[Pkg-xmpp-devel] Bug#742748: Re : Re: jabberd2: fails to build with clang instead of gcc

[Nicolas Sévelin-Radiguet]

----- Simon Josefsson <simon at josefsson.org> a écrit :
> Willem van den Akker <wvdakker at wilsoft.nl> writes:
> 
> > On Tue, 2015-04-28 at 12:27 +0200, Simon Josefsson wrote:
> >
> >> This patch looks wrong to me -- it changes a void function into an int
> >> function, and changes return values from void to 0, which the caller
> >> ignores.  
> >
> >
> > I dont see a problem there. It is only used for avoiding errors compling
> > with CLANG.
> > If or if not a caller ignores a return value is not an issue.
> >
> >
> >> As far as I can tell, the patch is not needed at all since
> >> upstream has fixed this issue.  
> >
> >
> > Not in version 2.3.3
> 
> I don't follow.  What error is there in 2.3.3 that needs to be patched?
> I believe the clang_FTBFS_Wreturn-type.diff patch we carried was the
> problem that caused the FTBFS, and removing the patch should fix
> everything.
> 
> The bug in 2.2.17 was this line:
> 
> https://github.com/jabberd2/jabberd2/blob/jabberd-2.2.17/sm/mod_roster.c#L463

That's right.
There was also an error here:
https://github.com/jabberd2/jabberd2/blob/jabberd-2.2.17/s2s/out.c#L630
fixed by commit ef675c9b2d9991899b267d5dbd01f14aa1e6abf8

> 
> The bug is that a void function can't return a value.  Gcc warns for
> this, clang throws an error.
> 
> The line is fixed in 2.3.3:
> 
> https://github.com/jabberd2/jabberd2/blob/jabberd-2.3.3/sm/mod_roster.c#L462
> 
> So the function is a proper void function.
> 
> /Simon

Right, again.
I should have double checked everything yesterday. I just ran a build of 2.3.3
*without* my patch and everything looks fine. Buildlog is here:
https://github.com/nonas/debian-clang/blob/master/buildlogs/jabberd2/jabberd2_2.3.3-1_amd64-20150428-1912.build

The build process is not automatic at all. So please ping me so that I can test
the next upload, just to be sure.

Cheers,
Nicolas