Re: [testsuite/c++] Why is there one fail in demangle.exp?

Re: [testsuite/c++] Why is there one fail in demangle.exp?

[Michael Elizabeth Chastain]

The comment above the test is:

	## 1999-04-19: "Fix from Dale Hawkins".  Shouldn't segfault.

This implies to me that an old version of gdb used to segfault on
that input.

Note that the release date of gdb-4.18 was 1999-04-09.  So I built one:

  mec@duracef:/tmp/mec/build$ gdb/gdb
  GNU gdb 4.18
  Copyright 1998 Free Software Foundation, Inc.
  GDB is free software, covered by the GNU General Public License, and you are
  welcome to change it and/or distribute copies of it under certain conditions.
  Type "show copying" to see the conditions.
  There is absolutely no warranty for GDB.  Type "show warranty" for details.
  This GDB was configured as "i686-pc-linux-gnu".
  (gdb) maintenance demangle __thunk_64__0RL__list__Q29CosNaming20_proxy_NamingConMtextUlRPt25_CORBA_Unbounded_Sequence1ZQ29CosNaming7BindingRPQ29CosNaming15BindingIterator
  Segmentation fault (core dumped)

> I can see that being wrong (_0RL__list as a function name is a little
> suspicious).  But the test expects "Can't demangle".

I don't know what the right answer is.  It looks like the intent of the
test is to check whether the demangler crashes on this input.  It's
quite possible that the input is purposely malformed.

I'm inclined to approve both "Can't demangle" and any reasonable
demangling of the input as PASS, with a comment to the effect that
gdb 4.18 segfaults on this input.

Michael C

Re: [testsuite/c++] Why is there one fail in demangle.exp?

[Daniel Jacobowitz]

On Tue, Dec 04, 2001 at 03:57:30AM -0600, Michael Elizabeth Chastain wrote:
> The comment above the test is:
> 
> 	## 1999-04-19: "Fix from Dale Hawkins".  Shouldn't segfault.
> 
> This implies to me that an old version of gdb used to segfault on
> that input.
> 
> Note that the release date of gdb-4.18 was 1999-04-09.  So I built one:
> 
>   mec@duracef:/tmp/mec/build$ gdb/gdb
>   GNU gdb 4.18
>   Copyright 1998 Free Software Foundation, Inc.
>   GDB is free software, covered by the GNU General Public License, and you are
>   welcome to change it and/or distribute copies of it under certain conditions.
>   Type "show copying" to see the conditions.
>   There is absolutely no warranty for GDB.  Type "show warranty" for details.
>   This GDB was configured as "i686-pc-linux-gnu".
>   (gdb) maintenance demangle __thunk_64__0RL__list__Q29CosNaming20_proxy_NamingConMtextUlRPt25_CORBA_Unbounded_Sequence1ZQ29CosNaming7BindingRPQ29CosNaming15BindingIterator
>   Segmentation fault (core dumped)
> 
> > I can see that being wrong (_0RL__list as a function name is a little
> > suspicious).  But the test expects "Can't demangle".
> 
> I don't know what the right answer is.  It looks like the intent of the
> test is to check whether the demangler crashes on this input.  It's
> quite possible that the input is purposely malformed.
> 
> I'm inclined to approve both "Can't demangle" and any reasonable
> demangling of the input as PASS, with a comment to the effect that
> gdb 4.18 segfaults on this input.

Dan Berlin commented in a separate message that the string has an
ambiguous meaning in the v2 mangling grammar.  I don't believe that the
FAIL: serves any purpose, though, so I'd like to make it pass but put a
comment to the effect that this isn't 100% accurate.  Patch later.


-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

[testsuite/c++] Why is there one fail in demangle.exp?

[Daniel Jacobowitz]

The test is:

drow@nevyn:~% c++filt                                                                 
__thunk_64__0RL__list__Q29CosNaming20_proxy_NamingContextUlRPt25_CORBA_Unbounded_Sequence1ZQ29CosNaming7BindingRPQ29CosNaming15BindingIterator
virtual function thunk (delta:-64) for CosNaming::_proxy_NamingContext::_0RL__list(unsigned long, _CORBA_Unbounded_Sequence<CosNaming::Binding> *&, CosNaming::BindingIterator *&)

I can see that being wrong (_0RL__list as a function name is a little
suspicious).  But the test expects "Can't demangle".

Anyone know if it is right or wrong?  I'll add a clarifying comment.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer