Re: libsndfile 1.2.0
On 2023/01/11 15:50:54 +0100, Jan Stary wrote: > Ah, looking at the nm(1) diff of the previous and current linsdfile.so, > the ogg_opus_seek_manual symbol is missing now (1.2.0) wrt previous (1.1.0). > Does that imply a SHARED_LIBS bump? that's a symbol used by the library, not provided by it, so no bump is needed. /usr/src/lib/check_sym is quite handy for these things, (assuming you have the 'old' version installed) you can use it to compare the system version in /usr/local/lib with the one in the fake tree /usr/ports/pobj//fake-amd64/usr/local/lib/ and look at its output.
Re: libsndfile 1.2.0
On 2023/01/11 15:50, Jan Stary wrote: > Ah, looking at the nm(1) diff of the previous and current linsdfile.so, > the ogg_opus_seek_manual symbol is missing now (1.2.0) wrt previous (1.1.0). > Does that imply a SHARED_LIBS bump? lower-case 't' not upper-case; that's a private internal function, not exported $ /usr/src/lib/check_sym /usr/local/lib/libsndfile.so.7.0 /usr/obj/ports/libsndfile-1.2.0/fake-amd64/usr/local/lib/libsndfile.so.7.0 /usr/local/lib/libsndfile.so.7.0 --> /usr/obj/ports/libsndfile-1.2.0/fake-amd64/usr/local/lib/libsndfile.so.7.0 No dynamic export changes
Re: libsndfile 1.2.0
On Wed, Jan 11, 2023 at 03:50:54PM +0100, Jan Stary wrote: > Ah, looking at the nm(1) diff of the previous and current linsdfile.so, > the ogg_opus_seek_manual symbol is missing now (1.2.0) wrt previous (1.1.0). > Does that imply a SHARED_LIBS bump? No. ogg_opus_seek_manual wasn't public: 000782f0 t ogg_opus_seek_manual If it had been an uppercase T then this would have warranted a major bump due to symbol removal. (I verified with /usr/lib/check_sym and checked by looking at the diff that your update didn't require a bump)
Re: libsndfile 1.2.0
Ah, looking at the nm(1) diff of the previous and current linsdfile.so, the ogg_opus_seek_manual symbol is missing now (1.2.0) wrt previous (1.1.0). Does that imply a SHARED_LIBS bump? Jan On Jan 11 13:31:17, h...@stare.cz wrote: > On Jan 10 20:03:49, h...@stare.cz wrote: > > I am working on an update of libsndfile to 1.2.0. > > diff below. > > Tested on current/amd64 and current/aarch64; > tested with sox (converting various formats). > Please test everywhere. > > On Jan 10 20:11:54, t...@theobuehler.org wrote: > > -GH_TAGNAME=1.1.0 > > +GH_TAGNAME=1.2.0 > > On Jan 10 19:35:22, s...@spacehopper.org wrote: > > Though now https://github.com/libsndfile/libsndfile/issues/816 has been > > fixed and the uploaded tar.xz should work with cmake, it may make sense > > to switch to that instead of the on-the-fly generated tar.gz. > > That's what I did; it downloads the "static" tar.xz, > but I kept the cmake build of the 1.1.0 port: it detects > the MPEG dependencies correctly, unlike the ./configure script. > > Jan > > > > Index: Makefile > === > RCS file: /cvs/ports/audio/libsndfile/Makefile,v > retrieving revision 1.41 > diff -u -p -r1.41 Makefile > --- Makefile 3 Apr 2022 11:12:38 - 1.41 > +++ Makefile 11 Jan 2023 12:13:49 - > @@ -1,11 +1,12 @@ > COMMENT= library to handle various audio file formats > > -GH_ACCOUNT= libsndfile > -GH_PROJECT= libsndfile > -GH_TAGNAME= 1.1.0 > +V= 1.2.0 > +DISTNAME=libsndfile-${V} > +EXTRACT_SUFX=.tar.xz > CATEGORIES= audio > > HOMEPAGE=https://libsndfile.github.io/libsndfile/ > +MASTER_SITES= > https://github.com/libsndfile/libsndfile/releases/download/${V}/ > > MAINTAINER= Jan Stary > > Index: distinfo > === > RCS file: /cvs/ports/audio/libsndfile/distinfo,v > retrieving revision 1.20 > diff -u -p -r1.20 distinfo > --- distinfo 3 Apr 2022 11:12:38 - 1.20 > +++ distinfo 11 Jan 2023 12:13:49 - > @@ -1,2 +1,2 @@ > -SHA256 (libsndfile-1.1.0.tar.gz) = > ZCqHa9YbY/k0ZijbpfigNWo611DH9vQgGdJs5gumoVs= > -SIZE (libsndfile-1.1.0.tar.gz) = 684409 > +SHA256 (libsndfile-1.2.0.tar.xz) = > DjDnBy+D3ISGPi5V8pkXXH4EpZAq55z7mdQknuj21go= > +SIZE (libsndfile-1.2.0.tar.xz) = 730268
Re: libsndfile 1.2.0
On Wed, Jan 11, 2023 at 01:31:17PM +0100, Jan Stary wrote: > On Jan 10 20:03:49, h...@stare.cz wrote: > > I am working on an update of libsndfile to 1.2.0. > > diff below. > > Tested on current/amd64 and current/aarch64; > tested with sox (converting various formats). > Please test everywhere. > > On Jan 10 20:11:54, t...@theobuehler.org wrote: > > -GH_TAGNAME=1.1.0 > > +GH_TAGNAME=1.2.0 > > On Jan 10 19:35:22, s...@spacehopper.org wrote: > > Though now https://github.com/libsndfile/libsndfile/issues/816 has been > > fixed and the uploaded tar.xz should work with cmake, it may make sense > > to switch to that instead of the on-the-fly generated tar.gz. > > That's what I did; it downloads the "static" tar.xz, > but I kept the cmake build of the 1.1.0 port: it detects > the MPEG dependencies correctly, unlike the ./configure script. Committed, thanks.
Re: libsndfile 1.2.0
On Jan 10 20:03:49, h...@stare.cz wrote: > I am working on an update of libsndfile to 1.2.0. diff below. Tested on current/amd64 and current/aarch64; tested with sox (converting various formats). Please test everywhere. On Jan 10 20:11:54, t...@theobuehler.org wrote: > -GH_TAGNAME=1.1.0 > +GH_TAGNAME=1.2.0 On Jan 10 19:35:22, s...@spacehopper.org wrote: > Though now https://github.com/libsndfile/libsndfile/issues/816 has been > fixed and the uploaded tar.xz should work with cmake, it may make sense > to switch to that instead of the on-the-fly generated tar.gz. That's what I did; it downloads the "static" tar.xz, but I kept the cmake build of the 1.1.0 port: it detects the MPEG dependencies correctly, unlike the ./configure script. Jan Index: Makefile === RCS file: /cvs/ports/audio/libsndfile/Makefile,v retrieving revision 1.41 diff -u -p -r1.41 Makefile --- Makefile3 Apr 2022 11:12:38 - 1.41 +++ Makefile11 Jan 2023 12:13:49 - @@ -1,11 +1,12 @@ COMMENT= library to handle various audio file formats -GH_ACCOUNT=libsndfile -GH_PROJECT=libsndfile -GH_TAGNAME=1.1.0 +V= 1.2.0 +DISTNAME= libsndfile-${V} +EXTRACT_SUFX= .tar.xz CATEGORIES=audio HOMEPAGE= https://libsndfile.github.io/libsndfile/ +MASTER_SITES= https://github.com/libsndfile/libsndfile/releases/download/${V}/ MAINTAINER=Jan Stary Index: distinfo === RCS file: /cvs/ports/audio/libsndfile/distinfo,v retrieving revision 1.20 diff -u -p -r1.20 distinfo --- distinfo3 Apr 2022 11:12:38 - 1.20 +++ distinfo11 Jan 2023 12:13:49 - @@ -1,2 +1,2 @@ -SHA256 (libsndfile-1.1.0.tar.gz) = ZCqHa9YbY/k0ZijbpfigNWo611DH9vQgGdJs5gumoVs= -SIZE (libsndfile-1.1.0.tar.gz) = 684409 +SHA256 (libsndfile-1.2.0.tar.xz) = DjDnBy+D3ISGPi5V8pkXXH4EpZAq55z7mdQknuj21go= +SIZE (libsndfile-1.2.0.tar.xz) = 730268
Re: libsndfile 1.2.0
On Tue, Jan 10, 2023 at 08:32:19PM +0100, Jan Stary wrote: > On Jan 10 20:11:54, t...