Re: java/6092: sparc-sun-solaris2.7 has hundreds of libjava failures with -m64

["Kaveh R. Ghazi" <ghazi@caip.rutgers.edu>]

The following reply was made to PR java/6092; it has been noted by GNATS.

From: "Kaveh R. Ghazi" <ghazi@caip.rutgers.edu>
To: tromey@redhat.com
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: java/6092: sparc-sun-solaris2.7 has hundreds of libjava failures with -m64
Date: Sat, 30 Mar 2002 15:33:58 -0500 (EST)

  > From: Tom Tromey <tromey@redhat.com>
  > 
  > >>>>> "Kaveh" == Kaveh R Ghazi <ghazi@caip.rutgers.edu> writes:
  > 
  > Kaveh> I looked at last night's testrun to see the context around each
  > Kaveh> failure.  Unfortunately there aren't any useful error messages, it
  > Kaveh> just compiles/PASSes and then spawn FAILs the run.  I've included a
  > Kaveh> bunch of them here using "grep -B 3 ^FAIL libjava.log" so you can see
  > Kaveh> what I mean.
  > 
  > Maybe running dejagnu with `--verbose' will give more information.
 
 Ok, I ran it with RUNTESTFLAGS containing --verbose in addition to the
 --target_board and it did say a little more between the spawn and the
 FAIL line.  Now I get mysterious crash messages like this:
 
   spawn [open ...]
   7 file6
   close result is child killed: illegal instruction
   got
 
 or this:
 
   spawn [open ...]^M
   7 file6
   close result is child killed: SIGABRT
   got
 
 or this:
 
   spawn [open ...]^M
   7 file6
   close result is child killed: segmentation violation
 
 (I don't know what the "7 file6" means.)  However, I also get lots of
 random other failures about "output from source compiled test" or
 "output from bytecode->native test" doesn't match.  So there doesn't
 seem to be an obvious pattern.
 
 
  > The libjava test suite doesn't remove an executable if it fails.  So
  > you could also try running the executable by hand to see if it fails
  > outside the test suite.  (This isn't always perfectly accurate due to
  > environmental differences; for instance the test suite sets
  > LD_LIBRARY_PATH.)
  > Thanks for looking at this.
  > Tom
 
 Hmm, actually it *does* remove the FAILed executables.  Well, let's
 see, I was running several passes and the -m64 pass goes first.  I
 guess the subsequent passes without -m64 PASSed these tests and
 removed the executable.  :-)
 
 So I reran it by hand and caught one of the executable files at random
 (Array_3) before the next pass started.  I had to setup
 LD_LIBRARY_PATH to contain the right directories (I think), here's
 what I get from `ldd Array_3' showing what shared libs will be used:
 
         libm.so.1 =>     /usr/lib/64/libm.so.1
         libgcc_s_sparcv9.so.1 =>
         /teal/caip5/ghazi/gcc-testing/branch/build/gcc/libgcc_s_sparcv9.so.1
         libgcj.so.3 =>
         /teal/caip5/ghazi/gcc-testing/branch/build/sparc-sun-solaris2.7/sparcv9/libjava/.libs/libgcj.so.3
         libpthread.so.1 =>       /usr/lib/64/libpthread.so.1
         librt.so.1 =>    /usr/lib/64/librt.so.1
         libsocket.so.1 =>        /usr/lib/64/libsocket.so.1
         libnsl.so.1 =>   /usr/lib/64/libnsl.so.1
         libdl.so.1 =>    /usr/lib/64/libdl.so.1
         libc.so.1 =>     /usr/lib/64/libc.so.1
         libaio.so.1 =>   /usr/lib/64/libaio.so.1
         libmp.so.2 =>    /usr/lib/64/libmp.so.2
         libthread.so.1 =>        /usr/lib/64/libthread.so.1
         /usr/platform/SUNW,Ultra-Enterprise-10000/lib/sparcv9/libc_psr.so.1
 
 That seems ok.  So I ran it and got:
 	Abort (core dumped)
 
 I tried it under gdb-4.18, but it said:
 	Array_3": not in executable format: File format not recognized
 
 I went to another machine that had gdb 5.0 installed and got:
 Starting program:
 	/teal/caip5/ghazi/gcc-testing/branch/build/sparc-sun-solaris2.7/libjava/testsuite/Array_3
 	procfs:4036 -- process not stopped.
 	procfs: ...giving up...
 
 After that I gave up.  (I don't know whether the gdb problem is because
 gdb doesn't understand -m64 executables or because it doesn't
 understand java produced code or maybe the local installation was
 broken.)
 
 If you'd like me to attempt anything else let me know.  Alternatively,
 you could get a solaris2 box and run make check with:
 setenv RUNTESTFLAGS "--verbose --target_board='unix{-m64,}'"
 to get both regular and -m64 passes.
 
 	Thanks for your help.
 		--Kaveh
 --
 Kaveh R. Ghazi			Director of Systems Architecture
 ghazi@caip.rutgers.edu		Qwest Global Services