Re: 3.1.52 fails to build in hppa/linux

[Jaroslav Hajek]

On Sun, Feb 22, 2009 at 6:42 PM, Marco Atzeri <address@hidden> wrote:
>
>
>
>
> --- Dom 22/2/09, Rafael Laboissiere  ha scritto:
>
>> Da: Rafael Laboissiere
>> Oggetto: Re: 3.1.52 fails to build in hppa/linux
>> A: address@hidden
>> Data: Domenica 22 febbraio 2009, 17:21
>> * Jaroslav Hajek  [2009-02-22
>> 16:41]:
>>
>> > On Sun, Feb 22, 2009 at 4:08 PM, Rafael Laboissiere
>> <> wrote:
>> > > A bug report has been filed against the octave3.1
>> in Debian regarding the
>> > > failure to build the package on the hppa
>> architecture.  You will find the
>> > > complete bug report in
>> http://bugs.debian.org/516588, including a link to
>> > > the full build log.  In sum, it fails with errors
>> like this:
>> > >
>> > > g++ -c  -fPIC -I. -I.. -I../liboctave -I../src
>> -I../libcruft/misc -DHAVE_CONFIG_H -I/usr/include/freetype2
>> -Wall -W -Wshadow -Wold-style-cast -O2 -g oct-inttypes.cc -o
>> pic/oct-inttypes.o
>> > > oct-inttypes.cc:511: error: ambiguous template
>> specialization 'mop<octave_int_cmp_op::lt>'
>> for 'bool octave_int_cmp_op::mop(double, int64_t)'
>> > >
>> > > Any ideas?
>> > >
>> > > --
>> > > Rafael
>> > >
>> >
>> > This is a bug in gcc. See, for instance,
>> >
>> http://www.nabble.com/Cast-ambiguity-compilation-error-td21532844.html
>> > or any earlier thread referenced therein. Upgrading
>> gcc should help.
>>
>> The version of g++ used in the hppa failing build mentioned
>> above is 4.3.3.
>> It is the same version that I use to build the packages on
>> my amd64 system.
>> Curiously enough, the automatic builds on armel [1] sparc
>> [2] architectures
>> fail in the same way.  On the other hand, building on i386
>> [3] and amd64 (my
>> development system) succeeds.
>>
>> Can that be caused by the way the int64 type is defined on
>> some
>> architectures?
>>
>> [1]
>> http://experimental.debian.net/fetch.php?&pkg=octave3.1&ver=3.1.52-4&arch=armel&stamp=1235272806&file=log&as=raw
>> [2]
>> http://experimental.debian.net/fetch.php?&pkg=octave3.1&ver=3.1.52-4&arch=sparc&stamp=1235287227&file=log&as=raw
>> [3]
>> http://experimental.debian.net/fetch.php?&pkg=octave3.1&ver=3.1.52-4&arch=i386&stamp=1235278606&file=log&as=raw
>>
>> --
>> Rafael
>
> we have the same problem on cygwin with gcc-4.3.2.
> I suspect is something around int64 type, but I have not yet
> identified the root cause.
>
> Marco
>


I don't say it's impossible, but I do not see how it can actually be
possible, unless int64_t is a typedef of double
(which would really be weird). So I still suspect that gcc is just
buggy on some platforms.

-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz