On 2018/08/11 17:50, Theo Buehler wrote: > On Sat, Aug 11, 2018 at 02:58:44PM +0100, Stuart Henderson wrote: > > On 2018/08/11 00:43, Solene Rapenne wrote: > > > this patch updates libpng to 1.6.35, the new version changelog is the > > > following: > > > > > > Changes since the last public release (1.6.34): > > > > > > Restored 21 of the contrib/pngsuite/i*.png, which do not cause test > > > failures. Placed the remainder in contrib/pngsuite/interlaced/i*.png. > > > Added calls to png_set_*() transforms commonly used by browsers to > > > the fuzzer. > > > Removed some unnecessary brackets in pngrtran.c > > > Fixed miscellaneous typos (Patch by github user "luzpaz"). > > > Change "ASM C" to "C ASM" in CMakeLists.txt > > > Fixed incorrect handling of bKGD chunk in sub-8-bit files (Cosmin) > > > Added hardware optimization directories to zip and 7z distributions. > > > Fixed incorrect bitmask for options. > > > Fixed many spelling typos. > > > Make png_get_iCCP consistent with man page (allow compression-type > > > argument > > > to be NULL, bug report by Lenard Szolnoki). > > > Replaced the remaining uses of png_size_t with size_t (Cosmin) > > > Fixed the calculation of row_factor in png_check_chunk_length > > > (reported by Thuan Pham in SourceForge issue #278) > > > Added missing parentheses to a macro definition > > > (suggested by "irwir" in GitHub issue #216) > > > > > > > > > It compiles fine and the test pass 100%, maybe a bulk build is required? > > > > I've read through a diff of old and new source, I don't think a bulk > > build is necessary. OK with me though I wouldn't mind another pair of > > eyes on it as it's high in the dependency chain and the release approaches. > > FWIW: I looked through the diff as well and couldn't spot anything > worrisome. I'm also ok with this. > > There's one build warning: > > /usr/ports/pobj/png-1.6.35/libpng-1.6.35/contrib/libtests/pngvalid.c:11661:4: > warning: implicit declaration of function 'feenableexcept' is invalid in C99 > [-Wimplicit-function-declaration] > feenableexcept(FE_DIVBYZERO | FE_INVALID | FE_OVERFLOW); > ^ > > due to the fact that feenableexcept() is guarded by __BSD_VISIBLE, which > is set to 0 if _POSIX_SOURCE is defined without _BSD_SOURCE being defined. > > Patch below would fix this. Tests still pass. > > If a more experienced porter agrees with this, please include it with > your diff.
I'd be OK either way on this, it was already there in 1.6.34, and I think it's basically a noop anyway. > Index: patches/patch-contrib_libtests_pngvalid_c > =================================================================== > RCS file: patches/patch-contrib_libtests_pngvalid_c > diff -N patches/patch-contrib_libtests_pngvalid_c > --- /dev/null 1 Jan 1970 00:00:00 -0000 > +++ patches/patch-contrib_libtests_pngvalid_c 11 Aug 2018 15:43:41 -0000 > @@ -0,0 +1,16 @@ > +$OpenBSD$ > + > +Avoid implicit declaration of feenableexcept(3). > + > +Index: contrib/libtests/pngvalid.c > +--- contrib/libtests/pngvalid.c.orig > ++++ contrib/libtests/pngvalid.c > +@@ -20,7 +20,7 @@ > + > + #define _POSIX_SOURCE 1 > + #define _ISOC99_SOURCE 1 /* For floating point */ > +-#define _GNU_SOURCE 1 /* For the floating point exception extension */ > ++#define _BSD_SOURCE 1 /* For the floating point exception extension */ > + > + #include <signal.h> > + #include <stdio.h> >