make core dumps

make core dumps

Post by Mike Harriso » Thu, 11 Mar 1999 04:00:00



I've come across a problem with make. Every time I try to use make, it core
dumps. This is on everything I download that needs to be maked. I don't
understand this at all. It was working fine before today. I have libc5.4.46
installed on Slackware 3.3, linux kernel 2.0.30.

Any ideas? Thanks,

Mike

 
 
 

make core dumps

Post by Paul D. Smi » Thu, 11 Mar 1999 04:00:00



  mh> I've come across a problem with make. Every time I try to use
  mh> make, it core dumps. This is on everything I download that needs
  mh> to be maked. I don't understand this at all. It was working fine
  mh> before today. I have libc5.4.46 installed on Slackware 3.3, linux
  mh> kernel 2.0.30.

You didn't say what version of make you were using.

Did you recently update your libc?  There was a problem with GNU make
3.75 and some versions of libc; see the README for the new libc you
installed, or upgrade to a newer version of GNU make.

--
-------------------------------------------------------------------------------

 "Please remain calm...I may be mad, but I am a professional." --Mad Scientist
-------------------------------------------------------------------------------
   These are my opinions---Nortel Networks takes no responsibility for them.

 
 
 

make core dumps

Post by Michael Davi » Thu, 11 Mar 1999 04:00:00




Quote:>I've come across a problem with make. Every time I try to use make, it core
>dumps. This is on everything I download that needs to be maked. I don't
>understand this at all. It was working fine before today. I have libc5.4.46
>installed on Slackware 3.3, linux kernel 2.0.30.

>Any ideas? Thanks,

Hi Interlog neighbor,

When 'make' bombs on me (actually on Solaris, which is what I work
on on my day job), it's because some of my environment variables
have values that are too long. Either PATH is too long, or
LD_LIBRARY_PATH is too long, or there are too many items
in another variable which gets split into a series of -I arguments
to the compier.

If you have the source to 'make', you might find a fixed-size
string that you can make bigger, or better yet dynamically
allocate...

Good luck

--
// Michael Davis -- Solaris code slave and happy Linux User.
//
// From sunny Toronto...

 
 
 

make core dumps

Post by Mike Harris » Thu, 11 Mar 1999 04:00:00





>  mh> I've come across a problem with make. Every time I try to use
>  mh> make, it core dumps. This is on everything I download that needs
>  mh> to be maked. I don't understand this at all. It was working fine
>  mh> before today. I have libc5.4.46 installed on Slackware 3.3, linux
>  mh> kernel 2.0.30.

>You didn't say what version of make you were using.

>Did you recently update your libc?  There was a problem with GNU make
>3.75 and some versions of libc; see the README for the new libc you
>installed, or upgrade to a newer version of GNU make.

Yes, I updated libc to 5.4.46 today (from 5.4.33), but make still
worked after the update.  I don't remember doing anything else that
might cause the problem, but who knows I must have.  I have tried both
make 3.74 and GNU make 3.74. I couldn't find a more recent version on
the internet, where is the best place to get the latest?

Thanks,

Mike
Mike Harrison

http://www.interlog.com/~mharris

 
 
 

make core dumps

Post by Paul D. Smi » Thu, 11 Mar 1999 04:00:00



  mh> Yes, I updated libc to 5.4.46 today (from 5.4.33), but make still
  mh> worked after the update.

Probably you hadn't fully installed; maybe you hadn't rebooted, or
re-run ldconfig, or something, and make was still using the older
version of libc somehow.

  mh> I don't remember doing anything else that might cause the problem,
  mh> but who knows I must have.  I have tried both make 3.74 and GNU
  mh> make 3.74.

Not sure why you're making a distinction here: there's only one make
typically available on any Linux system and that's GNU make.  Anyway,
any GNU make version less than 3.76 will show this problem.

  mh> I couldn't find a more recent version on the internet, where is
  mh> the best place to get the latest?

I'm pretty surprised you couldn't find one, since the GNU tools are
mirrored on hundreds of systems all over the world.

The canonical FTP site for all GNU tools is ftp://ftp.gnu.org/; you can
find GNU make at:

  ftp://ftp.gnu.org/gnu/make/

See http://www.gnu.org/ for information and mirrors, etc.  The latest is
3.77, but 3.76.1 will work too.

--
-------------------------------------------------------------------------------

 "Please remain calm...I may be mad, but I am a professional." --Mad Scientist
-------------------------------------------------------------------------------
   These are my opinions---Nortel Networks takes no responsibility for them.

 
 
 

make core dumps

Post by Robert Wues » Fri, 12 Mar 1999 04:00:00



> I've come across a problem with make. Every time I try to use make, it core
> dumps. This is on everything I download that needs to be maked. I don't
> understand this at all. It was working fine before today. I have libc5.4.46
> installed on Slackware 3.3, linux kernel 2.0.30.

> Any ideas? Thanks,

> Mike


An excerpt from from H.J. Lu's release.libc-5.4.46:

A dirent bug, which erroneously defined d->reclen to d->namlen if
USE_GNU was defined, has been fixed.  Unfortunately, some GNU packages
depend on this bug.  GNU make 3.xx is one of them. GNU make 3.75 or
above is fine. A patch is included here for 3.74 and below.

