Re: [NEW] cudd-3.0.0
On 2019/03/22 16:05, Alessandro DE LAURENZIS wrote: > Hello Stuart, > > On 22/03/2019 15:16, Stuart Henderson wrote: > > [...] > > Diff on top of yours. > > > > - fix fetches > > > > - it's c++ so use COMPILER > > Just for my comprehension, could you please elaborate a bit (or point me to > the relevant manpage)? On arches which still use GCC as the system compiler, a huge number of ports need to use the newer version of the compiler in ports, rather than the pre-c++11 one in base. Rather than playing whack-a-mole to fix breakage on those arches, and to avoid conflicts between the different C++ standard libraries for base and ports compilers, we're building everything using C++ with ports-gcc unless they're on the path to building ports-gcc itself. It's not ideal (especially because it's easily forgotten) but I don't have a better idea how to improve it yet.. > > > > - it builds with or without doxygen, but if present during the configure > > run then it's needed during build, so add a build dep > > > > - get rid of "-release @PACKAGE_VERSION@" to remove the lib symlink; > > if other things were looking for 'cudd-3.0.0' then you'd probably want > > to keep it but it doesn't sound like that's the case > > > > - symlink is simpler than patching for gdiff > > > > with this added, OK sthen@ to import. > [...] > > Works for me, no objections. Thanks a lot. > > -- > Alessandro DE LAURENZIS > [mailto:jus...@atlantide.t28.net] > Web: http://www.atlantide.t28.net > LinkedIn: https://www.linkedin.com/in/delaurenzis/ >
Re: [NEW] cudd-3.0.0
Hello Stuart, On 22/03/2019 15:16, Stuart Henderson wrote: [...] Diff on top of yours. - fix fetches - it's c++ so use COMPILER Just for my comprehension, could you please elaborate a bit (or point me to the relevant manpage)? - it builds with or without doxygen, but if present during the configure run then it's needed during build, so add a build dep - get rid of "-release @PACKAGE_VERSION@" to remove the lib symlink; if other things were looking for 'cudd-3.0.0' then you'd probably want to keep it but it doesn't sound like that's the case - symlink is simpler than patching for gdiff with this added, OK sthen@ to import. [...] Works for me, no objections. Thanks a lot. -- Alessandro DE LAURENZIS [mailto:jus...@atlantide.t28.net] Web: http://www.atlantide.t28.net LinkedIn: https://www.linkedin.com/in/delaurenzis/
Re: [NEW] cudd-3.0.0
On 2019/03/22 10:27, Alessandro DE LAURENZIS wrote:
> Hello Anthony,
>
> thanks a lot for your feedback.
>
> On 22/03/2019 06:16, Anthony J. Bentley wrote:
> > Hi Alessandro,
> >
> > Alessandro DE LAURENZIS writes:
> > > All tests are PASS, provided that we use gdiff(1) instead of diff(1) for
> > > the dddmp set.
> >
> > Careful...
> >
> > -# Store corresponding CNF in different format:
> > +# Store corresponding CNF in gdifferent format:
> > # NodeByNode method -> file 4.node1.tmp
> > # MaxtermByMaxterm -> file 4.max1.tmp
> > -# Best with different options:
> > +# Best with gdifferent options:
>
> Ooops. Fixed.
>
> >
> > > HOMEPAGE = https://sourceforge.net/projects/cudd-mirror/
> >
> > I think https://www.davidkebo.com/cudd is a better homepage. I would use
> > both SourceForge and https://www.davidkebo.com/source/cudd_versions/ in
> > MASTER_SITES.
>
> Actually David doesn't seem to be affiliated to the project in any way, and
> his page is just a tutorial (there are a few on the net).
>
> The most complete source I can find is a Github mirror:
>
> https://github.com/ivmai/cudd
>
> which contains a collection of the most recent releases and upstream's
> README as front page.
>
> I would suggest to switch to this as master site, but I'm facing issues in
> fetching the tarball: we're trying to download:
>
> https://github.com/ivmai/cudd/archive/3.0.0/cudd-3.0.0.tar.gz
>
> instead of:
>
> https://github.com/ivmai/cudd/archive/cudd-3.0.0.tar.gz
>
> I vaguely remember a similar problem some months ago... do we have to change
> something in ports' infrastructure?
>
> >
> > > One thing I'm not able to do is to force the name of the .so file to be
> > > libcudd.so.0.0, instead of libcudd-3.0.0.so.0.0. Could someone point me
> > > in the right direction?
> >
> > Well, it seems to create both. I suppose you could @comment out the
> > redundant library in PLIST, but personally I wouldn't bother; in ports
> > like textproc/apertium I just leave it be. Which one does OpenSTA pick up?
> >
>
> OpenSTA searches for libcudd-3.0.0.so.0.0 (so we would need
>
> WANTLIB += cudd-3.0.0
>
> which doesn't seem ideal to me, but maybe I'm wrong...).
>
> Also, I would prefer to add some documentation, given that it's available in
> the release (although not updated to the latest version), but I didn't find
> a way to do it using a target from the upstream's Makefile; I just added a
> post-install part in the port's one (this requires pdflatex, so I added
> print/texlive/base to BUILD_DEPENDS).
>
> Updated tarball (using GH_ACCOUNT/PROJECT/TAGNAME, so breaking 'make fetch')
> attached.
>
> --
> Alessandro DE LAURENZIS
> [mailto:jus...@atlantide.t28.net]
> Web: http://www.atlantide.t28.net
> LinkedIn: https://www.linkedin.com/in/delaurenzis/
Diff on top of yours.
- fix fetches
- it's c++ so use COMPILER
- it builds with or without doxygen, but if present during the configure
run then it's needed during build, so add a build dep
- get rid of "-release @PACKAGE_VERSION@" to remove the lib symlink;
if other things were looking for 'cudd-3.0.0' then you'd probably want
to keep it but it doesn't sound like that's the case
- symlink is simpler than patching for gdiff
with this added, OK sthen@ to import.
diff --git Makefile Makefile
index cb1d15c..c3314f0 100644
--- Makefile
+++ Makefile
@@ -2,13 +2,14 @@
COMMENT = Colorado University Decision Diagram
+V =3.0.0
GH_ACCOUNT = ivmai
GH_PROJECT = cudd
-GH_TAGNAME = 3.0.0
+GH_TAGNAME = cudd-$V
+DISTNAME = cudd-$V
CATEGORIES = devel
-SHARED_LIBS += cudd-${GH_TAGNAME} 0.0 # 0.0
SHARED_LIBS += cudd 0.0 # 0.0
MAINTAINER = Alessandro De Laurenzis
@@ -18,7 +19,10 @@ PERMIT_PACKAGE_CDROM = Yes
WANTLIB = ${COMPILER_LIBCXX} m
-BUILD_DEPENDS =print/texlive/base
+COMPILER = base-clang ports-gcc base-gcc
+
+BUILD_DEPENDS =devel/doxygen \
+ print/texlive/base
CONFIGURE_STYLE = gnu
@@ -32,4 +36,7 @@ post-install:
${INSTALL_DATA} ${WRKSRC}/doc/cudd.pdf ${PREFIX}/share/doc/cudd
${INSTALL_DATA} ${WRKSRC}/doc/phase.pdf ${PREFIX}/share/doc/cudd
+pre-test:
+ ln -s ${LOCALBASE}/bin/gdiff ${WRKDIR}/bin/diff
+
.include
diff --git distinfo distinfo
index 28e2db0..cfe8ddd 100644
--- distinfo
+++ distinfo
@@ -1,2 +1,2 @@
-SHA256 (cudd-3.0.0.tar.gz) = uOlmtFYslqA+f76iOXKVh9ezldU8rcw5pyA7Sc9+62k=
-SIZE (cudd-3.0.0.tar.gz) = 1175302
+SHA256 (cudd-3.0.0.tar.gz) = X+FFBBxZRonm589M1iPV8rfDYmFwi+jJpyrtcs9nrM4=
+SIZE (cudd-3.0.0.tar.gz) = 1170745
diff --git pkg/PLIST pkg/PLIST
index aae6309..02d57dd 100644
--- pkg/PLIST
+++ pkg/PLIST
@@ -1,6 +1,5 @@
@comment $OpenBSD: PLIST,v$
include/cudd.h
-@lib lib/libcudd-3.0.0.so.${LIBcudd-3.0.0_VERSION}
lib/libcudd.a
lib/libcudd.la
@lib lib/libcudd.so.${LIBcudd_VERSION}
diff --git patches/patch-Makefile_in patches/patch-Makefile_in
new file mode
Re: [NEW] cudd-3.0.0
Hello, On 22/03/2019 10:27, Alessandro DE LAURENZIS wrote: Hello Anthony, thanks a lot for your feedback. On 22/03/2019 06:16, Anthony J. Bentley wrote: Hi Alessandro, Alessandro DE LAURENZIS writes: All tests are PASS, provided that we use gdiff(1) instead of diff(1) for the dddmp set. Careful... -# Store corresponding CNF in different format: +# Store corresponding CNF in gdifferent format: # NodeByNode method -> file 4.node1.tmp # MaxtermByMaxterm -> file 4.max1.tmp -# Best with different options: +# Best with gdifferent options: Ooops. Fixed. HOMEPAGE = https://sourceforge.net/projects/cudd-mirror/ I think https://www.davidkebo.com/cudd is a better homepage. I would use both SourceForge and https://www.davidkebo.com/source/cudd_versions/ in MASTER_SITES. Actually David doesn't seem to be affiliated to the project in any way, and his page is just a tutorial (there are a few on the net). The most complete source I can find is a Github mirror: https://github.com/ivmai/cudd which contains a collection of the most recent releases and upstream's README as front page. I would suggest to switch to this as master site, but I'm facing issues in fetching the tarball: we're trying to download: https://github.com/ivmai/cudd/archive/3.0.0/cudd-3.0.0.tar.gz instead of: https://github.com/ivmai/cudd/archive/cudd-3.0.0.tar.gz I vaguely remember a similar problem some months ago... do we have to change something in ports' infrastructure? One thing I'm not able to do is to force the name of the .so file to be libcudd.so.0.0, instead of libcudd-3.0.0.so.0.0. Could someone point me in the right direction? Well, it seems to create both. I suppose you could @comment out the redundant library in PLIST, but personally I wouldn't bother; in ports like textproc/apertium I just leave it be. Which one does OpenSTA pick up? OpenSTA searches for libcudd-3.0.0.so.0.0 (so we would need WANTLIB += cudd-3.0.0 which doesn't seem ideal to me, but maybe I'm wrong...). I was wrong; OpenSTA picks libcudd.so.0.0 up (but still needs WANTLIB+=cudd-3.0.0 for a clean port-lib-depends-check) Also, I would prefer to add some documentation, given that it's available in the release (although not updated to the latest version), but I didn't find a way to do it using a target from the upstream's Makefile; I just added a post-install part in the port's one (this requires pdflatex, so I added print/texlive/base to BUILD_DEPENDS). Updated tarball (using GH_ACCOUNT/PROJECT/TAGNAME, so breaking 'make fetch') attached. -- Alessandro DE LAURENZIS [mailto:jus...@atlantide.t28.net] Web: http://www.atlantide.t28.net LinkedIn: https://www.linkedin.com/in/delaurenzis/
Re: [NEW] cudd-3.0.0
Hello Anthony, thanks a lot for your feedback. On 22/03/2019 06:16, Anthony J. Bentley wrote: Hi Alessandro, Alessandro DE LAURENZIS writes: All tests are PASS, provided that we use gdiff(1) instead of diff(1) for the dddmp set. Careful... -# Store corresponding CNF in different format: +# Store corresponding CNF in gdifferent format: # NodeByNode method -> file 4.node1.tmp # MaxtermByMaxterm -> file 4.max1.tmp -# Best with different options: +# Best with gdifferent options: Ooops. Fixed. HOMEPAGE = https://sourceforge.net/projects/cudd-mirror/ I think https://www.davidkebo.com/cudd is a better homepage. I would use both SourceForge and https://www.davidkebo.com/source/cudd_versions/ in MASTER_SITES. Actually David doesn't seem to be affiliated to the project in any way, and his page is just a tutorial (there are a few on the net). The most complete source I can find is a Github mirror: https://github.com/ivmai/cudd which contains a collection of the most recent releases and upstream's README as front page. I would suggest to switch to this as master site, but I'm facing issues in fetching the tarball: we're trying to download: https://github.com/ivmai/cudd/archive/3.0.0/cudd-3.0.0.tar.gz instead of: https://github.com/ivmai/cudd/archive/cudd-3.0.0.tar.gz I vaguely remember a similar problem some months ago... do we have to change something in ports' infrastructure? One thing I'm not able to do is to force the name of the .so file to be libcudd.so.0.0, instead of libcudd-3.0.0.so.0.0. Could someone point me in the right direction? Well, it seems to create both. I suppose you could @comment out the redundant library in PLIST, but personally I wouldn't bother; in ports like textproc/apertium I just leave it be. Which one does OpenSTA pick up? OpenSTA searches for libcudd-3.0.0.so.0.0 (so we would need WANTLIB += cudd-3.0.0 which doesn't seem ideal to me, but maybe I'm wrong...). Also, I would prefer to add some documentation, given that it's available in the release (although not updated to the latest version), but I didn't find a way to do it using a target from the upstream's Makefile; I just added a post-install part in the port's one (this requires pdflatex, so I added print/texlive/base to BUILD_DEPENDS). Updated tarball (using GH_ACCOUNT/PROJECT/TAGNAME, so breaking 'make fetch') attached. -- Alessandro DE LAURENZIS [mailto:jus...@atlantide.t28.net] Web: http://www.atlantide.t28.net LinkedIn: https://www.linkedin.com/in/delaurenzis/ cudd.tar.gz Description: application/gzip
Re: [NEW] cudd-3.0.0
Hi Alessandro, Alessandro DE LAURENZIS writes: > All tests are PASS, provided that we use gdiff(1) instead of diff(1) for > the dddmp set. Careful... -# Store corresponding CNF in different format: +# Store corresponding CNF in gdifferent format: # NodeByNode method -> file 4.node1.tmp # MaxtermByMaxterm -> file 4.max1.tmp -# Best with different options: +# Best with gdifferent options: > HOMEPAGE = https://sourceforge.net/projects/cudd-mirror/ I think https://www.davidkebo.com/cudd is a better homepage. I would use both SourceForge and https://www.davidkebo.com/source/cudd_versions/ in MASTER_SITES. > One thing I'm not able to do is to force the name of the .so file to be > libcudd.so.0.0, instead of libcudd-3.0.0.so.0.0. Could someone point me > in the right direction? Well, it seems to create both. I suppose you could @comment out the redundant library in PLIST, but personally I wouldn't bother; in ports like textproc/apertium I just leave it be. Which one does OpenSTA pick up? -- Anthony J. Bentley
