[ooo-build] ooo-build-3.1.0.99.3 fails building sddll0.hxx

[Christoph Willing]

On 25/08/2009, at 3:24 AM, Petr Mladek wrote:

> On Monday 24 August 2009, Rene Engelhard wrote:
>> Hi,
>>
>> On Mon, Aug 24, 2009 at 12:31:54PM -0400, John B wrote:
>>>   Until the boost developers fix their code (assuming they're even  
>>> going
>>> to do so), would it be possible to test for the version of boost  
>>> on a box
>>> and fall back to the boost in the libs-extern-sys package if boost  
>>> 1.37
>>> or later is detected? Cristoph Willing, the maintainer of the  
>>> Slackware
>>
>> No. Because distris *do* use system-boost with a fixed boost.  
>> Failing if
>> boost >= 1.37 would be bogus.
>
> You might find another workaround at
> http://bugzilla.novell.com/show_bug.cgi?id=533565#c4. It hacks OOo  
> sources to
> compile the two problematic files with exceptions enabled. It causes  
> that the
> wrong #ifdef in the boost header does not have any effect.
>
> I used it as a temporary hack. I do not want to commit it into ooo- 
> build. I
> expect that every distro should fix its boost. openSUSE is already  
> fixed, ...
>
> Anyway, feel free to send a patch against ooo-build that would add a  
> configure
> check and enable the hack if needed ;-)


While the --without-system-boost option is available, the hack doesn't  
seem essential.

Nevertheless, I just tried this hack on a build overnight and it  
completed fine. According to the bugzilla page, the patch re-enables  
exceptions. Isn't this just a "good thing" anyway?


I also tried the patch to fix boost (1.38) which Rene suggested and  
the ooo-build then worked. Unfortunately the boost patch is not going  
to be applied to the imminent Slackware 13.0 release (maybe later when  
more failures are reported).

However, while individual distros _could_ provide the fix, isn't it up  
to the boost people to fix their own releases? This fault was evident  
in 1.37, the patch works (and builds ooo) in 1.38. The patch applies  
cleanly to their 1.39 and 1.40 beta1 releases so I assume the fault is  
still there.

Is someone able to mount a case with the boost developers to have this  
fixed at the source?


chris


QCIF Access Grid Manager
University of Queensland