On Tue, Dec 6, 2016 at 5:15 PM, Diego Biurrun <[email protected]> wrote: > On Tue, Dec 06, 2016 at 04:23:09PM -0500, Vittorio Giovara wrote: >> On Tue, Dec 6, 2016 at 2:43 AM, Diego Biurrun <[email protected]> wrote: >> > On Tue, Dec 06, 2016 at 12:27:22AM -0500, Vittorio Giovara wrote: >> >> --- a/libavdevice/version.h >> >> +++ b/libavdevice/version.h >> >> @@ -27,8 +27,8 @@ >> >> >> >> -#define LIBAVDEVICE_VERSION_MAJOR 56 >> >> -#define LIBAVDEVICE_VERSION_MINOR 1 >> >> +#define LIBAVDEVICE_VERSION_MAJOR 57 >> >> +#define LIBAVDEVICE_VERSION_MINOR 0 >> >> #define LIBAVDEVICE_VERSION_MICRO 0 >> >> >> >> --- a/libavresample/version.h >> >> +++ b/libavresample/version.h >> >> @@ -27,7 +27,7 @@ >> >> >> >> -#define LIBAVRESAMPLE_VERSION_MAJOR 3 >> >> +#define LIBAVRESAMPLE_VERSION_MAJOR 4 >> >> #define LIBAVRESAMPLE_VERSION_MINOR 0 >> >> #define LIBAVRESAMPLE_VERSION_MICRO 0 >> >> >> >> --- a/libswscale/version.h >> >> +++ b/libswscale/version.h >> >> @@ -26,7 +26,7 @@ >> >> >> >> -#define LIBSWSCALE_VERSION_MAJOR 4 >> >> +#define LIBSWSCALE_VERSION_MAJOR 5 >> >> #define LIBSWSCALE_VERSION_MINOR 0 >> >> #define LIBSWSCALE_VERSION_MICRO 0 >> > >> > These bumps seem gratuitous, there are no deprecated APIs in these >> > libraries. >> >> I'm aware there are not ABI-changes in those libs, but historically >> we've always bumped every library to avoid any possible version >> conflict. See for example 6d3ea1957f681b3bf9c752e6d21a501cc8d4180d or >> 3bc2e89c76e88ae6f1fd5287e0b11abcfc3c601c. While most of the root >> causes for those incompatibilities have been long resolved, it's still >> a good practice and relatively harmless to do. > > There was this idea of a single version number for all libraries. > Maybe it's time to revisit it?
I'm not against it, this is probably a good time to discuss it. >> > Also, lots of tests currently break if you bump all version numbers because >> > the output of the framecrc muxer changes. That will have to be addressed. >> > Anton knows the exact reason and wanted to make a decision which way to >> > resolve this. >> > >> > https://oracle.libav.org/x86_64-linux-version-bump/20161204165430 >> >> This is because that test disables a particular API which changes the >> test duration, which is not yet included in this removal. > > I suggest we resolve this in the not-so-distant future and try to maintain > a green version-bump Oracle instance. The bumb changes hashes because a function becomes more precise, there is not much we can do besides applying a custom patch to that machine. -- Vittorio _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
