Re: C++/Java regressions

[mec dot gnu at mindspring dot com (Michael Elizabeth Chastain)]

Hi Ian and David,

My last spin was with gdb HEAD checked out on 2003-11-21 06:25:07.
I got some new demangler code, because my gdb HEAD had
"715 tests, 18 failures" instead of the old "715 tests, 31 failures".
However I did not get the "cp-demangle.c: complete rewrite" version.

I didn't see any demangler-related changes in my results.

I'll fire up some scripts and see if I can see any difference now.

ian> I'm not saying that the problem is not the demangler.  I don't know
ian> what the problem is.  If you can give more information, I would be
ian> happy to adjust the demangler.

Totally reasonable.

For what it's worth, how would I file a demangler bug report?
Do I just file it in gcc bugzilla, because the primary copy of
libiberty lives in gcc?

ian> I enclose the diff of gdb.sum.

Cool!  Just because I'm such a gdb test suite geek, I'll post a
little analysis of what you are seeing.

I'd like to keep improving the test suite so that
"diff gdb-1.sum gdb-2.sum" has less noise in it.
It has been getting better lately.

Michael C

-PASS: gdb.base/pc-fp.exp: get value of $fp (0xbfffe788)
+PASS: gdb.base/pc-fp.exp: get value of $fp (0xbfffe388)

  The test name has a machine address in it, so it fluctuates a bit
  from run to run (obviously).

-PASS: gdb.cp/cplusfuncs.exp: print &'overload1arg(unsigned)'
+PASS: gdb.cp/cplusfuncs.exp: print &'overload1arg(unsigned int)'

  You changed the demangler.

-KFAIL: gdb.cp/cplusfuncs.exp: print &'hairyfunc5' (PRMS: gdb/19)
-KFAIL: gdb.cp/cplusfuncs.exp: print &'hairyfunc6' (PRMS: gdb/19)
-KFAIL: gdb.cp/cplusfuncs.exp: print &'hairyfunc7' (PRMS: gdb/19)
+PASS: gdb.cp/cplusfuncs.exp: print &'hairyfunc5'
+PASS: gdb.cp/cplusfuncs.exp: print &'hairyfunc6'
+PASS: gdb.cp/cplusfuncs.exp: print &'hairyfunc7'

  You fixed a bug in the demangler (a very old bug, gdb/19).
  Cool!

-PASS: gdb.threads/print-threads.exp: Hit kill breakpoint, 9 (slow with kill breakpoint)
 PASS: gdb.threads/print-threads.exp: Hit thread_function breakpoint, 4 (slow with kill breakpoint)
+PASS: gdb.threads/print-threads.exp: Hit kill breakpoint, 9 (slow with kill breakpoint)

  Non-determinism in the order of execution of thread tests.

-PASS: gdb.threads/print-threads.exp: Hit thread_function breakpoint, 5 (slow with kill breakpoint)
 PASS: gdb.threads/print-threads.exp: Hit kill breakpoint, 11 (slow with kill breakpoint)
+PASS: gdb.threads/print-threads.exp: Hit thread_function breakpoint, 5 (slow with kill breakpoint)

  Ditto.

-PASS: gdb.threads/pthreads.exp: continue to bkpt at common_routine in thread 2
-PASS: gdb.threads/pthreads.exp: backtrace from thread 2 bkpt in common_routine
+FAIL: gdb.threads/pthreads.exp: continue to bkpt at common_routine in thread 2

  My understanding is that this is a real race condition between
  gdb and the operating system.  It happens about one time every
  500 or so test runs for me.  Daniel J knows more about this.
  Whatever it is, it's not a demangler problem.

-PASS: gdb.threads/schedlock.exp: other thread 2 didn't run
-PASS: gdb.threads/schedlock.exp: other thread 3 didn't run
 PASS: gdb.threads/schedlock.exp: current thread ran
+PASS: gdb.threads/schedlock.exp: other thread 3 didn't run
+PASS: gdb.threads/schedlock.exp: other thread 4 didn't run

  More thread non-determinism, sigh.  As long as they all pass it's
  okay.