[Monotone-devel] Re: [Monotone-debian] Bug#455646: FTBFS with GCC 4.3: missing #includes

[Zack Weinberg]

On Dec 10, 2007 7:41 PM, Martin Michlmayr <address@hidden> wrote:
> Package: monotone
> Version: 0.37-4
> Usertags: ftbfs-gcc-4.3
>
> Your package fails to build with GCC 4.3.  Version 4.3 has not been
> released yet but I'm building with a snapshot in order to find errors
> and give people an advance warning.

With my upstream hat on, I've committed *partial* fixes for this
problem.  (These will appear in 0.39, due out in probably a month.)  I
am cc:ing you back and monotone-devel as well, because there are still
some problems that I need help with:

1) "declaration changes meaning of symbol" errors.  I see two of these:

roster.cc:3503: error: declaration of 'virtual
void<unnamed>::a_scalar::set(const revision_id&,
<unnamed>::scalar_val, const std::set<revision<hexenc<id> >,
std::less<revision<hexenc<id> > >, std::allocator<revision<hexenc<id>
> > >&, roster_t&, marking_map&)'
bits/stl_set.h:95: error: changes meaning of 'set' from 'class
std::set<revision<hexenc<id> >, std::less<revision<hexenc<id> > >,
std::allocator<revision<hexenc<id> > > >'

and

boost/circular_buffer_base.hpp:502: error: declaration of 'typedef
class 
boost::reverse_iterator<boost::cb_details::cb_iterator<boost::circular_buffer<T,
Alloc>, boost::cb_details::cb_nonconst_traits<Alloc> > >
boost::circular_buffer<T, Alloc>::reverse_iterator'
boost/iterator/reverse_iterator.hpp:23: error: changes meaning of
'reverse_iterator' from 'class
boost::reverse_iterator<boost::cb_details::cb_iterator<boost::circular_buffer<T,
Alloc>, boost::cb_details::cb_nonconst_traits<Alloc> > >'

(note: boost::circular_buffer is an unofficial experimental boost
module that we picked up a long time ago - don't go looking for it in
the official boost packages.)  I have never seen this error before.
My immediate reaction is that I have no idea what's going on in boost
(but who does?) and that the code in roster.cc ought to be valid - but
I am by no means a C++ expert.  I was not able to construct a cut-down
version of the roster.cc code that provokes the same error; perhaps
I'll get around to delta-minimizing one from the real code, but I
doubt I'll have time soon.

2) sqlite/vdbeaux.c:2212: internal compiler error: in
get_addr_dereference_operands, at tree-ssa-operands.c:1746

GCC bug.  Should be reproducible with the official Debian sqlite
package.  I do not have time to file a GCC bug report.

3) backward/auto_ptr.h: In static member function 'static bool
Botan::DL_Group::generate_dsa_primes(Botan::BigInt&
, Botan::BigInt&, Botan::u32bit, Botan::u32bit, const
Botan::MemoryRegion<unsigned char>&)':
backward/auto_ptr.h:177: warning:
'hash.std::auto_ptr<Botan::HashFunction>::_M_ptr' is used
uninitialized in this function
botan/dsa_gen.cpp:57: note:
'hash.std::auto_ptr<Botan::HashFunction>::_M_ptr' was declared here

Looks spurious; the variable in question is declared with a
constructor expression.  (I am ignorant of what is going on inside
auto_ptr though.)

Note also to monotone-devel: some of the changes were in netxx.
That's the one we basically own at this point because its upstream
died, right?

zw