Tomáš Chvátal | 26 Apr 15:17 2012
Picon

Weird build error on 3.5 l10ntools

Hi guys,

Some users in Gentoo are reporting weird issue in l10ntools when using gcc-4.4.

Please see the attached log for the issue.

Any idea how to resolve it?

Might something broke during the migration from ByteString?

Cheers

Tom
Attachment (error.log): application/octet-stream, 6682 bytes
Hi guys,

Some users in Gentoo are reporting weird issue in l10ntools when using gcc-4.4.

Please see the attached log for the issue.

Any idea how to resolve it?

Might something broke during the migration from ByteString?

Cheers

Tom
(Continue reading)

Lubos Lunak | 26 Apr 17:27 2012
Picon

Re: Weird build error on 3.5 l10ntools

On Thursday 26 of April 2012, Tomáš Chvátal wrote:
> Hi guys,
>
> Some users in Gentoo are reporting weird issue in l10ntools when using
> gcc-4.4.
>
> Please see the attached log for the issue.
>
> Any idea how to resolve it?

../inc/tagtest.hxx:68: note: candidates are: TokenInfo& 
TokenInfo::operator=(TokenInfo&)

 Assuming that gcc is not just lazy to report the signature properly, this is 
wrong, there is const missing in the argument, and that's most probably 
causing the problem. But I do not see any such operator there in the source, 
so it looks like the autogenerated operator is wrong, i.e. compiler bug.

--

-- 
 Lubos Lunak
 l.lunak <at> suse.cz
_______________________________________________
LibreOffice mailing list
LibreOffice <at> lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice
Stephan Bergmann | 26 Apr 17:38 2012
Picon

Re: Weird build error on 3.5 l10ntools

On 04/26/2012 05:27 PM, Lubos Lunak wrote:
> On Thursday 26 of April 2012, Tomáš Chvátal wrote:
>> Some users in Gentoo are reporting weird issue in l10ntools when using
>> gcc-4.4.
>>
>> Please see the attached log for the issue.
>>
>> Any idea how to resolve it?
>
> ../inc/tagtest.hxx:68: note: candidates are: TokenInfo&
> TokenInfo::operator=(TokenInfo&)
>
>   Assuming that gcc is not just lazy to report the signature properly, this is
> wrong, there is const missing in the argument, and that's most probably
> causing the problem. But I do not see any such operator there in the source,
> so it looks like the autogenerated operator is wrong, i.e. compiler bug.

The autogenerated operator= would take non-const ref if any of the data 
members' operator= required that.  The only suspiciously-looking one in 
TokenInfo is StringHashMap aProperties, which is a boost::unordered_map. 
  Something wrong with the boost version maybe (assuming that the 
problem happens --with-system-boost)?

Stephan
_______________________________________________
LibreOffice mailing list
LibreOffice <at> lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice
Caolán McNamara | 26 Apr 21:11 2012
Picon

Re: Weird build error on 3.5 l10ntools

On Thu, 2012-04-26 at 17:38 +0200, Stephan Bergmann wrote:
> On 04/26/2012 05:27 PM, Lubos Lunak wrote:
> > On Thursday 26 of April 2012, Tomáš Chvátal wrote:
> >> Some users in Gentoo are reporting weird issue in l10ntools when using
> >> gcc-4.4.
> >>
> >> Please see the attached log for the issue.
> >>
> >> Any idea how to resolve it?
> >
> > ../inc/tagtest.hxx:68: note: candidates are: TokenInfo&
> > TokenInfo::operator=(TokenInfo&)
> >
...

>  boost::unordered_map. 

sounds like...

https://svn.boost.org/trac/boost/ticket/5119

we've got boost.5119.unordered_map-cp-ctor.patch in local boost for
that.

C.

_______________________________________________
LibreOffice mailing list
LibreOffice <at> lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libreoffice
(Continue reading)


Gmane