GCC 3.0.3: Bugs to Fix

GCC 3.0.3: Bugs to Fix

[Mark Mitchell]

Here are some bugs that it would be nice, although not imperative,
to fix before 3.0.3:

PR4299:

  A crash when compiling the Linux kernel on an x86.  It would be
  very good if we could fix this bug; making GCC 3.0.x a usable
  compiler for buliding the Linux kernel is important.

PR3720:

  A crash in the C++ run-time library on simple formatted input.

PR3048

  A C++ name lookup problem.  I will try to fix this one.

PR3381

  A C++ template instantiation problem.

Are there volunteers avilable for PRs 4299, 3048, and 3381?

Again, if there are other PRs that represent regressions from
previous 3.0.x releases, please let me know ASAP.

Thanks,

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: GCC 3.0.3: Bugs to Fix

[Richard Henderson]

On Thu, Nov 29, 2001 at 01:53:18PM -0800, Mark Mitchell wrote:
> PR4299:
> 
>   A crash when compiling the Linux kernel on an x86.  It would be
>   very good if we could fix this bug; making GCC 3.0.x a usable
>   compiler for buliding the Linux kernel is important.

I'll take this one.


r~

Re: GCC 3.0.3: Bugs to Fix

[Mark Mitchell]

--On Thursday, November 29, 2001 03:12:05 PM -0800 Richard Henderson 
<rth@redhat.com> wrote:

> On Thu, Nov 29, 2001 at 01:53:18PM -0800, Mark Mitchell wrote:
>> PR4299:
>>
>>   A crash when compiling the Linux kernel on an x86.  It would be
>>   very good if we could fix this bug; making GCC 3.0.x a usable
>>   compiler for buliding the Linux kernel is important.
>
> I'll take this one.

Thanks!

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: GCC 3.0.3: Bugs to Fix

[Mark Mitchell]

--On Thursday, November 29, 2001 03:12:05 PM -0800 Richard Henderson 
<rth@redhat.com> wrote:

> On Thu, Nov 29, 2001 at 01:53:18PM -0800, Mark Mitchell wrote:
>> PR4299:
>>
>>   A crash when compiling the Linux kernel on an x86.  It would be
>>   very good if we could fix this bug; making GCC 3.0.x a usable
>>   compiler for buliding the Linux kernel is important.
>
> I'll take this one.

Thanks!

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: GCC 3.0.3: Bugs to Fix

[Richard Henderson]

On Thu, Nov 29, 2001 at 01:53:18PM -0800, Mark Mitchell wrote:
> PR4299:
> 
>   A crash when compiling the Linux kernel on an x86.  It would be
>   very good if we could fix this bug; making GCC 3.0.x a usable
>   compiler for buliding the Linux kernel is important.

I'll take this one.


r~

Re: GCC 3.0.3: Bugs to Fix

[Nathan Sidwell]

Mark Mitchell wrote:

> PR3381
> 
>   A C++ template instantiation problem.
I'll look at this.

nathan

-- 
Dr Nathan Sidwell   ::   http://www.codesourcery.com   ::   CodeSourcery LLC
         'But that's a lie.' - 'Yes it is. What's your point?'
nathan@codesourcery.com : http://www.cs.bris.ac.uk/~nathan/ : nathan@acm.org

Re: GCC 3.0.3: Bugs to Fix

[Nathan Sidwell]

Mark Mitchell wrote:

> PR3381
> 
>   A C++ template instantiation problem.
I'll look at this.

nathan

-- 
Dr Nathan Sidwell   ::   http://www.codesourcery.com   ::   CodeSourcery LLC
         'But that's a lie.' - 'Yes it is. What's your point?'
nathan@codesourcery.com : http://www.cs.bris.ac.uk/~nathan/ : nathan@acm.org

Re: GCC 3.0.3: Bugs to Fix

[Franz Sirl]

On Thursday 29 November 2001 22:53, Mark Mitchell wrote:
> Here are some bugs that it would be nice, although not imperative,
> to fix before 3.0.3:

What about PR3993? Another KDE2 blocking bug...

