On Sun, Jun 04, 2023 at 07:56:59PM +0200, Theo Buehler wrote:
> On Sun, Jun 04, 2023 at 05:55:02PM +0000, Klemens Nanni wrote:
> > On Sun, Jun 04, 2023 at 05:48:24PM +0000, Brian Callahan wrote:
> > > Hi ports --
> > > 
> > > Attached is an update to devel/dwz. The main benefit for us is a modest
> > > speedup.
> > > 
> > > The full announcement with changes is here:
> > > https://sourceware.org/pipermail/dwz/2022q4/001385.html
> > > 
> > > Seems OK on my end on amd64. But since dwz is integral for debug
> > > packages, looking for more eyeballs.
> > > 
> > > OK?
> > > 
> > > ~Brian
> > 
> > You fetch xxhash and turn it into a local include.
> > Did you try adding the following as dependency instead?
> > 
> >     $ pkglocate include/xxhash.h
> >     xxhash-0.8.1p0:sysutils/xxhash:/usr/local/include/xxhash.h
> 
> I think that would create a loop. xxhash has DEBUG_PACKAGES set and
> dwz is needed to extract the debug info.

Makes sense now, thanks :)

> > > Index: patches/patch-dwz_c
> > > ===================================================================
> > > RCS file: /cvs/ports/devel/dwz/patches/patch-dwz_c,v
> > > retrieving revision 1.3
> > > diff -u -p -r1.3 patch-dwz_c
> > > --- patches/patch-dwz_c   11 Mar 2022 18:50:00 -0000      1.3
> > > +++ patches/patch-dwz_c   29 Nov 2022 23:40:09 -0000
> > > @@ -1,39 +1,46 @@
> > >  Create error() function, Linux-ism.
> > >  Our libelf doesn't have some of these defines.
> > >  obstack is not a system header
> > > +xxhash is not a system header

Perhaps "Use local copy to break dependency cycle" to clarify that?
Then I wouldn't have asked in the first place, as I read your comment
as '' <xxhash.h> does not exist, force "xxhash.h" ''

> > > +-#include <xxhash.h>
> > > ++#include "xxhash.h"

Reply via email to