Re: UPDATE archivers/zstd

On 2018/03/30 20:19, Björn Ketelaars wrote:
> On Fri 30/03/2018 18:22, Stuart Henderson wrote:
> > On 2018/03/30 19:01, Björn Ketelaars wrote:
> > > On Fri 30/03/2018 13:45, Klemens Nanni wrote:
> > > > On Fri, Mar 30, 2018 at 05:30:55AM +0200, Björn Ketelaars wrote:
> > > > > zstd received an update, which focuses on performance. Changelog can be
> > > > > found at https://github.com/facebook/zstd/blob/dev/NEWS. The major
> > > > > number of the shared lib has been bumped as a couple of symbols have
> > > > > been removed, and new ones have been added.
> > > > >
> > > > > I've been running with this update for a couple of days on AMD64 using
> > > > > its only customer: borgbackup.
> > > > >
> > > > > OK?
> > > > Has anyone contacted upstream already to merge the trivial
> > > > detection/support bits we're carrying in our patches? Would be nice to
> > > > get those in, effectively limiting required patches on our site to
> > > > library version handling.
> > > >
> > > > Tests fail with a bunch of "use of undeclared identifier", did they pass
> > > > before?
> > > >
> > > > fatal error: too many errors emitted, stopping now [-ferror-limit=]
> > > > 2 warnings and 20 errors generated.
> > > > 2 warnings and 20 errors generated.
> > >
> > >
> > > I'm having a deja vu [0] ;-)
> > >
> > > Could you send me your test.log?
> > >
> > > [0] https://marc.info/?l=openbsd-ports&m=151473111522358&w=2
> > >
> >
> > I don't see that here. kn, do you have anything funny in CFLAGS/CDIAGFLAGS/etc?
> >
> > I have a different problem with tests though:
> >
> > - Compress without dictID
> > tmp : 21.56% ( 42430 => 9148 bytes, tmp.zst)
> > tmp.zst : Decoding error (36) : Restored data doesn't match checksum
> > gmake[1]: *** [Makefile:307: zstd-playTests] Error 1
> > gmake[1]: Leaving directory '/usr/obj/ports/zstd-1.3.4/zstd-1.3.4/tests'
>
> I ran 'make clean && make test' 3 times, one of which gave the same
> error. The other 2 runs were ok. Included a test.log of one of the
> successful runs.
>
> Any suggestions on how to debug this issue?

Oh, the in-tree one does the same - since it's not worse than that
I'm OK with updating, but maybe see if upstream have any suggestions
about the "without dictID" test?