Re: Fix Java PR gdb/1322 (was Re: [rfa/doc] Document pr gdb/1322 in PROBLEMS)

Re: Fix Java PR gdb/1322 (was Re: [rfa/doc] Document pr gdb/1322 in PROBLEMS)

[Michael Elizabeth Chastain]

drow> Could someone test that it fixes the Java problems on HEAD?

It works for me.  It passes the "continue to exit" test
in gdb.java/jmisc1.exp.  (That's the specific test for this bug).

I didn't do a full before-and-after with this patch, my system is busy
today.  But it looks like it could not possibly cause any regressions,
It's really just changing the order of two tests that already happen:
an fde==NULL test and the test inside the SECT_OFF_TEXT macro.

Recommended for approval in gdb HEAD.

Michael C

	* dwarf2-frame.c (dwarf2_frame_find_fde): Check whether any FDEs are
	available before calling SECT_OFF_TEXT.

[rfa/doc] Document pr gdb/1322 in PROBLEMS

[Michael Elizabeth Chastain]

This patch adds pr gdb/1322 to PROBLEMS.  It describes the problem
and offers a workaround and a link to a patch.

OK to apply to gdb HEAD?

OK to apply to gdb gdb_6_0-branch?

Michael C

2003-08-18  Michael Chastain  <mec@shout.net>

	* PROBLEMS: Document pr gdb/1322, the Java anonymous
	objfile bug.

Index: PROBLEMS
===================================================================
RCS file: /cvs/src/src/gdb/PROBLEMS,v
retrieving revision 1.14.30.2
diff -c -3 -p -r1.14.30.2 PROBLEMS
*** PROBLEMS	26 Jun 2003 16:25:53 -0000	1.14.30.2
--- PROBLEMS	18 Aug 2003 18:28:19 -0000
*************** gcc 3.x generates these multiple object 
*** 19,21 ****
--- 19,35 ----
  implement virtual base classes.  gcc 2.x generated just one object code
  function with a hidden parameter, but gcc 3.x conforms to a multi-vendor
  ABI for C++ which requires multiple object code functions.
+ 
+ gdb/1322: "internal-error: sect_index_text not initialized" after printing a java type
+ 
+ When gdb debugs a Java program which uses shared libraries, it can
+ generate this internal error.
+ 
+ If you get this error when debugging Java, you can try working around
+ the problem by relinking your program with no shared libraries at all
+ (even the standard C library).  With the GNU linker, use the "-static"
+ argument to do this.
+ 
+ You can also try this patch for gdb:
+ 
+   http://sources.redhat.com/ml/gdb-prs/2003-q3/msg00190.html

Fix Java PR gdb/1322 (was Re: [rfa/doc] Document pr gdb/1322 in PROBLEMS)

[Daniel Jacobowitz]

On Mon, Aug 18, 2003 at 02:32:20PM -0400, Michael Elizabeth Chastain wrote:
> + 
> + gdb/1322: "internal-error: sect_index_text not initialized" after printing a java type
> + 
> + When gdb debugs a Java program which uses shared libraries, it can
> + generate this internal error.
> + 
> + If you get this error when debugging Java, you can try working around
> + the problem by relinking your program with no shared libraries at all
> + (even the standard C library).  With the GNU linker, use the "-static"
> + argument to do this.
> + 
> + You can also try this patch for gdb:
> + 
> +   http://sources.redhat.com/ml/gdb-prs/2003-q3/msg00190.html

How about we fix it instead?  This patch depends on Mark K's
per-objfile-data patch, but only because it touches the same area, I
think.

Dwarf maintainers, is this OK?  Could someone test that it fixes the
Java problems on HEAD?

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

2003-08-24  Daniel Jacobowitz  <drow@mvista.com>

	PR gdb/1322
	* dwarf2-frame.c (dwarf2_frame_find_fde): Check whether any FDEs are
	available before calling SECT_OFF_TEXT.