@theobuehler.org wrote: > > On Tue, Jan 10, 2023 at 08:03:49PM +0100, Jan Stary wrote: > > > I am working on an update of libsndfile to 1.2.0. > > > > > > Sadly, I was not around for the 1.1.0 update (thank you Brad), > > > when apparently the decision was made to switch to the cmake build. > > > Is there a particular reason for that? > > > > > > Under https://github.com/libsndfile/libsndfile/releases > > > they provide a tar.gz (as just a tag really) without ./configure, > > > so to use that is to pull in autoreconf and automake etc. However, > > > there is also a tar.xz which provides ./configure; I propose to switch > > > back to the gnu ./configure build, considering the archivers/xzcat > > > dependency a lesser evil than the cmake dependency. > > > > > > Are there any advantages to the cmake build? > > > > I don't think there is a disadvantage to it. It's already done. > > > > -GH_TAGNAME=1.1.0 > > +GH_TAGNAME=1.2.0 > > > > make makesum > > make FETCH_PACKAGES= > > Yes, I have that already; but I would like to get rid > of the cmake dependency, and the python dependency. Then I suggest that you finish that up, submit the update get that committed, then you can play with switching to autoconf and send a diff for that.
Re: libsndfile 1.2.0
On 2023/01/10 20:11, Theo Buehler wrote: > > I don't think there is a disadvantage to it. It's already done. > > -GH_TAGNAME=1.1.0 > +GH_TAGNAME=1.2.0 > > make makesum > make FETCH_PACKAGES= > > and you should be ready to test (and see if you need to bump the shared > library version, etc). Though now https://github.com/libsndfile/libsndfile/issues/816 has been fixed and the uploaded tar.xz should work with cmake, it may make sense to switch to that instead of the on-the-fly generated tar.gz.
Re: libsndfile 1.2.0
On Jan 10 20:11:54, t...@theobuehler.org wrote: > On Tue, Jan 10, 2023 at 08:03:49PM +0100, Jan Stary wrote: > > I am working on an update of libsndfile to 1.2.0. > > > > Sadly, I was not around for the 1.1.0 update (thank you Brad), > > when apparently the decision was made to switch to the cmake build. > > Is there a particular reason for that? > > > > Under https://github.com/libsndfile/libsndfile/releases > > they provide a tar.gz (as just a tag really) without ./configure, > > so to use that is to pull in autoreconf and automake etc. However, > > there is also a tar.xz which provides ./configure; I propose to switch > > back to the gnu ./configure build, considering the archivers/xzcat > > dependency a lesser evil than the cmake dependency. > > > > Are there any advantages to the cmake build? > > I don't think there is a disadvantage to it. It's already done. > > -GH_TAGNAME=1.1.0 > +GH_TAGNAME=1.2.0 > > make makesum > make FETCH_PACKAGES= Yes, I have that already; but I would like to get rid of the cmake dependency, and the python dependency. Jan > and you should be ready to test (and see if you need to bump the shared > library version, etc). > > > Are there any objections? > > > > Jan > > >
Re: libsndfile 1.2.0
On Tue, Jan 10, 2023 at 08:03:49PM +0100, Jan Stary wrote: > I am working on an update of libsndfile to 1.2.0. > > Sadly, I was not around for the 1.1.0 update (thank you Brad), > when apparently the decision was made to switch to the cmake build. > Is there a particular reason for that? > > Under https://github.com/libsndfile/libsndfile/releases > they provide a tar.gz (as just a tag really) without ./configure, > so to use that is to pull in autoreconf and automake etc. However, > there is also a tar.xz which provides ./configure; I propose to switch > back to the gnu ./configure build, considering the archivers/xzcat > dependency a lesser evil than the cmake dependency. > > Are there any advantages to the cmake build? I don't think there is a disadvantage to it. It's already done. -GH_TAGNAME=1.1.0 +GH_TAGNAME=1.2.0 make makesum make FETCH_PACKAGES= and you should be ready to test (and see if you need to bump the shared library version, etc). > Are there any objections? > > Jan >