[commit/java] Work around java/1565 in jmisc.exp

[commit/java] Work around java/1565 in jmisc.exp

[Andrew Cagney]

[-- Attachment #1: Type: text/plain, Size: 219 bytes --]

Hello,

This patch modifies jmisc.exp to work around the very very long standing 
bug java/1565 (just that no one thought to report it?).  I'll separatly 
follow up with a testcase for just that bug.

committed,
Andrew

[-- Attachment #2: diffs --]
[-- Type: text/plain, Size: 877 bytes --]

2004-02-24  Andrew Cagney  <cagney@redhat.com>

	* gdb.java/jmisc.exp: Use the fully quallified name of "main" when
	setting a breakpoint.  Work around PR java/1565.

Index: testsuite/gdb.java/jmisc.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.java/jmisc.exp,v
retrieving revision 1.2
diff -u -r1.2 jmisc.exp
--- testsuite/gdb.java/jmisc.exp	6 Mar 2001 08:21:58 -0000	1.2
+++ testsuite/gdb.java/jmisc.exp	24 Feb 2004 16:08:44 -0000
@@ -66,7 +66,9 @@
 gdb_test "set print sevenbit-strings" ".*"
 
 if ![set_lang_java] then {
-    runto ${testfile}.main
+    # Ref PR gdb:java/1565.  Don't use the simpler "break jmisc.main".
+    # As of 2004-02-24 it wasn't working and is being tested separatly.
+    runto "\'${testfile}.main(java.lang.String\[\])\'"
 
     send_gdb "ptype jmisc\n"   
     gdb_expect {   

Re: [commit/java] Work around java/1565 in jmisc.exp

[Daniel Jacobowitz]

On Tue, Feb 24, 2004 at 11:12:11AM -0500, Andrew Cagney wrote:
> Hello,
> 
> This patch modifies jmisc.exp to work around the very very long standing 
> bug java/1565 (just that no one thought to report it?).  I'll separatly 
> follow up with a testcase for just that bug.

FYI, about that bug: GCC's debug info is screwy; we need to strip
arguments and parentheses from DW_AT_name for language_java and then I
expect it will start working.  I want to finish my current massacre of
dwarf2read before I go playing with that, though - it's always been on
a lower burner.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [commit/java] Work around java/1565 in jmisc.exp

[Andrew Cagney]

> On Tue, Feb 24, 2004 at 11:12:11AM -0500, Andrew Cagney wrote:
> 
>>> Hello,
>>> 
>>> This patch modifies jmisc.exp to work around the very very long standing 
>>> bug java/1565 (just that no one thought to report it?).  I'll separatly 
>>> follow up with a testcase for just that bug.
> 
> 
> FYI, about that bug:

FYI:

 > GCC's debug info is screwy;

FYI, http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6587

 > we need to strip arguments and parentheses from DW_AT_name for 
language_java and then I
> expect it will start working.

FYI, a hack stripping that stuff didn't appear to make it work, and 
doesn't fix the underlying problem of corrupt debug info.  As with Ada, 
Java needs a better symbol search method.  What happened to that thread?

   I want to finish my current massacre of
> dwarf2read before I go playing with that, though - it's always been on
> a lower burner.

Um, you appear to be spreading yourself very very thin here :-(

Andrew

Re: [commit/java] Work around java/1565 in jmisc.exp

[Daniel Jacobowitz]

On Tue, Feb 24, 2004 at 11:42:43AM -0500, Andrew Cagney wrote:
> > we need to strip arguments and parentheses from DW_AT_name for 
> language_java and then I
> >expect it will start working.
> 
> FYI, a hack stripping that stuff didn't appear to make it work, and 
> doesn't fix the underlying problem of corrupt debug info.  As with Ada, 
> Java needs a better symbol search method.  What happened to that thread?

I imagine it got queued, just like everything else.

>   I want to finish my current massacre of
> >dwarf2read before I go playing with that, though - it's always been on
> >a lower burner.
> 
> Um, you appear to be spreading yourself very very thin here :-(

That's why I'm working from oldest to newest :)  Two of the oldest
messages in my gdb-patches TODO folder rely on the intercu work I've
done this week.  After that I think I may be stuck changing my TOOD
algorithm, or going back to working on multiple address breakpoints -
which means, going back to symbol searching.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [commit/java] Work around java/1565 in jmisc.exp

[David Carlton]

On Tue, 24 Feb 2004 11:12:11 -0500, Andrew Cagney <cagney@gnu.org> said:

> This patch modifies jmisc.exp to work around the very very long
> standing bug java/1565 (just that no one thought to report it?).

It's bug java/1164 (with a corresponding GDB PR).  As Michael pointed
out, one of the other Java tests already does what you want.  It would
be good for this test to be KFAILed, though, given that you're the
second person to have proposed that change in the last two or three
months.

