGWindowWidth=600;
    OutputAreaHeight=111;
    EditorAreaWidth=759;
    OutputSizeMax=1000000;
    PicturesHistoryMax=50;
    RecentFilesMax=25;
    ...
    PrintCommand="";
    ExternalEditor="notepad.exe %1";
    SaveOptionsAtExit=0;
    NumberOfProcesses_v2=1;
    MPIRunCommand_v2="c:\\Program\ Files\\MPICH2\\bin\\mpiexec.exe -n %2 %1";
    ServerType="sequential";
    RemoteServerCommand="ssh localhost cmd.exe /c C:\\\\Program Files\\\\FreeFem++-cs-11.1\\\\Contents\\\\Windows\\\\..\\\\..\\\\ffs";
    RemoteServerDirectory="";
    RemoteServerClientName="localhost";
    AutoSave=0;
    AutoSaveSeconds=300;
    AutoSaveFile="autosave.edp";
    AutoReload=0;
    AutoRun=0;
    ShowHowto=0;
    TemporaryDirectory="";
    EnableHighlighting=1;
    NoWait=0;
    antoine 11/02/11 21:38
    Thank you for the tests. This line is quite interesting: "Fri, 11 Feb 2011 20:23:09: ffs: error: Socket::Connect: Connection refused". I looked up the meaning of this system error: "WINSOCK Error: Connection refused. No connection could be made because the target computer actively refused it. This usually results from trying to connect to a service that is inactive on the foreign host. You might have gone to the wrong host, or the server application you are trying to contact might not be running" (found at http://technet.microsoft.com/en-us/library/cc975854.aspx).

    So we need to find why ffs (the FreeFem++ server) is not able to talk to ffc (the FreeFem++ client) on the same host (the local machine). It could be a firewall question (has the firewall been modified since last week?), or a configuration question (has the FFCS configuration been changed to use a remote server for instance?). Could you send me the contents of c:\Users\<your username>\ffcs.ini?
    Marlon 11/02/11 21:14
    i tried with the new version but it's the same

    \ffcs.log

    Fri, 11 Feb 2011 20:22:41: ffc: forgetting option 'PathToDoc' (sp=0)
    Fri, 11 Feb 2011 20:22:41: ffc: forgetting option 'PathToExe' (sp=0)
    Fri, 11 Feb 2011 20:22:41: ffc: forgetting option 'PathToExamples' (sp=0)
    Fri, 11 Feb 2011 20:22:41: ffc: forgetting option 'ServerType' (sp=0)
    Fri, 11 Feb 2011 20:23:08: ffs: forgetting option 'PathToDoc'
    Fri, 11 Feb 2011 20:23:08: ffs: forgetting option 'PathToExe'
    Fri, 11 Feb 2011 20:23:08: ffs: forgetting option 'PathToExamples'
    Fri, 11 Feb 2011 20:23:08: ffs: forgetting option 'ServerType'
    Fri, 11 Feb 2011 20:23:09: ffs: error: Socket::Connect: Connection refused
    antoine 11/02/11 21:10
    Thank you. There are a few things that we can try to understand this bug:

    - There is a new version compiled on Win7-64 available since yesterday. Would you like to try this new version first and see if it works better?

    - If the FFCS window crashes when the error message appears, I would also be interested in the contents of the file c:\Users\<your username>\ffcs.log

    - Since the error happens when a computation is running, would it be possible for you to send an example script where this error shows? I would like to try it on my own Win7 machine.

    Thank you in advance.
    anonymous 11/02/11 20:51
    i am using Win7 64, the error appears when I compile; the last weak it was working fine
    antoine 11/02/11 19:51
    @marlon:

    Thank you for your report. Which version of Windows are you using (XP, Vista, 7, 32, 64)? When does this message appear?
    Marlon 11/02/11 19:44
    There is a microsoft visual c++ runtime library error

    "this application has requested the Runtime to terminate it in a inusual way"
    antoine 19/11/10 20:41
    @Bob:

    There is indeed a problem with the examples directory under Windows. I will get back to you as soon as I have a fix (probably sometime next week).
    Bob Jackson 19/11/10 20:20
    I can't find the files generated by the examples, for example in chap3 membrane.edp the graph.txt and the Th.msh file don't seem to get written to disk anywhere. I tried giving the filename a specific path, "c:\graph.txt" , but it didn't make any difference.
    antoine 06/09/10 14:26
    @Fred:

    Ok, I have installed a Slackware system. The new version of FFCS (10.14 for 32-bit Linux) seems to run straight out of the box (no need to recompile). Could you try this new version and let me know if it works for you? If it does not, then I will check if FFCS recompiles ok on this system.
    anonymous 23/08/10 16:26
    Antoine,

    Thanks for the reply. I'm still looking at it from time to time and will try it on the latest version of Slackware (13.1) soon. I have been busy with other things but check back periodically. I'll let you know if I make any progress and what the difficulty was/is if I find anything new.

    Thanks,

    Fred
    antoine 16/08/10 11:59
    @Fred:

    Sorry for getting back to you so late. It looks like there is some library problem for FF on Slackware 13, but I can't determine which library is at fault without testing it on a Slackware system, and I don't have one at the moment. I will get back to you as soon as I can (probably in a matter of weeks).
    anonymous 03/08/10 16:08
    Hi,

    When trying to compile on Slackware 13.0 Linux (32bit), the following error occurs during the freefem++ build:

    Making all in x11
    make[6]: Entering directory `/home/fred/src/ffcs-10.13/ff/local/src/x11'
    g++ -DNDEBUG -O3 -march=athlon -DDRAWING -DBAMG_LONG_LONG -DNCHECKPTR -rdynamic -fPIC -o FreeFem++-x11 Xrgraph.o ffcsstream.o ../lglib/liblg.a ../fflib/libff.a -lumfpack -lamd -L/home/fred/src/ffcs-10.13/ff/local/download/lib -larpack -lblas -lXxf86vm -lXext -lX11 -lXpm -lm -ldl -ff2c
    /home/fred/src/ffcs-10.13/ff/local/download/lib/libarpack.a(dnaupd.o): In function `dnaupd_':
    dnaupd.f:(.text+0x760): undefined reference to `dlamch_'
    /home/fred/src/ffcs-10.13/ff/local/download/lib/libarpack.a(dsaupd.o): In function `dsaupd_':
    dsaupd.f:(.text+0x858): undefined reference to `dlamch_'
    /home/fred/src/ffcs-10.13/ff/local/download/lib/libarpack.a(dneupd.o): In function `dneupd_':
    dneupd.f:(.text+0x40): undefined reference to `dlamch_'
    dneupd.f:(.text+0x33c): undefined reference to `dlapy2_'
    dneupd.f:(.text+0xb4f): undefined reference to `dlaset_'
    dneupd.f:(.text+0xbe2): undefined reference to `dlahqr_'
    dneupd.f:(.text+0xd93): undefined reference to `dtrsen_'
    dneupd.f:(.text+0xe4f): undefined reference to `dgeqr2_'
    dneupd.f:(.text+0xec7): undefined reference to `dorm2r_'
    dneupd.f:(.text+0xf05): undefined reference to `dlacpy_'
    dneupd.f:(.text+0x102a): undefined reference to `dlapy2_'
    dneupd.f:(.text+0x10ba): undefined reference to `dlapy2_'

    The same thing happens when trying to compile freefem++-3.9. I have used the following configure options:
    ./configure --prefix=/opt/ffcs --enable-optim --with-freefem-options="--with-flib='-ff2c' --enable-opengl --enable-download" Really though it doesn't matter what options are used for configuring, the same error occurs anyway. It looks to me like some sort of linker problem but alas, I am really not sure.

    Thanks,

    Fred

    antoine 10/05/10 12:07
    Hi Bouke,

    Thank you very much for your feedback. I have added a note about X11 for Mac in the future release of FFCS.

    There is indeed a problem with the FFCS 10.6 icon on MacOS. Hopefully this will be solved in the next version.

    Regards,

    Antoine.
    Bouke 10/05/10 11:05
    Hi Antoine,
    Thanks for the fast reply.
    I just compiled FreeFEM++-cs from source and it was a breeze. Very impressive!

    I still can't start it by double clicking the icon, but from the terminal it works just fine. I do have to start X11 manually (in Applications/Utilities) otherwise it can not open a display but that may be caused by my OSX or X11 version. That's all ok for me but you may want to note it on your website for other (more easily daunted) OSX 10.4 users.

    Thanks for making this available. I'm going to play with it now. (I'll check for the new version & see if I can get that to work, too).

    Regards,

    Bouke
    antoine 10/05/10 10:40
    Hi Bouke,

    A new version of FreeFem++-cs for Mac will be published in the next few days. Maybe you should test that one first. The library problems will probably still be the same (since I only have access to MacOS 10.5 and 10.6), but at least you will have the most recent version to compile. All the library problems are very likely to go away when compiling from source.

    Please let me know how you get on.

    Regards,

    Antoine.
    Bouke 10/05/10 09:57
    I just downloaded the binary version 10.6 for MacOS 10.5 or earlier (Intel) but it does not run on my iMac (OSX 10.4.11). From the Finder, it just dies quietly when I double click, when I try to start ffg (or FreeFem__(-cs) in the FreemFem++-cs application folder from a terminal window it says /usr/X11/lib/libXpm.4.dylib is not loaded. I have libXpm.4.dylib on my system but it is in /usr/X11R6/lib.

    I though I could do a quick fix by making a symbolic link in /usr from X11R6 to X11 but then it says: "Incompatible library version: FreeFem++-cs requires version 16.0.0 or later, libXpm.4.dylib provides version 4.11.0".

    Is something missing from my system? Surely my X11 can't be 11 major releases behind?

    I'm going to try to compile from source, any ideas in the mean time?
    Joel 27/02/10 21:57
    Thanks Antoine,

    If none of it works, I will have our lone mac user on the old interface and trust that it is working in the same way as the rest of us on Ubuntu, which is not the worst case.

    Further complicating everything is that she is only on OSX 10.4 which it seems has very little support anymore. We were force to use fink rather than macports and even fink only has the X11 version of vtk.

    All that is a long way to say, if you fix the X11 patchset, that would be excellent. If it still doesn't work, we will find some way to provide the software for her.

    Thank you so much for this excellent piece of software, and for helping us, no matter the result!
    antoine 26/02/10 23:02
    Hi Joel,

    The MacOS case is quite complex. Carbon used to work fine, but it is not available on the new 64-bit MacOS 10.6 anymore. Cocoa will not work for FreeFem++-cs because of a problem somewhere in VTK or FLTK. So X11 is probably the way to go. But the patch problem shows that the MacOS+X11 version of FreeFem++-cs is currently broken.

    If you wait for a few days, I will have a version that compiles with X11 on MacOS.

    Regards,

    Antoine.
    Joel 26/02/10 20:26
    Well I am able to "finish" compiling, but the .app says that it is broken. When I enter it and run the FreeFem++-cs directly it opens and I see the GUI but then immediately crashes. I suspect this may be because the VTK I have isn't cocoa. Therefore I tried the --disable-cocoa switch but then during make a patch that it attempts fails. While I try to sort out some of this silliness I wonder if you know what patch I should be looking at? It says the files in question are orig/configure and local/configure.

    I will try the cocoa version again and give a proper report on the crash, maybe it will mean something.
    Joel 26/02/10 19:52
    Thanks for the feedback, I will give it a try
    antoine 26/02/10 19:15
    Hi Joel,

    First of all, sorry for the trouble with the spam-checking routine. Fortunately, I saw your message in the log so I could validate it by hand. Hopefully, the spam-checking bug is corrected as well.

    "No rule to make .../FreeFem++-mpi" is indeed, as you have rightly guessed, a packaging bug. As a workaround, just remove the line "${workarea}/Contents/${platform}/FreeFem++-mpi${EXEEXT} \" from the definition "$prog=" at line 227 of pack/Makefile. And the compilation process will complete (the MPI program is not used anyway).

    Please let me know if everything goes as planned.

    And thank you for the Ubuntu bug report.

    Regards,

    Antoine.
    Joel 26/02/10 17:16
    Now I am trying to make on the MacOSX 10.4 computer that we also have in the lab and we get:

    make[2]: *** No rule to make target `../ff/local/src/mpi/FreeFem++-mpi', needed by `FreeFem++-cs 10.3.app/Contents/MacOS/FreeFem++-mpi'. Stop.
    make[1]: *** [all-recursive] Error 1
    make: *** [all] Error 2

    it seems to be from the "pack" Makefile so I think I was almost done, but I cannot seem to get past this. The file it is looking for to cp into pack "FreeFem++-mpi" doesn't seem to exist, though I didn't get any errors that it wasn't built.

    Any thoughts?
    Joel 24/02/10 20:34
    I needed to add
    #include <stdio.h>
    to server.cpp to compile.

    Specifics:
    version 10.3
    ubuntu linux 9.10
    antoine 25/11/09 08:48
    Hi Cesare,

    Thank you very much for solving this problem. Sorry that I could not get back to you earlier. I will include your modifications in the FFCS source as soon as possible.

    Best regards,

    Antoine.
    anonymous 24/11/09 20:16
    solved (I hope..) problem with fltk

    I substitute in file filename_list.cxx the following line (line 70):
    int n = scandir(d, list, 0, (int(*)(const void*,const void*))sort);

    with this line:
    int n = scandir(d, list, 0, (int(*)( const dirent **,const dirent **))sort);

    now make works (until compile freefem, this give errors, but are known under ubuntu 9.1 by freefem mailing list)

    i hope thats is right

    Cesare
    antoine 02/11/09 17:29
    @cesare:

    This looks like a compilation problem in the FLTK library. Let me try and reproduce it on a local machine and I'll get back to you.

    Thank you for the report.

    Antoine.
    cesare 02/11/09 14:41
    Hi,
    I have the following problem:

    Under ubuntu karmic(9.1) the compilation procedure gives error, as :

    filename_list.cxx: In function ‘int fl_filename_list(const char*, dirent***, int (*)(dirent**, dirent**))’:
    filename_list.cxx:70: error: invalid conversion from ‘int (*)(const void*, const void*)’ to ‘int (*)(const dirent**, const dirent**)’
    filename_list.cxx:70: error: initializing argument 4 of ‘int scandir(const char*, dirent***, int (*)(const dirent*), int (*)(const dirent**, const dirent**))’
    make[4]: *** [filename_list.o] Errore 1
    make[4]: uscita dalla directory «/home/cesare/ffcs-9.11/fltk/local/src»
    make[3]: *** [all] Errore 1
    make[3]: uscita dalla directory «/home/cesare/ffcs-9.11/fltk/local»
    make[2]: *** [all] Errore 2
    make[2]: uscita dalla directory «/home/cesare/ffcs-9.11/fltk»
    make[1]: *** [all-recursive] Errore 1
    make[1]: uscita dalla directory «/home/cesare/ffcs-9.11»
    make: *** [all] Errore 2

    Why? with the previous version of ubuntu, Jaunty(9.04) all works fine.



    Thanks
    antoine 20/05/09 16:21
    Hi Tadeusz,

    Found the bug! It was a problem with the packaging procedure on Windows (It kept old versions of some DLLs, which explains why some old examples worked better). This is solved in version 9.11 for Windows (you will find it on the Download page). Sorry for the inconvenience.
    antoine 20/05/09 11:18
    Hi Tadeusz,

    Thank you for pointing this problem out. I have no idea what causes the new version of some examples to fail. This seems to happen only on Windows. I can reproduce this bug locally so I'll get back to you as soon as I have found an explanation for it.
    tadeusz 20/05/09 08:49
    Dear Antoine
    Many examples give me this token related errors. FreeFem++ ver 3.0-6 examples work fine. OS Windows XP pro latest SPacks. Any clue appreciated. Version with Lua editor worked OK as well.

    example Laplace3d:

    Error line number 53, in file C:\DOCUME~1\tadeusz\LOCALS~1\Temp\.ffcs18.edp, before token )
    medit: Sorry no way to save this kind of data
    current line = 53
    Compile error : medit: Sorry no way to save this kind of data
    line number :53, )
    error Compile error : medit: Sorry no way to save this kind of data
    line number :53, )
    code = 1 mpirank: 0
    Dan 03/03/09 17:49
    Dear Antoine,

    This time, it's OK, I mean the newest precompiled version (9.4) works like a charm. Thank you very much,

    Dan Calugaru,
    Research Engineer
    Laboratory of Mathematics
    Besancon, France
    antoine 19/02/09 16:49
    There was obviously a difference between your setup and mine (since I could not reproduce your bug), so I tried to apply all software updates to CentOS. And then I saw exactly the same bug as you: the window appeared, then vanished at any mouse move! I tracked the bug down to libdl.so, so now you have a new version (9.4) to test, if you want to. Sorry for all the trouble!
    Dan 19/02/09 14:59
    I'm sorry Antoine, but it doesn't yet work. Precomplied Linux versions (9.2 and 9.3) give me the same error (when the mouse is moved on the FreeFem++-cs window):

    $ Path_to_FreeFem++-cs-v9.2/FreeFem++-cs
    Path_to_FreeFem++-cs-v9.2/FreeFem++-cs: line 23: 655 Segmentation fault "$exepath/ffc" -pathtodoc "$docpath" -pathtoexe "$exepath" -pathtoexamples "$examplepath" $@

    $ Path_to_FreeFem++-cs-v9.3/FreeFem++-cs
    Path_to_FreeFem++-cs v9.3/FreeFem++-cs: line 23: 661 Segmentation fault "$exepath/ffc" -pathtodoc "$docpath" -pathtoexe "$exepath" -pathtoexamples "$examplepath" $@

    The only one difference is "655" instead of "661".

    Concerning libGL.so, below are some outputs from our machine which could be useful :

    $ rpm -qa | grep GL
    mesa-libGL-6.5.1-7.5.el5
    mesa-libGL-devel-6.5.1-7.5.el5
    mesa-libGLU-devel-6.5.1-7.5.el5
    mesa-libGLU-6.5.1-7.5.el5

    $ rpm -ql mesa-libGL-devel-6.5.1-7.5.el5
    ...
    /usr/lib/libGL.so
    ...
    antoine 19/02/09 11:47
    Hi Dan,

    I tried CentOS 5.1 and I found a bug in the precompiled Linux version of FreeFem++-cs which prevented it from working (a problem related to libGL.so). Maybe you can try the new precompiled Linux package (version 9.3) and let me know if it works for you?
    antoine 18/02/09 14:28
    Hi Dan,

    Thank you for all the details. I'll now try and reproduce the bug you are seeing in this environment and I'll get back to you.
    Dan 18/02/09 14:09
    Thanks to your replies. This is the output :

    $ g++ -v
    Using built-in specs.
    Target: i386-redhat-linux
    Configured with: ../configure --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --enable-shared --enable-threads=posix --enable-checking=release --with-system-zlib --enable-__cxa_atexit --disable-libunwind-exceptions --enable-libgcj-multifile --enable-languages=c,c++,objc,obj-c++,java,fortran,ada --enable-java-awt=gtk --disable-dssi --enable-plugin --with-java-home=/usr/lib/jvm/java-1.4.2-gcj-1.4.2.0/jre --with-cpu=generic --host=i386-redhat-linux
    Thread model: posix
    gcc version 4.1.2 20071124 (Red Hat 4.1.2-42)

    Our cluster system is Rocks 5.1 based on CentOS 5.1 (which is a rebuid of Red Hat EL 5.1).

    We have also installed the precompiled 9.2 version of ff++-cs on this cluster, but it doesn't work well : when "FreeFem++-cs" is launched, the graphical interface is opened, but it is brutally closed if the mouse cursor is moved on it (without any clicking).
    antoine 18/02/09 13:36
    Hi Dan,

    This "undefined reference to `__cxa_get_exception_ptr'" problem seems related to the version of g++. Could you give me the output of "g++ -v" on your machine (and maybe the version of your Linux distribution if you have it) so that I try and test the same version locally?
    Dan 18/02/09 13:19
    Thank you for the new package. Indeed, the previous error is corrected. Unfortunately, the compilation is not yet successfully done because, after about 20 minutes on our old 32 bit master node, it is finished with the following error :
    -----------------------------
    g++ -DHAVE_CONFIG_H -I. -I.. -DSERVER -DNDEBUG -O3 -mmmx -msse -msse2 -MT ffs-main.o -MD -MP -MF .deps/ffs-main.Tpo -c -o ffs-main.o `test -f 'main.cpp' || echo './'`main.cpp
    mv -f .deps/ffs-main.Tpo .deps/ffs-main.Po
    g++ -DSERVER -DNDEBUG -O3 -mmmx -msse -msse2 -DNDEBUG -O3 -mmmx -msse -msse2 -DDRAWING -DBAMG_LONG_LONG -DNCHECKPTR -rdynamic -o ffs ffs-main.o libffs.a ../ff/local/src/lglib/liblg.a ../ff/local/src/fflib/libff.a /share/apps/ffcs-9.2/ff/local/download/lib/libumfpack.a /share/apps/ffcs-9.2/ff/local/download/lib/libamd.a -L/share/apps/ffcs-9.2/ff/local/download/lib -larpack -L/share/apps/ffcs-9.2/ff/local/download/blas -lcblas -lf77blas -lXxf86vm -lXext -lX11 -lXpm -lm -ldl -L/usr/lib/gcc/i386-redhat-linux/3.4.6 -lg2c -lpthread -lSM -lICE -lXext -lXpm -lX11 -lm
    libffs.a(libffs_a-server.o): In function `servermain(int, char**)':
    server.cpp:(.text+0x394b): undefined reference to `__cxa_get_exception_ptr'
    server.cpp:(.text+0x39a6): undefined reference to `__cxa_get_exception_ptr'
    server.cpp:(.text+0x3ae9): undefined reference to `__cxa_get_exception_ptr'
    libffs.a(libffs_a-options.o): In function `Options::load(std::basic_string<char, std::char_traits<char>, std::allocator<char> >)':
    options.cpp:(.text+0x4229): undefined reference to `__cxa_get_exception_ptr'
    collect2: ld returned 1 exit status
    make[3]: *** [ffs] Error 1
    make[3]: Leaving directory `/share/apps/ffcs-9.2/src'
    make[2]: *** [all] Error 2
    make[2]: Leaving directory `/share/apps/ffcs-9.2/src'
    make[1]: *** [all-recursive] Error 1
    make[1]: Leaving directory `/share/apps/ffcs-9.2'
    make: *** [all] Error 2
    ------------------------------

    I would also say that we have already installed "freefem++-3.0-5" (from source). Making "./configure" step with or without the "--with-freefem=PATH_to_freefem++-3.0-5" option doesn't change anything (the same error message is obtained at the end of the compilation).
    antoine 18/02/09 11:30
    Hi Dan,

    Thank you for the bug report and the analysis. I have packaged a new source version (as version 9.2) with the missing files (including fltk/orig/configure). It should now compile without problems. Please let me know if it works for you.
    Dan 18/02/09 10:44
    Other potential problem for the \"./configure\" step : the option --with-fltk=\"path\" seems to be ignored, because even with this option, when we pass to the \"make\" step, it try to \"cd fltk ... cd local\" and to configure the local version.
    Dan 17/02/09 19:10
    Hello,

    When I try to install FreeFem++-cs-9.1 on a standard linux machine, after the "./configure" step succeeded, I obtain for the "make" step the following error :

    $ make
    make all-recursive
    make[1]: Entering directory `/share/apps/ffcs-9.1'
    Making all in fltk
    make[2]: Entering directory `/share/apps/ffcs-9.1/fltk'
    cd local && env CXX=g++ CC=gcc \\
    ./configure --enable-threads --enable-localjpeg --enable-localpng --enable-localzlib --enable-gl --disable-cygwin
    env: ./configure: No such file or directory
    make[2]: *** [local/config.status] Error 127
    make[2]: Leaving directory `/share/apps/ffcs-9.1/fltk'
    make[1]: *** [all-recursive] Error 1
    make[1]: Leaving directory `/share/apps/ffcs-9.1'
    make: *** [all] Error 2

    There is no "configure" file script in "fltk/local" repertory. What could I do to solve this issue ?
    antoine 03/02/09 10:27
    There is indeed a problem with The MPI option in FreeFem++-cs. Thank you Albert for the bug report. I will make it work again as soon as I can but it looks like quite a difficult problem to solve. In the meantime, as a workaround, you can use FreeFem++-mpi (found in the FreeFem++ archive) to run parallel programs.
    Albert 03/02/09 03:37
    I have a problem with MPI test: after I activate "Use MPI" in the Options menu and run "essai.edp", the following message comes out:

    Internal error: _spawnv() failed: Invalid argument
    antoine 16/01/09 09:31
    FreeFem++ dlls are now included in FreeFem++-cs version 9.1 for Windows. Please go to the Download page to install it.
    antoine 19/12/08 14:14
    Yes, at the moment dlls are not included in the Windows distribution because of a compilation option problem. But if all goes well there will be a new distribution with dlls within a couple of months.
    paul 19/12/08 08:42
    As far as I understand, the .dll files corresponding to the several "load" possibilities (e.g. "popen") are not included in the windows distribution.
    antoine 06/12/08 09:25
    The "autosave" bug is fixed in version 8.12.
    antoine 29/11/08 08:47
    There is a problem with the "autosave" feature : when a modified file is automatically saved, its name is permanently changed to "autosave.edp". This will be corrected very shortly. In the meanwhile, a workaround is to deactivate the Auto-Save option in the Options menu.