Index: dwarf2-frame.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2-frame.c,v
retrieving revision 1.11
diff -u -p -r1.11 dwarf2-frame.c
--- dwarf2-frame.c	21 Aug 2003 22:35:33 -0000	1.11
+++ dwarf2-frame.c	24 Aug 2003 16:52:24 -0000
@@ -1024,9 +1024,12 @@ dwarf2_frame_find_fde (CORE_ADDR *pc)
       struct dwarf2_fde *fde;
       CORE_ADDR offset;
 
-      offset = ANOFFSET (objfile->section_offsets, SECT_OFF_TEXT (objfile));
-      
       fde = objfile_data (objfile, dwarf2_frame_data);
+      if (fde == NULL)
+	continue;
+
+      offset = ANOFFSET (objfile->section_offsets, SECT_OFF_TEXT (objfile));
+
       while (fde)
 	{
 	  if (*pc >= fde->initial_location + offset

Re: Fix Java PR gdb/1322 (was Re: [rfa/doc] Document pr gdb/1322 in PROBLEMS)

[Mark Kettenis]

Daniel Jacobowitz <drow@mvista.com> writes:

> On Mon, Aug 18, 2003 at 02:32:20PM -0400, Michael Elizabeth Chastain wrote:
> > + 
> > + gdb/1322: "internal-error: sect_index_text not initialized" after printing a java type
> > + 
> > + When gdb debugs a Java program which uses shared libraries, it can
> > + generate this internal error.
> > + 
> > + If you get this error when debugging Java, you can try working around
> > + the problem by relinking your program with no shared libraries at all
> > + (even the standard C library).  With the GNU linker, use the "-static"
> > + argument to do this.
> > + 
> > + You can also try this patch for gdb:
> > + 
> > +   http://sources.redhat.com/ml/gdb-prs/2003-q3/msg00190.html
> 
> How about we fix it instead?  This patch depends on Mark K's
> per-objfile-data patch, but only because it touches the same area, I
> think.
> 
> Dwarf maintainers, is this OK?  Could someone test that it fixes the
> Java problems on HEAD?

I'm not exactly a DWARF maintainer, but since I wrote the code in that
file I assume I can approve your patch :-).  I'll do so if you'll add a

   gdb_assert (objfile->section_offsets);

just before the

   offset = ANOFFSET (...

to check the assumption you're making that the section offsets will be
available when we have DWARF2 CFI.

Oh, and change the PR first line of the ChangeLog into java/1322 such
that GNATS will notice your patch.  I believe it will ignore it if you
use gdb/1322.

Mark

Re: Fix Java PR gdb/1322 (was Re: [rfa/doc] Document pr gdb/1322 in PROBLEMS)

[Daniel Jacobowitz]

On Sun, Aug 24, 2003 at 09:42:16PM +0200, Mark Kettenis wrote:
> Daniel Jacobowitz <drow@mvista.com> writes:
> 
> > On Mon, Aug 18, 2003 at 02:32:20PM -0400, Michael Elizabeth Chastain wrote:
> > > + 
> > > + gdb/1322: "internal-error: sect_index_text not initialized" after printing a java type
> > > + 
> > > + When gdb debugs a Java program which uses shared libraries, it can
> > > + generate this internal error.
> > > + 
> > > + If you get this error when debugging Java, you can try working around
> > > + the problem by relinking your program with no shared libraries at all
> > > + (even the standard C library).  With the GNU linker, use the "-static"
> > > + argument to do this.
> > > + 
> > > + You can also try this patch for gdb:
> > > + 
> > > +   http://sources.redhat.com/ml/gdb-prs/2003-q3/msg00190.html
> > 
> > How about we fix it instead?  This patch depends on Mark K's
> > per-objfile-data patch, but only because it touches the same area, I
> > think.
> > 
> > Dwarf maintainers, is this OK?  Could someone test that it fixes the
> > Java problems on HEAD?
> 
> I'm not exactly a DWARF maintainer, but since I wrote the code in that
> file I assume I can approve your patch :-).  I'll do so if you'll add a
> 
>    gdb_assert (objfile->section_offsets);
> 
> just before the
> 
>    offset = ANOFFSET (...
> 
> to check the assumption you're making that the section offsets will be
> available when we have DWARF2 CFI.
> 
> Oh, and change the PR first line of the ChangeLog into java/1322 such
> that GNATS will notice your patch.  I believe it will ignore it if you
> use gdb/1322.

Thanks.  I've checked this in on HEAD.  In a couple of days, if all
goes well, I'll move it to the branch.

I removed it from PROBLEMS too, since hopefully it's no longer a
problem.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer