Re: address@hidden: Re: [Gcl-devel] apparent GCL 2.5.0 bug on Redhat 7.3]

[Camm Maguire]

Greetings, and thanks for your report!

Matt Kaufmann <address@hidden> writes:

> Hi --
> 
> I re-ran using gcl_2.5.0.cvs20020625-64_i386.deb, and still got an error
> processing mult.lisp (after issuing "make certify-books-fresh" from the main
> ACL2 directory), and also a new one processing simple.lisp.  Snippets from the
> corresponding output files (with the directories as indicated) are shown 
> below.

OK, I cannot reproduce this.  I've downloaded and compiled 2.7, and am
running certify-books for quite a while now without error, way past
mult.cert. 

I don't understand which files contain the log information you quote
below.  But all the certs in data-structures have already been
finished in my process, and no .out file therein contains an error.  

I'm also not quite clear about your GCL setup.  Did you unpack the
.deb on a Redhat box by hand, or are you running on a Debian machine?
Which version of which OS?  Most specifically, which gcc?  If this
machine has the infamous buggy redhat gcc (I think it was 2.96)
released some time ago, I'm afraid we cannot support that.  GCL should
work with all *released* versions of gcc (2.96 was taken by Redhat
from a cvs snapshot), i.e. 2.95, 3.0, 3.1, and 3.2.  I'm still using
2.95 on my machine.

The other possibility, especially if you've unpacked the .deb by hand,
is that the library dependencies are messed up.  The package manager
takes care of such details in normal circumstances automatically.  By
hand, you might want to run ldd on
/usr/lib/gcl-2.5.0/unixport/saved_gcl, and pay particular attention to
the gmp library on your system.  

I'd suggest the following:

1) Please report OS version, gcc version, and the versions of any
   libraries saved_gcl is loading dynamically on your system.
2) If on Redhat, please try the .rpm in the same cvs subdir of the ftp
   site.  Your package manager should hopefully complain if there are
   dependency problems. (Though the rpm is built from the .deb
   automatically, and has not to my knowledge been tested).
3) If none of the above clears this up, the only way I'm going to be
   able to help is to either run the offending process under gdb,
   using a gcl compiled with debugging, or have someone else do so and
   report the results.

My certify process just executed:

Making expt.cert on Fri Nov  8 12:18:59 EST 2002

Take care,
 


> 
> ==================== books/misc/mult.out ====================
> 
> "/home/kaufmann/apps/acl2/v2-7/books/data-structures/deflist.lisp"
> [SGC for 103 STRING pages..(2337 writable)..(T=5).GC finished]
> [SGC for 103 STRING pages..(2355 writable)..(T=5).GC finished]
> Loading /home/kaufmann/apps/acl2/v2-7/books/data-structures/alist-defuns.o
> start address -T 0x9bf1000 Finished loading 
> /home/kaufmann/apps/acl2/v2-7/books/data-structures/alist-defuns.o
> [SGC for 103 STRING pages..(2375 writable)..(T=5).GC finished]
> [SGC for 103 STRING pages..(2401 writable)..(T=6).GC finished]
> Loading /home/kaufmann/apps/acl2/v2-7/books/data-structures/defalist.o
> [SGC off]
> Error: Caught fatal error [memory may be damaged]
> Fast links are on: do (si::use-fast-links nil) for debugging
> Error signalled by LOAD-COMPILED-FILE-IF-MORE-RECENT.
> Broken at COND.  Type :H for Help.
> ACL2>>
> 
> ==================== books/powerlists/simple.out ====================
> 
> ACL2::ACL2-DEFAULTS-TABLE
> [SGC for 1004 CONS pages..(3111 writable)..(T=9).GC finished]
> [SGC for 116 STRING pages..(3143 writable)..(T=10).GC finished]
> Loading /home/kaufmann/apps/acl2/v2-7/books/data-structures/utilities.o
> [SGC off]
> Error: Caught fatal error [memory may be damaged]
> Fast links are on: do (si::use-fast-links nil) for debugging
> Error signalled by ACL2::LOAD-COMPILED-FILE-IF-MORE-RECENT.
> Broken at COND.  Type :H for Help.
> U>>
> 
> -- Matt
> ------- Start of forwarded message -------
> Date: 7 Nov 2002 10:42:42 -0600
> From: Matt Kaufmann <address@hidden>
> To: address@hidden
> CC: address@hidden, address@hidden
> In-reply-to: <address@hidden> (message from Camm Maguire
>       on 07 Nov 2002 11:26:27 -0500)
> Subject: Re: [Gcl-devel] apparent GCL 2.5.0 bug on Redhat 7.3
> 
> Hi --
> 
> >> Could you please tell me where you got the GCL you tested on Redhat
> >> and on Debian?
> 
> If I'm not mistaken, I used the binary provided by the following CVS snapshot,
> which I unpacked using ar, gunzip, and tar xvf (much as explained in
> HOWTO-UNPACK-DEBS).
> 
> ftp://ftp.gnu.org/gnu/gcl/cvs/gcl_2.5.0.cvs20020625-56_i386.deb
> 
> I did not compile it myself.  Would you like me to try that?
> 
> Later this morning I tried gcl_2.5.0.cvs20020625-64_i386.deb, though it wasn't
> much of a test since I didn't recompile all compiled ACL2 files being loaded.
> 
> >> Lastly, is there remote access to this machine?
> 
> No, sorry; it's my home computer and someone else set up my home network, so I
> wouldn't begin to know how to provide access.  (If you're in Austin you're
> welcome to come sit at it!)
> 
> I'll start from scratch as soon as I send this off and let you know tonight
> (gotta head off to work now; I'll have email contact again there).
> 
> By the way, we distribute ACL2 images built on GCL.  Is there any disclaimer
> you'd like us to put on the web about using CVS snapshots of GCL?
> 
> Finally, the Debian GCL I used without error was installed by Univ. of Texas
> sysadmins.  Perhaps the following answers your question; if not, let me know
> and I'll try to find out from them, since I couldn't find anything definitive
> on the filesystem.
> 
>   glenmorangie.cs.utexas.edu% cat gcl-2.5.0-linuxP3/CVS/Entries
>   /AC_FD_CC/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /AC_FD_MSG/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /COPYING.LIB-2.0/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /ChangeLog/1.14/Sun Feb  3 18:44:06 2002/-ko/
>   /ChangeLog.old/1.1/Mon Feb  4 02:11:53 2002//
>   /acconfig.h/1.16/Wed Aug 14 22:02:02 2002//
>   /add-defs/1.2/Sat Dec  9 07:10:35 2000/-ko/
>   /add-defs.bat/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /add-defs1/1.6/Thu Jun  6 18:23:11 2002/-ko/
>   /config.guess/1.2/Mon Dec 17 20:48:33 2001/-ko/
>   /config.sub/1.2/Mon Dec 17 20:48:33 2001/-ko/
>   /configure/1.74/Wed Aug 21 22:22:50 2002/-ko/
>   /configure-new.ac/1.2/Tue Jul 30 06:29:35 2002//
>   /configure.in/1.77/Wed Aug 21 22:22:50 2002/-ko/
>   /eval.html/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /eval.tcl/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /faq/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /gcl.jpg/1.1.1.1/Mon Dec  6 22:43:55 1999/-ko/
>   /gcl.png/1.2/Sun Jan 20 02:09:23 2002//
>   /gcl1.jpg/1.1.1.1/Mon Dec  6 22:43:55 1999/-ko/
>   /gcl2.jpg/1.1.1.1/Mon Dec  6 22:43:55 1999/-ko/
>   /gmp.patch/1.1/Wed Jul 31 03:13:52 2002//
>   /install.sh/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /ltconfig/1.1/Fri Aug  9 05:49:32 2002//
>   /ltmain.sh/1.1/Fri Aug  9 05:49:32 2002//
>   /machine/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /machines/1.2/Wed Jul 10 08:02:21 2002/-ko/
>   /majvers/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /makdefs/1.2/Tue Dec 18 18:50:17 2001/-ko/
>   /makedefc.in/1.10/Wed Aug 21 22:22:50 2002//
>   /makefile/1.39/Sun Aug 18 16:15:10 2002/-ko/
>   /merge.c/1.1.1.1/Mon Dec  6 22:43:54 1999/-ko/
>   /minvers/1.5/Sun Feb  3 18:44:06 2002/-ko/
>   /readme/1.4/Mon May 15 19:24:26 2000/-ko/
>   /readme.gmp/1.1/Fri May 11 05:56:54 2001//
>   /readme.mingw/1.7/Wed Jul 31 07:10:03 2002//
>   D
>   glenmorangie.cs.utexas.edu% 
> 
> - -- Matt
>    Cc: address@hidden, address@hidden
>    From: Camm Maguire <address@hidden>
>    Date: 07 Nov 2002 11:26:27 -0500
> 
>    Greetings!
> 
>    Matt Kaufmann <address@hidden> writes:
> 
>    > Hi --
>    > 
>    > I just ran across an apparent GCL 2.5.0 bug on Redhat 7.3.  I performed
>    > essentially the same action on a Debian Gnu Linux platform (GCL 2.5.0) 
> as well
>    > as many others (including Solaris GCL 2.3.8 and many other Lisps) without
>    > running into this problem.  The relevant part of the log is as follows.
>    > 
> 
>    Could you please tell me where you got the GCL you tested on Redhat
>    and on Debian?  Of course, 2.5.0 has not yet been released, so this
>    must be a CVS snapshot.  Did you compile yourself?  Use the binary
>    packages on the ftp site?  Which versions?  Lastly, is there remote
>    access to this machine?  The error is likely occurring because some
>    compiled objects refer to symbols which cannot be found in the main
>    image.  
> 
>    Thanks!
> 
> 
>    >   [SGC for 94 STRING pages..(5707 writable)..(T=24).GC finished]
>    >   Loading 
> /home/kaufmann/apps/acl2/acl2-sources/books/data-structures/structures.o
>    >   [SGC for 0 CONTIGUOUS-BLOCKS pages..(5722 writable)..(T=25).GC 
> finished]
>    >   [SGC off][GC for 500 RELOCATABLE-BLOCKS pages..(T=50).GC finished]
>    >   [SGC on]
>    >   Error: Cannot get relocated section contents
>    > 
>    >   Fast links are on: do (si::use-fast-links nil) for debugging
>    >   Error signalled by INCLUDE-BOOK-FN.
>    >   Broken at COND.  Type :H for Help.
>    >   ACL2>>
>    > 
>    > Here is how you can re-create this bug (at least, on Redhat 7.3; at 
> least on
>    > *my* laptop).  First download
>    > ftp://ftp.cs.utexas.edu:/pub/moore/acl2/v2-7/acl2.tar.gz.  Then (where 
> you can
>    > replace "gcl" below with a path to your GCL):
>    > 
>    >   gunzip acl2.tar.gz
>    >   tar xvf acl2.tar
>    >   cd acl2-sources
>    >   make LISP=gcl
>    >   make certify-books
>    > 
>    > The error will appear in books/misc/mult.out.  It will take a little 
> while
>    > (about 15 min. on my 850 MH Duron laptop) to get to the problem after 
> make
>    > certify-books is started.  The above mult.out file is the result of 
> ACL2's
>    > processing ("certification") of file mult.lisp; none of the other 
> several dozen
>    > such certifications completed so far have caused an error in the above 
> make
>    > certify-books.
>    > 
>    > By the way, the above acl2.tar.gz may not be our final release of ACL2 
> Version
>    > 2.7, though it's close.  So please do not redistribute it, but feel free 
> to use
>    > it in your testing.  In fact, we plan to finalize the release next 
> Wednesday
>    > (11/13), so if you find any issues before then, we'd be very interested!
>    > 
>    > Thanks --
>    > -- Matt
>    > 
>    > 
>    > _______________________________________________
>    > Gcl-devel mailing list
>    > address@hidden
>    > http://mail.gnu.org/mailman/listinfo/gcl-devel
>    > 
>    > 
> 
>    -- 
>    Camm Maguire                                               address@hidden
>    ==========================================================================
>    "The earth is but one country, and mankind its citizens."  --  Baha'u'llah
> ------- End of forwarded message -------
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah