Streamline build process

Developers
2009-10-30
2012-09-14
  • Stefan Schroeder

    Hello Gxsm-developers, howdy PZ,

    after several years of abstinence I was trying to compile Gxsm2, and I am
    slightly disappointed with the results. The build process is still rough
    around the edges. Unclear error messages and we're still using this autogen
    script, which was a hack from the start. I get a segfault when starting gxsm2
    and tons of 'deprecated' warnings. Do you have them, too?

    They look like this:

    warning: deprecated conversion from string constant to 'char*'

    I will browse my old code and do some polishing. Or is there sth else to do
    that so. without actual hardware could tackle.

    yours

    Stefan

     
  • Percy Zahl

    Percy Zahl - 2009-10-31

    Hey Stefan!

    nice to get you back, how are you?

    not sure what system you are using, but stable Debian and latest Ubuntu works
    trouble free -- all default debs only.

    Yes, there are these dumb gchar / const gchar warnings with latest compilers
    -- they are hard to clean up (and non of them is a problem) -- as some
    dependencies are circular and biting in your own tail or going down to libs --
    easy fixes of those I clean up already, i.e. added "const" as/where
    possible :-(...

    In particular (static and in general equivalent to const)) array definitions
    are trouble some.... just find out your self -- any help is very welcome :-)

    PS: I am just running several Linuxes/Debain/Ubuntu.. in VirtualBox OSE (Sun)
    for testing purposes -- including USB support, just figuring it out! This is
    pretty cool and quiet impressive fast on a multi CPU machine -- like my quad
    Phenom X4 (3Ghz/8G/3T RAID5)!

     
  • Stefan Schroeder

    I am fine. Work as a programmer for Siemens-Braunschweig together with Winnie
    and Bierkandt and Fühner and Hullmann.

    My machine is

    Linux gloriaxfce 2.6.28-15-generic #52-Ubuntu SMP GNU/Linux

    Well, it's a step forward that we can use the netcdf-packages from the
    repository. I remember the times when you had to install the manually...

    Anyway, those warnings should be easy to fix. What about the autogen.sh? Why
    can't we use 'configure' like other packages? I will have a peek. Some time
    ago you patched the python-interface (sth with the defines IIRC).

    I am really happy that Gxsm has a vibrant community, what about AK?

     
  • Percy Zahl

    Percy Zahl - 2009-11-01

    Always need autogen.sh if "bootstrap code" from CVS, not if from
    tarball. Since long now, no need for any "non packaged" add ons.

     
  • Percy Zahl

    Percy Zahl - 2009-11-06

    I started fixing a whole lot of those stupid const gchar / gchar warnig issues
    -- so far almost don in Gxsm-2.0/src (core). -- some deep part recursive
    dependencies to resolve...

     
  • Percy Zahl

    Percy Zahl - 2009-11-07

    Fixed all const gchar * issues in core code (/src), plugins pending.

    Does any one has a clue what these two wicked warnings are all about????

    xsmtypes.h:203: warning: "typedef" was ignored in this declaration

    .... many more of the same

    code like "typedef struct abc { int x,y; };" seams to be not any longer fine
    :-(

    and this type of warnings look also strange and new to be ?!?!?:

    gapp_service.h:209: warning: format not a string literal and no format
    arguments

     
  • Stefan Schroeder

    I started cleaning up the plugins, beginning with common/, but then I
    hesitated, because I wasn't sure whether we should check in advance, if all
    plugins are really required. Some of them might be severly obsolete. Can we
    run a survey on this via the forums. Sth. like: "We want to throw away plugin
    XY, you (dear user) have 6 month to reply if you want to have this plugin
    kept." This would allow us to concentrate on the real functionality without
    cleaning up dead code.

    The typedef was an easy one. Simply remove "typedef."

    I guess the reason being that no alias is wanted usually

    Eg.

    typedef enum AUTO_SAVE_MODE { MANUAL_SAVE_AS, AUTO_NAME_SAVE,
    AUTO_NAME_PARTIAL...};

    does not have an actual name for the typedef'd alias,

    that'll be:

    typedef enum AUTO_SAVE_MODE { MANUAL_SAVE_AS, AUTO_NAME_SAVE,
    AUTO_NAME_PARTIAL...} AUTO_SAVE_MODE;

    Then the 2nd one:

    When you find this error, you can replace the string

    mld

    with

    "%s", mld

    Beats me why.

    Next week I won't be able to do any commits b/o a business trip, but I guess
    the changes are übersichtlich anyway.

     
  • Percy Zahl

    Percy Zahl - 2009-11-07

    OK, just do one by one plugin and commit ASAP and done with it to prevent
    conflicts. I usually do csv commits in short intervals for ever major and
    successful edit, etc.

    Let me do the active Hwi plugins, i.e. hard/sranger*

    Let me know if you want to go over all the type def issues....

     
  • Percy Zahl

    Percy Zahl - 2009-11-08

    Alright, did a major code cleanup with respect to these warnings. Almost done.
    Just a few plugins pending.

     
  • Stefan Schroeder

    yομ αяε яεαℓℓy α ωizzαяδ

    :-)

     
  • Percy Zahl

    Percy Zahl - 2009-11-10

    :-)

    there are a few plugins left to treat -- one all the same issue is the plugin
    def structure needs a non const for description field, just put the const
    string into a gstrdup() -- as i did for some already.

     
  • Percy Zahl

    Percy Zahl - 2009-11-13

    cleaned up a bunch of plugins...

     
  • Percy Zahl

    Percy Zahl - 2009-11-14

    two wicked disturbing gtk-critical messages -- seam not to cause any trouble,
    but dislike. Tracked down to plugin spm_scancontrol -- at least they do not
    show if i disable this plugin, but it has a few side effects/relations as it
    hooked up dependencies...:

    (spm_scancontrol.C: spm_scancontrol_query) Gxsm-PlugIn-DEBUG-MESSAGE :
    - spm_scancontrol_query:done

    (gxsm2:23200): Gtk-CRITICAL **: gtk_object_get_data: assertion `GTK_IS_OBJECT
    (object)' failed

    (gxsm2:23200): Gtk-CRITICAL **: gtk_label_set_text: assertion `GTK_IS_LABEL
    (label)' failed

    -- can not locate any corresponding lines/code in the pi. Any idea how to
    track down? Tried via ddd, but seams endless to catch....

     
  • Percy Zahl

    Percy Zahl - 2009-11-15

    cleanup of warnings done with exception of a few plugins I do not maintain.
    Please check yours if you are author of any plugin.

     
  • Thorsten Wagner

    Thorsten Wagner - 2009-11-15

    These plugins are still giving warnings during compilation on my system:

    plugin_ctrl.C 1)

    epsfutils. 2)

    sdfimport.C 3)

    kmdsp_hwi.C 3)

    gmetopng.C 4)

    uksoft2001topng.C 1)

    rhkspm32topng.C 3)

    1) Dereferencing type pun pointer, strict aliasing rule

    2) Undefined operation on "i"

    3) const char*

    4) unknown letter to convert

    Sorry for not posting the compiler output directly - I am running a German
    distribution. I hope that my translation give you a sense of what is the
    problem. Anyway, I can take care of the last three const char * warnings -
    hopefully!

    By the way: Is the critical GTK runtime error fixed? I cannot reproduce it on
    my system. But some plugins are not inserted correctly to the menues, eg. the
    SPM scan control. That's because - as stated already above - I am trying on my
    German ubuntu. So this might give a hint, what is the reason for that error. I
    also found something on that error with the ..._get_text-issue. The posting
    said something about that wrapping is not longer supported.

    Thorsten :-)

     
  • Thorsten Wagner

    Thorsten Wagner - 2009-11-15

    Fixed sdfimport.C, kmdsp_hwi.C 3, and rhkspm32topng.C. No further const char*
    warning should appear now during compilation.

    Thorsten :-)

     
  • Percy Zahl

    Percy Zahl - 2009-11-16

    1) Dereferencing type pun pointer, strict aliasing rule

    not sure what it exactly dislikes/how to fix/do better here in plugin_ctrl.C

    2) Undefined operation on "i"

    no real need to fix -- order of i++'s does not matter here -- would be easy
    fix, just more typing....

    --

    thx for the other remaining fixes :-)

     
  • Percy Zahl

    Percy Zahl - 2009-11-16

    made gcc happy with epsutils now -- recoded.

    add -Wstrict-aliasing=0 to Makefile.am, to turn off complains on plugin_ctrl.C

    • zero warnings left in core.
     

Log in to post a comment.