Franz.

Re: GCC 3.0.3: Bugs to Fix

[Franz Sirl]

On Thursday 29 November 2001 22:53, Mark Mitchell wrote:
> Here are some bugs that it would be nice, although not imperative,
> to fix before 3.0.3:

What about PR3993? Another KDE2 blocking bug...

Franz.

Re: GCC 3.0.3: Bugs to Fix

[Nathan Sidwell]

Franz Sirl wrote:
> 
> On Thursday 29 November 2001 22:53, Mark Mitchell wrote:
> > Here are some bugs that it would be nice, although not imperative,
> > to fix before 3.0.3:
> 
> What about PR3993? Another KDE2 blocking bug...
Is there a reason that dlopen ("foo", RTLD_NOW | RTLD_GLOBAL)
is unacceptable?

The ABI was designed on the assumption that it is possible at link
or load time to collapse objects with the same name to a single address,
and thus use address comparisons to determine equality. Loading without
RTLD_GLOBAL break that.

Presumably you'll also observe that it's possible for the 'same function'
has two different addresses, one in the library and one not in the library.

[5.10] says 'Two pointers ... compare equal if and only if ... both point
to the same object or function ...' 
Ooh, that's not actually the same as
	'Two pointers compare unequal if and only if they point to different
objects or functions'
(but I always thought that was true, defect anyone?)

Anyhow, from the ABI spec point of view, you've fallen outside the ABI
by disjointing the namespace.  To make it work without RTLD_GLOBAL would
pessimize type comparisons because we'd have to compare strings rather than
pointers to strings.  What's worse is that it's the 'unequal to' comparison
that could not be optimized, and IMO, dynamic cast & catch have more
not-equal cases than equal cases.

If we HAD to make this work at ANY cost, then I'd change the ABI to secrete
a hash of the typename that could be compared in O(0), rather than O(length).
But see - that's an ABI change.

nathan

-- 
Dr Nathan Sidwell   ::   http://www.codesourcery.com   ::   CodeSourcery LLC
         'But that's a lie.' - 'Yes it is. What's your point?'
nathan@codesourcery.com : http://www.cs.bris.ac.uk/~nathan/ : nathan@acm.org

Re: GCC 3.0.3: Bugs to Fix

[Andreas Schwab]

Nathan Sidwell <nathan@codesourcery.com> writes:

|> [5.10] says 'Two pointers ... compare equal if and only if ... both point
                                               ^^^^^^^^^^^^^^
|> to the same object or function ...' 
|> Ooh, that's not actually the same as
|> 	'Two pointers compare unequal if and only if they point to different
|> objects or functions'
|> (but I always thought that was true, defect anyone?)

"if and only if" describes an equivalence, thus the negation holds as
well.

Andreas.

-- 
Andreas Schwab                                  "And now for something
Andreas.Schwab@suse.de				completely different."
SuSE Labs, SuSE GmbH, Schanzäckerstr. 10, D-90443 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5

Re: GCC 3.0.3: Bugs to Fix

[Nathan Sidwell]

Andreas Schwab wrote:

> "if and only if" describes an equivalence, thus the negation holds as
> well.
doh! I'd forgotten the precise meaning of 'iff', thanks

nathan

-- 
Dr Nathan Sidwell :: Computer Science Department :: Bristol University
           The voices in my head told me to say this
nathan@acm.org  http://www.cs.bris.ac.uk/~nathan/  nathan@cs.bris.ac.uk

Re: GCC 3.0.3: Bugs to Fix

[Mark Mitchell]

--On Monday, December 03, 2001 09:56:37 AM +0000 Nathan Sidwell 
<nathan@codesourcery.com> wrote:

> Franz Sirl wrote:
>>
>> On Thursday 29 November 2001 22:53, Mark Mitchell wrote:
>> > Here are some bugs that it would be nice, although not imperative,
>> > to fix before 3.0.3:
>>
>> What about PR3993? Another KDE2 blocking bug...
> Is there a reason that dlopen ("foo", RTLD_NOW | RTLD_GLOBAL)
> is unacceptable?
>
> The ABI was designed on the assumption that it is possible at link
> or load time to collapse objects with the same name to a single address,
> and thus use address comparisons to determine equality. Loading without
> RTLD_GLOBAL break that.

When the ABI was designed, the committee made a conscious decision not
to support any kind of dynamic linking that did not merge names.  I do
not think we should revisit this at this point.

Therefore, if that is what this bug concerns, the bug should simply
be closed with a note explaining the issue.  For bonus points, an
entry should be added to the FAQ.

Thanks,

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: GCC 3.0.3: Bugs to Fix

[Joe Buck]

Franz Sirl wrote:
> >> What about PR3993? Another KDE2 blocking bug...

<nathan@codesourcery.com> wrote:
> > Is there a reason that dlopen ("foo", RTLD_NOW | RTLD_GLOBAL)
> > is unacceptable?
> >
> > The ABI was designed on the assumption that it is possible at link
> > or load time to collapse objects with the same name to a single address,
> > and thus use address comparisons to determine equality. Loading without
> > RTLD_GLOBAL break that.

Mark Mitchell:
> When the ABI was designed, the committee made a conscious decision not
> to support any kind of dynamic linking that did not merge names.  I do
> not think we should revisit this at this point.
> 
> Therefore, if that is what this bug concerns, the bug should simply
> be closed with a note explaining the issue.  For bonus points, an
> entry should be added to the FAQ.

But someone claimed that this problem doesn't exist in the 3.1 branch.
So how can that be right, if this is right?  Or am I confused again?

Joe

Re: GCC 3.0.3: Bugs to Fix

[Mark Mitchell]

> Mark Mitchell:
>> When the ABI was designed, the committee made a conscious decision not
>> to support any kind of dynamic linking that did not merge names.  I do
>> not think we should revisit this at this point.
>
> But someone claimed that this problem doesn't exist in the 3.1 branch.
> So how can that be right, if this is right?  Or am I confused again?

I don't know.  I hadn't seen that follow-up.  It could be random chance;
something else happens to make the program work, even though the problem
is still there.

I'm almost certain that nothing changed in the library that would make
this work.

Note that it *may* work on systems that do not have weak symbols,
where we already fall back to strcmp for comparing type info.  There
may still be other similar problems though.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: GCC 3.0.3: Bugs to Fix

[Florian Kainz]

Mark Mitchell wrote:

> --On Monday, December 03, 2001 09:56:37 AM +0000 Nathan Sidwell
> <nathan@codesourcery.com> wrote:
>
> > Franz Sirl wrote:
> >>
> >> On Thursday 29 November 2001 22:53, Mark Mitchell wrote:
> >> > Here are some bugs that it would be nice, although not imperative,
> >> > to fix before 3.0.3:
> >>
> >> What about PR3993? Another KDE2 blocking bug...
> > Is there a reason that dlopen ("foo", RTLD_NOW | RTLD_GLOBAL)
> > is unacceptable?
> >
> > The ABI was designed on the assumption that it is possible at link
> > or load time to collapse objects with the same name to a single address,
> > and thus use address comparisons to determine equality. Loading without
> > RTLD_GLOBAL break that.
>
> When the ABI was designed, the committee made a conscious decision not
> to support any kind of dynamic linking that did not merge names.  I do
> not think we should revisit this at this point.
>
> Therefore, if that is what this bug concerns, the bug should simply
> be closed with a note explaining the issue.  For bonus points, an
> entry should be added to the FAQ.

But

dlopen ("foo", RTLD_NOW | RTLD_GLOBAL)

should work, right?  Well, it doesn't;  dynamic_cast still
returns 0 when it shouldn't, at least with gcc-3.0.2 and
gcc-3.0.1.

Bug #4292 has already been closed, so I am going to re-submit
it, with an updated example that uses RTLD_GLOBAL.

