Re: python-libconfig failing to build with musl

Re: python-libconfig failing to build with musl

[Rich Felker]

On Tue, Feb 28, 2017 at 10:12:33PM +0100, Thomas Petazzoni wrote:
> Hello,
> 
> The python-libconfig doesn't build with the musl C library, with some
> very weird error coming all the way from Boost Python:
> 
>   error: invalid use of incomplete type 'struct _IO_FILE'
> 
> See:
> 
>   http://autobuild.buildroot.net/results/f0f/f0f6cdc8c38c024772615d5e677b0f4ad63ef7ec/build-end.log
> 
> for the complete build log of python-libconfig.
> 
> If you want a way to easily reproduce, let me know, I can provide the
> 4/5 commands that can allow you to reproduce it with Buildroot.

It's intentional that FILE is an incomplete type. Do you know what
they're trying to do by trying to use it as if it were a complete
type? This is a bug that needs a patch, preferably one suitable for
upstream.

Rich

Re: Re: python-libconfig failing to build with musl

[Rich Felker]

On Tue, Feb 28, 2017 at 04:49:31PM -0500, Rich Felker wrote:
> On Tue, Feb 28, 2017 at 10:12:33PM +0100, Thomas Petazzoni wrote:
> > Hello,
> > 
> > The python-libconfig doesn't build with the musl C library, with some
> > very weird error coming all the way from Boost Python:
> > 
> >   error: invalid use of incomplete type 'struct _IO_FILE'
> > 
> > See:
> > 
> >   http://autobuild.buildroot.net/results/f0f/f0f6cdc8c38c024772615d5e677b0f4ad63ef7ec/build-end.log
> > 
> > for the complete build log of python-libconfig.
> > 
> > If you want a way to easily reproduce, let me know, I can provide the
> > 4/5 commands that can allow you to reproduce it with Buildroot.
> 
> It's intentional that FILE is an incomplete type. Do you know what
> they're trying to do by trying to use it as if it were a complete
> type? This is a bug that needs a patch, preferably one suitable for
> upstream.

On further investigation, it looks like there's something weird that
boost is doing that's causing it to get expanded in a template that
uses FILE rather than FILE*. Not clear why but this is probably a bug
in boost. Maybe someone who understands C++ better could advise.

BTW, any idea what happened to your mail to the list? It seems to have
gotten lost.

Rich

Re: Re: python-libconfig failing to build with musl

[Thomas Petazzoni]

Hello,

On Tue, 28 Feb 2017 21:25:43 -0500, Rich Felker wrote:

> > It's intentional that FILE is an incomplete type. Do you know what
> > they're trying to do by trying to use it as if it were a complete
> > type? This is a bug that needs a patch, preferably one suitable for
> > upstream.  
> 
> On further investigation, it looks like there's something weird that
> boost is doing that's causing it to get expanded in a template that
> uses FILE rather than FILE*. Not clear why but this is probably a bug
> in boost. Maybe someone who understands C++ better could advise.

Yeah, I also tried to have a look, but C++ and Boost are too unfamiliar
to me, and I had to give up.

> BTW, any idea what happened to your mail to the list? It seems to have
> gotten lost.

My e-mails for some reason don't make it to the mailing list. I am
registered on the list as I'm receiving all the traffic, and I'm using
the same From: address as the one used for my mailing list
subscription. My previous e-mail to the list (about SSP on x86) also
didn't make it to the list.

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

Re: Re: python-libconfig failing to build with musl

[Rich Felker]

On Wed, Mar 01, 2017 at 09:03:10AM +0100, Thomas Petazzoni wrote:
> > BTW, any idea what happened to your mail to the list? It seems to have
> > gotten lost.
> 
> My e-mails for some reason don't make it to the mailing list. I am
> registered on the list as I'm receiving all the traffic, and I'm using
> the same From: address as the one used for my mailing list
> subscription. My previous e-mail to the list (about SSP on x86) also
> didn't make it to the list.

Should be fixed now.

Rich