David Carlton
carlton@kealia.com

Re: [commit/java] Work around java/1565 in jmisc.exp

[Andrew Cagney]

> On Tue, 24 Feb 2004 11:12:11 -0500, Andrew Cagney <cagney@gnu.org> said:
> 
> 
>>> This patch modifies jmisc.exp to work around the very very long
>>> standing bug java/1565 (just that no one thought to report it?).
> 
> 
> It's bug java/1164 (with a corresponding GDB PR).  As Michael pointed
> out, one of the other Java tests already does what you want.  It would
> be good for this test to be KFAILed, though, given that you're the
> second person to have proposed that change in the last two or three
> months.

No, "jmain.exp" does what I want - test for three potential bugs.  The 
old "jmisc.exp" really isn't doing anything other than causing cascading 
failures :-(

BTW, I looked at KFAILing "jmisc.exp" but dropped the idea like a hot 
potato - the FAIL is burried deep in the depths of "runto" :-(

Andrew

Re: [commit/java] Work around java/1565 in jmisc.exp

[Daniel Jacobowitz]

On Tue, Feb 24, 2004 at 12:47:13PM -0500, Andrew Cagney wrote:
> >On Tue, 24 Feb 2004 11:12:11 -0500, Andrew Cagney <cagney@gnu.org> said:
> >
> >
> >>>This patch modifies jmisc.exp to work around the very very long
> >>>standing bug java/1565 (just that no one thought to report it?).
> >
> >
> >It's bug java/1164 (with a corresponding GDB PR).  As Michael pointed
> >out, one of the other Java tests already does what you want.  It would
> >be good for this test to be KFAILed, though, given that you're the
> >second person to have proposed that change in the last two or three
> >months.
> 
> No, "jmain.exp" does what I want - test for three potential bugs.  The 
> old "jmisc.exp" really isn't doing anything other than causing cascading 
> failures :-(
> 
> BTW, I looked at KFAILing "jmisc.exp" but dropped the idea like a hot 
> potato - the FAIL is burried deep in the depths of "runto" :-(

All you'd need then would be a setup_kfail, no?  Sure, you don't get to
match the explicit failing pattern that way...

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [commit/java] Work around java/1565 in jmisc.exp

[Andrew Cagney]

>>BTW, I looked at KFAILing "jmisc.exp" but dropped the idea like a hot 
>>> potato - the FAIL is burried deep in the depths of "runto" :-(
> 
> 
> All you'd need then would be a setup_kfail, no?

And randomly kfail any any of?
   fail "setting breakpoint at $function (timeout)"
   fail "running to $function in runto"
   fail "running to $function in runto (timeout)"
er, no.

>  Sure, you don't get to
> match the explicit failing pattern that way...

Andrew

Re: [commit/java] Work around java/1565 in jmisc.exp

[David Carlton]

On Tue, 24 Feb 2004 12:47:13 -0500, Andrew Cagney <cagney@gnu.org> said:
>> On Tue, 24 Feb 2004 11:12:11 -0500, Andrew Cagney <cagney@gnu.org> said:

>>>> This patch modifies jmisc.exp to work around the very very long
>>>> standing bug java/1565 (just that no one thought to report it?).

>> It's bug java/1164 (with a corresponding GDB PR).

That should say "GCC PR", of course.

>> As Michael pointed out, one of the other Java tests already does
>> what you want.  It would be good for this test to be KFAILed,
>> though, given that you're the second person to have proposed that
>> change in the last two or three months.

> No, "jmain.exp" does what I want - test for three potential bugs.
> The old "jmisc.exp" really isn't doing anything other than causing
> cascading failures :-(

I posted the above before I saw your jmain.exp; I agree that breaking
up the "can we break on main" functionality into a separate test is a
much better idea than the current setup.

David Carlton
carlton@kealia.com

Re: [commit/java] Work around java/1565 in jmisc.exp

[Michael Elizabeth Chastain]

This is redundant.  jmisc2.exp already does exactly what you want.

  % diff jmisc.exp jmisc2.exp
  1c1
  < # Copyright 2000 Free Software Foundation, Inc.
  ---
  > # Copyright 2002 Free Software Foundation, Inc.
  69,70d68
  <     runto ${testfile}.main
  <
  77a76,77
  >     runto ${testfile}.main(java.lang.String\[\])
  >

Also, did Anthony Green approve your patch before you commited it?

Michael C

===

2004-02-24  Andrew Cagney  <cagney@redhat.com>

	* gdb.java/jmisc.exp: Use the fully quallified name of "main" when
	setting a breakpoint.  Work around PR java/1565.

Re: [commit/java] Work around java/1565 in jmisc.exp

[Andrew Cagney]

> This is redundant.  jmisc2.exp already does exactly what you want.

Hmm, my just posted jmain.exp makes having both jmisc.exp and jmisc2.exp 
completly redundant.  The old jmisc.exp just cascaded failures.

Anthony, with jmain.exp, should jmisc2.exp be deleted?

Andrew