(This patch is from libc-5.4.38)

RCS file: /home/cvs/gnu/make/dir.c,v
retrieving revision 1.1.1.1
diff -c -r1.1.1.1 dir.c
*** 1.1.1.1     1995/06/25 03:27:16
--- dir.c       1995/06/25 03:55:22
***************
*** 20,26 ****

  #if   defined (POSIX) || defined (HAVE_DIRENT_H) || defined
(__GNU_LIBRARY__)
  #include <dirent.h>
! #ifndef       __GNU_LIBRARY__
  #define D_NAMLEN(d) strlen((d)->d_name)
  #else /* GNU C library.  */
  #define D_NAMLEN(d) ((d)->d_namlen)
--- 20,26 ----

  #if   defined (POSIX) || defined (HAVE_DIRENT_H) || defined
(__GNU_LIBRARY__)
  #include <dirent.h>
! #ifndef       __BAD_GNU_LIBRARY__
  #define D_NAMLEN(d) strlen((d)->d_name)
  #else /* GNU C library.  */
  #define D_NAMLEN(d) ((d)->d_namlen)
===================================================================
RCS file: /home/cvs/gnu/make/glob/glob.c,v
retrieving revision 1.1.1.1
diff -c -r1.1.1.1 glob.c
*** 1.1.1.1     1995/06/25 03:27:18
--- glob/glob.c 1995/06/25 03:56:01
***************
*** 64,70 ****

  #if   defined (POSIX) || defined (HAVE_DIRENT_H) || defined
(__GNU_LIBRARY__)
  #include <dirent.h>
! #ifndef       __GNU_LIBRARY__
  #define D_NAMLEN(d) strlen((d)->d_name)
  #else /* GNU C library.  */
  #define D_NAMLEN(d) ((d)->d_namlen)
--- 64,70 ----

  #if   defined (POSIX) || defined (HAVE_DIRENT_H) || defined
(__GNU_LIBRARY__)
  #include <dirent.h>
! #ifndef       __BAD_GNU_LIBRARY__
  #define D_NAMLEN(d) strlen((d)->d_name)
  #else /* GNU C library.  */
  #define D_NAMLEN(d) ((d)->d_namlen)

 
 
 

make core dumps

Post by Paul D. Smi » Sat, 13 Mar 1999 04:00:00



  md> When 'make' bombs on me (actually on Solaris, which is what I work
  md> on on my day job), it's because some of my environment variables
  md> have values that are too long. Either PATH is too long, or
  md> LD_LIBRARY_PATH is too long, or there are too many items
  md> in another variable which gets split into a series of -I arguments
  md> to the compier.

This has nothing to do with make.  These size limitations are imposed by
the kernel and/or your shell.  Limitations on the size of the values of
environment variables like PATH are typically imposed by the shell (csh,
for example, has a notoriously small limit).  Limitations on the number
of arguments you can pass to subprocesses are imposed by the kernel,
which has a maximum size that all arguments and environment variables
must fit into.

There's absolutely nothing make can do about either of these issues.
You'd have the same problems invoking similar commands from your shell
prompt.

  md> If you have the source to 'make', you might find a fixed-size
  md> string that you can make bigger, or better yet dynamically
  md> allocate...

You don't say what make you're using: in GNU make there are no such
fixed-size strings.  Your command script can be as large as can fit in
memory... much larger than the kernel will actually allow you to
execute.

Obviously we have no idea if there are any fixed size strings in Solaris
make since we don't have the source.

--
-------------------------------------------------------------------------------

 "Please remain calm...I may be mad, but I am a professional." --Mad Scientist
-------------------------------------------------------------------------------
   These are my opinions---Nortel Networks takes no responsibility for them.

 
 
 

1. DP2 Java 1.3 makes core dump

Hi,

I have a problem with the new DP2 disk, and I'm not
sure how to troubleshoot it further.

I started with a fresh disk (to be sure it was repeatable)
and installed from CDROM DP2 (just issued a couple weeks
ago).

I picked the X Developer package.

Next I installed the Jetty port ( cd to
/usr/ports/java/jetty )
and type make - this caused (by dependency) the JDK1.3 to
install
and it did at patch level 7.  Everything seemed to go fine
(wow
amazing - downloading stuff from Sun's web site and all!)
and I
tested the JDK with a couple of simple programs -
everything looks
fine.

BUT, when I run the Jetty demo, I get a BUSS ERROR and a
core dump
(a file called java.core appears in the directory).  I
tried to write
a couple of simple programs to see if they would cause the
buss error,
like a two threaded "Hello World" but nothing did.  I'm not
sure
if I can get source for Jetty to try and find exactly where
it dies -
(It's not at the contructor for the server, it's when I
actually try and
start it with a
 server.start();
call.

Here is the code that crashes, although running the Jetty
demo will
produce
the same buss error:

HttpServer server = new HttpServer();
try
{
    server.addListener(new InetAddrPort(8080));

HandlerContext context = server.addContext("/");
ServletHandlerContext servHand= new ServletHandlerContext(
server, "/"
);
servHand.setDynamicServletPathSpec( "/servlet/*");

server.addContext( null, servHand );
context.setResourceBase("./docroot/");
context.setServingResources(true);

try
{
    server.start();  //blammo! this line crashes

I'd very much like to help if I could, but I'm not sure
where to go
or if this might already be fixed in -Current

Best Regards,

-Will


with "unsubscribe freebsd-java" in the body of the message

2. Out-of-the-box web page

3. dump core or not dump core

4. Apache 1.2

5. core dumps core dumps everywhere...

6. Bug in sort (yes, really)

7. Making a setgid program dump core?

8. who wants to marry a sysadmin?

9. Dynamic Libraries: core dump after changes have been made

10. Core dump not making proper filename

11. 2.5 + X makes Emacs dump core (+ patch)

12. making linux core dump

13. What makes program core dump when output piped?