QT problem with ar/ranlib - SGI

This is a discussion on QT problem with ar/ranlib - SGI ; Hi NG, I managed to compile the QT-Library itself but I failed to compile the designer or the KDE artslib. I replaced the "missing" ranlib with an simple script with ar s %1 or ar rs %1 but I get ...

+ Reply to Thread
Results 1 to 7 of 7

Thread: QT problem with ar/ranlib

  1. QT problem with ar/ranlib

    Hi NG,

    I managed to compile the QT-Library itself but I failed to compile the
    designer or the KDE artslib. I replaced the "missing" ranlib with an simple
    script with

    ar s %1

    or

    ar rs %1

    but I get an Error:

    /usr/bin/ar[3]: cannot fork: too many processes

    Tried to compile some qt example programs without any trouble.

    Any hints ?

    Holger


  2. Re: QT problem with ar/ranlib

    Holger Banko wrote:

    > Hi NG,
    >
    > I managed to compile the QT-Library itself but I failed to compile the
    > designer or the KDE artslib. I replaced the "missing" ranlib with an simple
    > script with
    >
    > ar s %1
    >
    > or
    >
    > ar rs %1
    >
    > but I get an Error:
    >
    > /usr/bin/ar[3]: cannot fork: too many processes


    thread limit? processes job limit? nproc kernel tuneable? maxup kernel tuneable?

    I'm also puzzled by the syntax -- "ar rs" expects two arguments (an archive
    and a file); ar s expects another option besides the name of an archive.

    Why not "ar -ts %1 >/dev/null" ?


  3. Re: QT problem with ar/ranlib

    Alexis Cousein wrote:
    >>
    >> /usr/bin/ar[3]: cannot fork: too many processes

    >
    > I'm also puzzled by the syntax -- "ar rs" expects two arguments (an
    > archive and a file); ar s expects another option besides the name of an
    > archive.
    >
    > Why not "ar -ts %1 >/dev/null" ?


    Seems that helped a little bit, it compiled a little bit more but then it
    failed again with "cannot fork: too many processes".

    I tried to replace the ar just with a "test %1" but then the compilation
    fails because it has problems to find *.o files.

    So I thin the ar is really needed.

    > thread limit? processes job limit? nproc kernel tuneable? maxup kernel
    > tuneable?


    How can I change this ? I am an IRIX/SGI newbie.

    Holger


  4. Re: QT problem with ar/ranlib

    Alexis Cousein wrote:


    > Do limit and check "threads".
    >
    > Do "jlimits" (if it doesn't exist, you're safe there).


    no command limit or jlimits.

    > systune | grep maxup


    is 778

    > systune | grep nproc


    is 3112

    Ok I will check the manpage of systune.

    --Holger


  5. Re: QT problem with ar/ranlib

    Brent Casavant wrote:
    > ranlib is unnecessary on IRIX. See the hints at:
    >
    > http://freeware.sgi.com/howto.html
    >
    > for a brief discussion.
    >
    > I also seem to recall successfully substituting the true(1) command for
    > ranlib in various compilations circa 1998.


    I replaced ar with nothing but it seems that then the compilation has
    problems to find .o files.

    g++ -c -fno-exceptions -Wall -W -O2 -fPIC -DQT_NO_DEBUG -DQT_THREAD_SUPPORT
    -DQT_PLUGIN -DQT_SHARED -I/usr/lib/qt3/mkspecs/irix-g++ -I.
    -I../../interfaces -I../../../../include -I.moc/release-shared-mt/ -o
    ..obj/release-shared-mt/moc_mainwindowwizard.o
    ..moc/release-shared-mt/moc_mainwindowwizard.cpp
    test -d ../../../../plugins/designer/ || mkdir -p
    .../../../../plugins/designer/
    rm -f libwizards.so
    g++ -fno-exceptions -shared -Wl,-rpath,/usr/lib/qt3/plugins/designer
    -o libwizards.so .obj/release-shared-mt/main.o
    ..obj/release-shared-mt/sqlformwizardimpl.o
    ..obj/release-shared-mt/sqlformwizard.o
    ..obj/release-shared-mt/mainwindowwizard.o
    ..obj/release-shared-mt/qmake_image_collection.o
    ..obj/release-shared-mt/moc_sqlformwizardimpl.o
    ..obj/release-shared-mt/moc_sqlformwizard.o
    ..obj/release-shared-mt/moc_mainwindowwizard.o -Wl,-rpath,/usr/lib/qt3/lib
    -L/usr/lib/qt3/lib -lC -lqt-mt -lz -lGLU -lGL -lXmu -lXext -lX11 -lm -lSM
    -lICE -lpthread
    g++: .obj/release-shared-mt/main.o: No such file or directory
    g++: .obj/release-shared-mt/sqlformwizardimpl.o: No such file or directory
    g++: .obj/release-shared-mt/sqlformwizard.o: No such file or directory
    g++: .obj/release-shared-mt/mainwindowwizard.o: No such file or directory
    g++: .obj/release-shared-mt/qmake_image_collection.o: No such file or
    directory
    g++: .obj/release-shared-mt/moc_sqlformwizardimpl.o: No such file or
    directory
    g++: .obj/release-shared-mt/moc_sqlformwizard.o: No such file or directory
    g++: .obj/release-shared-mt/moc_mainwindowwizard.o: No such file or
    directory
    *** Error code 1 (bu21)

    I experimented with it a little bit. Very strange, that when I replace the
    ar with test or whatever, sometimes the compilation continues after the
    "fork error". Then it breaks, again with the .o problem and I switch back
    to ar. It continues again to break on another point. But now it always
    stops at this point. Same with artslib... Qt Examples comiles without
    problems, running perfect. I dont really need this QT-Designer where it
    fails, but I really need artslib.



  6. Re: QT problem with ar/ranlib

    Holger Banko wrote:

    > Brent Casavant wrote:
    >
    >>ranlib is unnecessary on IRIX. See the hints at:
    >>
    >>http://freeware.sgi.com/howto.html
    >>
    >>for a brief discussion.
    >>
    >>I also seem to recall successfully substituting the true(1) command for
    >>ranlib in various compilations circa 1998.

    >
    >
    > I replaced ar with nothing but it seems that then the compilation has
    > problems to find .o files.
    >
    > g++ -c -fno-exceptions -Wall -W -O2 -fPIC -DQT_NO_DEBUG -DQT_THREAD_SUPPORT
    > -DQT_PLUGIN -DQT_SHARED -I/usr/lib/qt3/mkspecs/irix-g++ -I.
    > -I../../interfaces -I../../../../include -I.moc/release-shared-mt/ -o
    > .obj/release-shared-mt/moc_mainwindowwizard.o


    are you sure that .obj directory exists? It rather looks to me like it
    means ./obj/...


  7. Re: QT problem with ar/ranlib

    * Holger Banko
    | > I also seem to recall successfully substituting the true(1)
    | > command for ranlib in various compilations circa 1998.
    |
    | I replaced ar with nothing but it seems that then the compilation has
    | problems to find .o files.

    You definitely need `ar' to create static libs, but you don't need
    `ranlib' on IRIX to update the static lib index. I `fix' problems
    like these by appending
    RANLIB = :
    to the Makefiles in question, assuming that RANLIB is the make
    variable which is used for `ranlib', and that `sh' is called by make
    as SHELL (csh's ":" command does not accept any arguments).

    R'

+ Reply to Thread