Florian Kainz
Industrial Light + Magic

Re: GCC 3.0.3: Bugs to Fix

[Mark Mitchell]

> But
>
> dlopen ("foo", RTLD_NOW | RTLD_GLOBAL)
>
> should work, right?  Well, it doesn't;  dynamic_cast still
> returns 0 when it shouldn't, at least with gcc-3.0.2 and
> gcc-3.0.1.

Yes, that should work.

> Bug #4292 has already been closed, so I am going to re-submit
> it, with an updated example that uses RTLD_GLOBAL.

Thanks.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: GCC 3.0.3: Bugs to Fix

[Alexandre Oliva]

On Dec  3, 2001, Florian Kainz <kainz@ilm.com> wrote:

> Bug #4292 has already been closed, so I am going to re-submit
> it, with an updated example that uses RTLD_GLOBAL.

I noted it was a dup of 3993, and noted in 3993 that 4292 had been
closed as a dup.  This should be enough.

IMO, it would be best to add data to 3993 instead of creating more
PRs.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist    *Please* write to mailing lists, not to me

Re: GCC 3.0.3: Bugs to Fix

[Alexandre Oliva]

On Dec  3, 2001, Nathan Sidwell <nathan@codesourcery.com> wrote:

> Franz Sirl wrote:
>> 
>> On Thursday 29 November 2001 22:53, Mark Mitchell wrote:
>> > Here are some bugs that it would be nice, although not imperative,
>> > to fix before 3.0.3:
>> 
>> What about PR3993? Another KDE2 blocking bug...
> Is there a reason that dlopen ("foo", RTLD_NOW | RTLD_GLOBAL)
> is unacceptable?

FWIW, I suggested someone @ilm.com who reported a duplicate to try
just that, and got a reply that it wasn't enough.  I couldn't look
into it any further, though.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist    *Please* write to mailing lists, not to me

Re: GCC 3.0.3: Bugs to Fix

[Mark Mitchell]

--On Friday, November 30, 2001 11:23:51 PM +0100 Franz Sirl 
<Franz.Sirl-kernel@lauterbach.com> wrote:

> On Thursday 29 November 2001 22:53, Mark Mitchell wrote:
>> Here are some bugs that it would be nice, although not imperative,
>> to fix before 3.0.3:
>
> What about PR3993? Another KDE2 blocking bug...

I'll put it on my list, but you should have told me when I made my
initial call for important PRs a while back.

I think that KDE is working on the 3.1 branch, BTW, so that we will
not have this mess in the 3.1 series.

Thanks,

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com

Re: GCC 3.0.3: Bugs to Fix

[Franz Sirl]

On Monday 03 December 2001 17:43, Mark Mitchell wrote:
> --On Friday, November 30, 2001 11:23:51 PM +0100 Franz Sirl
>
> <Franz.Sirl-kernel@lauterbach.com> wrote:
> > On Thursday 29 November 2001 22:53, Mark Mitchell wrote:
> >> Here are some bugs that it would be nice, although not imperative,
> >> to fix before 3.0.3:
> >
> > What about PR3993? Another KDE2 blocking bug...
>
> I'll put it on my list, but you should have told me when I made my
> initial call for important PRs a while back.

Sorry, I thought I did this earlier, but I may misremember that. Currently my 
spare time is a bit limited and I have to queue stuff up for the wwekend :-(.

> I think that KDE is working on the 3.1 branch, BTW, so that we will
> not have this mess in the 3.1 series.

Hopefully :-)

Franz.

GCC 3.0.3: Bugs to Fix

[Mark Mitchell]

Here are some bugs that it would be nice, although not imperative,
to fix before 3.0.3:

PR4299:

  A crash when compiling the Linux kernel on an x86.  It would be
  very good if we could fix this bug; making GCC 3.0.x a usable
  compiler for buliding the Linux kernel is important.

PR3720:

  A crash in the C++ run-time library on simple formatted input.

PR3048

  A C++ name lookup problem.  I will try to fix this one.

PR3381

  A C++ template instantiation problem.

Are there volunteers avilable for PRs 4299, 3048, and 3381?

Again, if there are other PRs that represent regressions from
previous 3.0.x releases, please let me know ASAP.

Thanks,

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com