Loading ...
Sorry, an error occurred while loading the content.
 

6.0w configure failure ...

Expand Messages
  • Ron Aaron
    Hi, My clean machine [Mandrake Linux 7.2, with the addition of libiconv added], fails compiling 6.0w: make distclean ./configure make The failure is that it
    Message 1 of 8 , Feb 28, 2001
      Hi,

      My clean machine [Mandrake Linux 7.2, with the addition of 'libiconv' added],
      fails compiling 6.0w:

      make distclean
      ./configure
      make

      The failure is that it seems to be unable to determine the size of an int (!):

      configure:5704: checking size of int
      configure:5723: gcc -o conftest -g -O2 -Wall -I/usr/local/include -L/usr/local/lib conftest.c -lncurses -lelf -liconv -lgpm 1>&5
      configure:5715: warning: return-type defaults to `int'
      configure: failed program was:
      #line 5712 "configure"
      #include "confdefs.h"
      #include <stdio.h>
      main()
      {
      FILE *f=fopen("conftestval", "w");
      if (!f) exit(1);
      fprintf(f, "%d\n", (int)sizeof(int));
      exit(0);
      }


      Far be it from me to whine, but shouldn't inability to determine 'sizeof(int)'
      be a FATAL error that causes all the compilation to cease? I mean, the
      current behavior of just setting it to '0' is ludicrous:
      1) there is never an int size of zero
      2) it makes it incredibly difficult to pinpoint the problem, since the
      compile just 'goes on' even though there is something pretty wrong!
      3) if we just blithely set SIZEOF_INT to something, wouldn't it make
      somewhat more sense to choose '4' as the value, as that is what 98% of
      the machines out there are using?

      Anyway, I don't have any clue why this behaviour is occurring. My laptop,
      also running Mandrake 7.2 (albeit with *tons* of patches and additions)
      compiles just fine. Sometimes it is interesting to wipe a disk and start
      over...

      Ron


      --
      Ron Aaron Moss Bay Engineering
      425-649-8817 Precision Crafted Software
      http://www.mossbayeng.com/~ron/ronsresume.html

      "If you won't be better tomorrow than you are today,
      what need have you for tomorrow?"
      R' Nahman of Bratslav
    • Bram Moolenaar
      ... The problem is probably that linking failed. Strange that there is no error message. Check the other lines in the config.log around this, I suspect other
      Message 2 of 8 , Mar 1, 2001
        Ron Aaron wrote:

        > My clean machine [Mandrake Linux 7.2, with the addition of 'libiconv' added],
        > fails compiling 6.0w:
        >
        > make distclean
        > ./configure
        > make
        >
        > The failure is that it seems to be unable to determine the size of an int (!):
        >
        > configure:5704: checking size of int
        > configure:5723: gcc -o conftest -g -O2 -Wall -I/usr/local/include -L/usr/local/lib conftest.c -lncurses -lelf -liconv -lgpm 1>&5
        > configure:5715: warning: return-type defaults to `int'
        > configure: failed program was:
        > #line 5712 "configure"
        > #include "confdefs.h"
        > #include <stdio.h>
        > main()
        > {
        > FILE *f=fopen("conftestval", "w");
        > if (!f) exit(1);
        > fprintf(f, "%d\n", (int)sizeof(int));
        > exit(0);
        > }
        >
        >
        > Far be it from me to whine, but shouldn't inability to determine
        > 'sizeof(int)' be a FATAL error that causes all the compilation to cease?

        The problem is probably that linking failed. Strange that there is no error
        message. Check the other lines in the config.log around this, I suspect other
        tests that compile and link a program also failed.

        > I mean, the current behavior of just setting it to '0' is ludicrous:
        > 1) there is never an int size of zero
        > 2) it makes it incredibly difficult to pinpoint the problem, since the
        > compile just 'goes on' even though there is something pretty wrong!
        > 3) if we just blithely set SIZEOF_INT to something, wouldn't it make
        > somewhat more sense to choose '4' as the value, as that is what 98%
        > of the machines out there are using?

        Using a default for SIZEOF_INT would just hide the problem. And when you
        would continue to compile Vim, you run into the check in vim.h, which clearly
        explains what the problem is when SIZEOF_INT is zero. Finding a solution is
        a bit more difficult though...

        > Anyway, I don't have any clue why this behaviour is occurring. My laptop,
        > also running Mandrake 7.2 (albeit with *tons* of patches and additions)
        > compiles just fine. Sometimes it is interesting to wipe a disk and start
        > over...

        I would suspect a problem with the libraries. Although it's unclear to me why
        a check would find a library and then later fail to link with it. Perhaps a
        problem with the ordering of -L arguments?

        --
        hundred-and-one symptoms of being an internet addict:
        25. You believe nothing looks sexier than a man in boxer shorts illuminated
        only by a 17" inch svga monitor.

        /// Bram Moolenaar -- Bram@... -- http://www.moolenaar.net \\\
        ((( Creator of Vim - http://www.vim.org -- ftp://ftp.vim.org/pub/vim )))
        \\\ Help me helping AIDS orphans in Uganda - http://iccf-holland.org ///
      • Ron Aaron
        I ll have to check it out tonight. Possibly the old libiconv is conflicting. Ron
        Message 3 of 8 , Mar 1, 2001
          I'll have to check it out tonight. Possibly the old libiconv is
          conflicting.

          Ron

          >-----Original Message-----
          >From: Bram Moolenaar [mailto:Bram@...]
          >Sent: Thursday, March 01, 2001 4:19 AM
          >To: Ron Aaron
          >Cc: VIM Developers
          >Subject: Re: 6.0w configure failure ...
          >
          >
          >
          >Ron Aaron wrote:
          >
          >> My clean machine [Mandrake Linux 7.2, with the addition of
          >'libiconv' added],
          >> fails compiling 6.0w:
          >>
          >> make distclean
          >> ./configure
          >> make
          >>
          >> The failure is that it seems to be unable to determine the
          >size of an int (!):
          >>
          >> configure:5704: checking size of int
          >> configure:5723: gcc -o conftest -g -O2 -Wall
          >-I/usr/local/include -L/usr/local/lib conftest.c -lncurses
          >-lelf -liconv -lgpm 1>&5
          >> configure:5715: warning: return-type defaults to `int'
          >> configure: failed program was:
          >> #line 5712 "configure"
          >> #include "confdefs.h"
          >> #include <stdio.h>
          >> main()
          >> {
          >> FILE *f=fopen("conftestval", "w");
          >> if (!f) exit(1);
          >> fprintf(f, "%d\n", (int)sizeof(int));
          >> exit(0);
          >> }
          >>
          >>
          >> Far be it from me to whine, but shouldn't inability to determine
          >> 'sizeof(int)' be a FATAL error that causes all the
          >compilation to cease?
          >
          >The problem is probably that linking failed. Strange that
          >there is no error
          >message. Check the other lines in the config.log around this,
          >I suspect other
          >tests that compile and link a program also failed.
          >
          >> I mean, the current behavior of just setting it to '0' is ludicrous:
          >> 1) there is never an int size of zero
          >> 2) it makes it incredibly difficult to pinpoint the
          >problem, since the
          >> compile just 'goes on' even though there is
          >something pretty wrong!
          >> 3) if we just blithely set SIZEOF_INT to something,
          >wouldn't it make
          >> somewhat more sense to choose '4' as the value, as
          >that is what 98%
          >> of the machines out there are using?
          >
          >Using a default for SIZEOF_INT would just hide the problem.
          >And when you
          >would continue to compile Vim, you run into the check in
          >vim.h, which clearly
          >explains what the problem is when SIZEOF_INT is zero. Finding
          >a solution is
          >a bit more difficult though...
          >
          >> Anyway, I don't have any clue why this behaviour is
          >occurring. My laptop,
          >> also running Mandrake 7.2 (albeit with *tons* of patches and
          >additions)
          >> compiles just fine. Sometimes it is interesting to wipe a
          >disk and start
          >> over...
          >
          >I would suspect a problem with the libraries. Although it's
          >unclear to me why
          >a check would find a library and then later fail to link with
          >it. Perhaps a
          >problem with the ordering of -L arguments?
          >
          >--
          >hundred-and-one symptoms of being an internet addict:
          >25. You believe nothing looks sexier than a man in boxer
          >shorts illuminated
          > only by a 17" inch svga monitor.
          >
          > /// Bram Moolenaar -- Bram@... --
          >http://www.moolenaar.net \\\
          >((( Creator of Vim -
          >http://www.vim.org -- ftp://ftp.vim.org/pub/vim )))
          > \\\ Help me helping AIDS orphans in Uganda -
          >http://iccf-holland.org ///
          >
        • Ron Aaron
          ... OK, I looked into it more, and still haven t a clue. I manually went into auto/config.h and changed SIZEOF_INT to 4, and everything compiled and works
          Message 4 of 8 , Mar 1, 2001
            Bram Moolenaar <Bram@...> writes:
            >Ron Aaron wrote:
            >
            >> The failure is that it seems to be unable to determine the size of an int (!):
            >>
            >> configure:5704: checking size of int
            >> configure:5723: gcc -o conftest -g -O2 -Wall -I/usr/local/include -L/usr/local/lib conftest.c -lncurses -lelf -liconv -lgpm 1>&5
            >> configure:5715: warning: return-type defaults to `int'
            >> configure: failed program was:
            >> #line 5712 "configure"
            >> #include "confdefs.h"
            >> #include <stdio.h>
            >> main()
            >> {
            >> FILE *f=fopen("conftestval", "w");
            >> if (!f) exit(1);
            >> fprintf(f, "%d\n", (int)sizeof(int));
            >> exit(0);
            >> }
            >>

            >The problem is probably that linking failed. Strange that there is no error
            >message. Check the other lines in the config.log around this, I suspect other
            >tests that compile and link a program also failed.

            OK, I looked into it more, and still haven't a clue.

            I manually went into auto/config.h and changed SIZEOF_INT to 4, and everything
            compiled and works fine. :-<

            I compared the config.log with one from a machine without the problem, and
            there was nothing obvious.

            Ron
          • Bram Moolenaar
            ... It appears that only running the test program failed. Otherwise there would have been error messages. Perhaps this fopen() failed: FILE
            Message 5 of 8 , Mar 2, 2001
              Ron Aaron wrote:

              > Bram Moolenaar <Bram@...> writes:
              > >Ron Aaron wrote:
              > >
              > >> The failure is that it seems to be unable to determine the size of an int (!):
              > >>
              > >> configure:5704: checking size of int
              > >> configure:5723: gcc -o conftest -g -O2 -Wall -I/usr/local/include -L/usr/local/lib conftest.c -lncurses -lelf -liconv -lgpm 1>&5
              > >> configure:5715: warning: return-type defaults to `int'
              > >> configure: failed program was:
              > >> #line 5712 "configure"
              > >> #include "confdefs.h"
              > >> #include <stdio.h>
              > >> main()
              > >> {
              > >> FILE *f=fopen("conftestval", "w");
              > >> if (!f) exit(1);
              > >> fprintf(f, "%d\n", (int)sizeof(int));
              > >> exit(0);
              > >> }
              > >>
              >
              > >The problem is probably that linking failed. Strange that there is no error
              > >message. Check the other lines in the config.log around this, I suspect other
              > >tests that compile and link a program also failed.
              >
              > OK, I looked into it more, and still haven't a clue.
              >
              > I manually went into auto/config.h and changed SIZEOF_INT to 4, and everything
              > compiled and works fine. :-<
              >
              > I compared the config.log with one from a machine without the problem, and
              > there was nothing obvious.

              It appears that only running the test program failed. Otherwise there would
              have been error messages.

              Perhaps this fopen() failed:

              FILE *f=fopen("conftestval", "w");

              Could happen when the file already exists and isn't writable.

              --
              Just remember...if the world didn't suck, we'd all fall off.

              /// Bram Moolenaar -- Bram@... -- http://www.moolenaar.net \\\
              ((( Creator of Vim - http://www.vim.org -- ftp://ftp.vim.org/pub/vim )))
              \\\ Help me helping AIDS orphans in Uganda - http://iccf-holland.org ///
            • Ron Aaron
              Oy! The fundamental problem is, I m an idiot. I installed libiconv 1.5.1 . Recompiled vim, clean. FORGOT to ldconfig and add /usr/local/lib to my
              Message 6 of 8 , Mar 2, 2001
                Oy!

                The fundamental problem is, I'm an idiot.

                I installed 'libiconv 1.5.1'. Recompiled vim, clean. FORGOT to 'ldconfig'
                and add /usr/local/lib to my 'ld.so.conf' file.

                So indeed the linking *was* failing, but in some subtle way which didn't whine
                too loudly.

                Now I'm recompiling, and it went past the configure stage just fine, thank you
                very much... it even works :-~)


                <embarrassed in Seattle>

                Ron
              • Zdenek Sekera
                Ron Aaron wrote: ... Must be the earthquake :-)
                Message 7 of 8 , Mar 2, 2001
                  Ron Aaron wrote:
                  ...
                  >
                  > <embarrassed in Seattle>
                  >

                  Must be the earthquake :-)

                  ---Zdenek
                • Ron Aaron
                  ... Thanks, Zdenek, for saving my pride :-) Ron
                  Message 8 of 8 , Mar 2, 2001
                    Zdenek Sekera <zs@...> writes:
                    >Ron Aaron wrote:
                    >...
                    >>
                    >> <embarrassed in Seattle>
                    >>
                    >
                    >Must be the earthquake :-)
                    >
                    >---Zdenek

                    Thanks, Zdenek, for saving my pride :-)

                    Ron
                  Your message has been successfully submitted and would be delivered to recipients shortly.