This is an automated email from the git hooks/post-receive script. gregoa pushed a change to annotated tag upstream/0.43 in repository libffi-platypus-perl.
at c761434 (tag) tagging 388c02e7463bf3431bc388a858cb2084cd4d035e (commit) replaces upstream/0.42.01 tagged by gregor herrmann on Fri Jul 8 11:19:49 2016 +0200 - Log ----------------------------------------------------------------- Upstream version 0.43 Bakkiaraj M (1): Added GNOME based example to get system uptime Dylan Cali (4): add test for checking ffi function in bool context implement FFI::Platypus::Function bool overload add test for implicit void return type on attach use void return type for attach if not specified Fitz Elliott (3): Fix typo in Type.pod Fix more typos in Type.pod Fix typos in code example in Platypus.pm Graham Ollis (664): Initial commit travis update diagnostic update release tests experimental include Test::Vars in travis use XS instead of XS_INTERNAL for older versions of Perl that should be a hash basic types travis needs Exporter::Tidy simple library interface initial code to create a sub rm old proof of concepts stuff better interface for ffi_sub basics of ffi_pl_call caveat some initial simple compiler checks make travis happy again start of standard testing lib test for alloca void not allowed as argument support for non-pointer argument types other than int fix refcount bug signature can be passed in as a array ref save some memory by not having each type have its own name some 64bit fixes, and use Math::Int64 include PkgConfig as part of release use (public?) https interface try various versions of pkg-config if -libffi does not work documentation (environment) update doco add method for disabling alloca add libffi use bundled version of libffi if system libffi is not available don't ever bother to build dynamic version of libffi cygwin support fixes for linux from last commit use alloca in calling ffi functions (if available) detect endian-ness use Math::Int64 for arguments more types support any integer or floating point return type pointer arguments in, pointer return values bug in test helper for debugging in c environment for specifiying extra build flags code to debug ffi call arguments less casting rebuild testlib when one of its sources has changed this turns out to be a bug somewhere in libff + clang land check for bug with a wierd combination of clang libffi and debian call it ffi instead of none fix silly misunderstanding of croak in xs rm memory leak when no alloca available lets call it scratch1 instead of scratch pointers out return pointer types break some code out of FFI/Platypus.xs into refcount.c basic closures string to ptr test return non pointer types from closure/callback rename member fix scope test closure call was mucking up the stack do not need this scope test anymore closure: other return types fix order problem use CvXSUBANY(cv).any_ptr instead of a hash revert some closure with arguments skip long double test until it can be fixed remove old crufty stuff remove dupe option to strip optimize flags from build meta info for ffi_sub anonymous ffi_sub comment add examples latest PkgConfig latest version of PkgConfig reduce duplication update release tests doco update travis, release tests and diagnostic update release tests rm unused var update release tests v0.01 start over unworking prototype update release tests add ability to build from git do not include META.json in git repo automatically build distclean before build, iff Build exists nice idea, but not really doable without lots of innerds in Perl basics of creating a testlib add find_symbol method some basic autoconfiguring stubs for windows that will need to be filled in notes size detection build with or without dzil clean resets stuff in my never ending quest to come up with the right config find symbols in the main exe basics of a type rename complete set of ffi types plus strings ppport.h needs to be checked in for this to work. start of something that might one day work as a call :P the basics of a function.call method sign detection type and types methods add function method spelling and other house keeping declare it rm debug doco type names as perl constants unused var attach functions start of non-basic types spelling and doco the basics of pointers eh? start with version 0.02 since 0.01 was the previous prototype tests for signed eight byte integers add some array support only need Math::Int64 if IVs are 32 bits functions that return arrays optional Math::Int64 c API trigger rebuild when ffi_platypus_call.h is modified move include to .xs since .c files will not have perl guts do not checkin .tmp files rename test library to match system convention on cygwin travis travis on all versions of perl return fixed length array pointer tests use singular in test name all integer types (64 bit is borked) fixes previous breakage on 64 bit only minor stuff tests and fixes for strings check the opposite of is_null doco start of windows implementation for libdl figured out the deal with psapi.h (at least a work around for the issue) build libtest on strawberry ! use EXTERN macro for extern in libtest functions libtest also depends on ac floats for libtest documentation fixes v0.02 mb needs to be dynamic no wonder that did not work synopsis and see also make sure the indexer sees these api for creating closure make it so you do not have to declare int and such s/sv/platypus_sv/g check for header to decide if ac needs to be redone rather than using config_data add a little smartness to the build process avoid warning gitignore get meta data for types that needed to be a string eval closure types the basics of closures are here other integer types for closures drain the leak spelling v0.03 this is why we can't have nice things on free bsd apparently the optimize flag is just -O some 64 bit fixes for 32 bit Perls some examples stolen from FFI::Raw rm space some basic memory functions tweaks add cast function tweaks progress of a sort export memory functions by default newer version of FFI::CheckLib update meta float support support for double better examples call them opaque instead of pointer strdup beware duplicate symbols bearing gifts opaque arrays add test for opaque types make configure require perl 5.8.1 explicit pointer to opaque works a little differently in pre 5.20 closures and pointers passing strings into closures add perl implementation of strdup for platforms that do not provide it doco for memory functions spelling v0.04 environment variable for forcing the use of Math::Int64 doco doco again typo new abstract (maybe?) custom types for arguments in doco sapcing finish port of examples todo float custom test (broken) fix for when YAML not installed remove userdata refactor but still broken custom opaque type fix bug with cif related to custom types :P typos make some dies into warns unable to create a closure becomes warning instead of die start to break up some large files return; from a custom type callback should not set an argument warn on die inside of a closure (#12) allow argument count mismatch (#14) fix for perls without warn_sv fix for Perl 5.8.8 (#3) fix for (#6) comment fix for #2 use unions instead of casting for return value interface change, callbacks should be in a hash ref output custom type if no sub ref is provided then pass through perl_to_ffi_post hook rm accdiental dep on yaml v0.05 typo v0.06 lets call them native types start new type documentation better see also proper for for 5.8.8 (#3) do not allocate massive amounts of memory on the stack (#8) clean up cruft from windows libdl implementation (#13) fix memory leak doco fix for #9 doco for sticky spelling doco doco specify lib at new time attach should return $self forgot we do support strings as closure argument types for now increment the ref count. may have to think of something else v0.07 little closer but still wrong more debug progress on memory leak (#4) operating directly on @_ in the cast function seems to rm FIXME that has been fixeded closure type checking another todo that has been todun include buffer functions from FFI::Util v0.08 make cast and sizeof instance methods rename FFI::Platypus::Declare#function -> attach update examples for API changes ffi_to_perl and perl_to_ffi become native_to_perl and perl_to_native update changes more contributing tips native_type as part of the hash now v0.09 string pointer implementation #10 doco pure perl version of StringPointer define custom types in declare import start of buffer support extra files fixes for arguments api Remove the OO from the custom types API v0.10 doco use "opaque" instead of "pointer" in examples doco doco doco fix test bug when not using alloca() no warnings on ref to undef (there is a legitimate use for returning native type as a pointer) yet another broken zmq example :P ...and finally a working example yay doco more doco tweaks tweaking abstract yet again doco make unthreaded perl with Alien::FFI->install_type = 'system' be defensive of c++ doco doco v0.11 80 characters doco v0.12 fix some borked links v0.14 add ignore_not_found attribute tag new features with the version in which they were introduced new line merge Module::Build::FFI into this dist doco doco v0.14_01 needs Text::ParseWords strawberry and the short bus do not include internal release in changes package method declare interface uses bundled C if available u prefix shortforms v0.14_02 do not include internal revision in changes forgotten update v0.14_03 doco MB::FFI exit when compiler required and not there doco hide the warning eh? remove internal release from changes file v0.15 v0.16 remove FFI::Util from example where it is not needed more diagnostics reverse the polarity pointer to ::StringArray travis "5.21" on travis is really "5.21.0" which is virutally identical to 5.20 so not really worth testing travis add version travis maybe make it a bit faster travis: UUID::FFI and FFI::Util fortran programmers are people too new bzip2 example sans warning doco rust example (#20) doco: I wish I had known I could do this earlier... more rust examples (#20) rearrange the deck chairs allow sizeof custom types faster sizeof start working on records (#22) a record type can be a class record(My::Foo) record implementation argument input (#21,#22) remove debug record as a return type (#21,#22) record doco 32bit for pack and unpack refer to the tutorial simple struct example round trip record examples in the Type.pod (#21,#22) and the oo example ... instructions for how to help even simpler compare and contrast records with opaque pointers rework getSystemTime example to use new record type give windows example a win32 prefix add a thanks section to dist.ini add bool as a primitive type Check that the size of the record is correct spelling v0.17 add better support for C++ in Module::Build::FFI support for alt foreign languages rename 'with' to 'lang' add foreign language support for declarative interface rust examples are moving to FFI::Platypus::Lang::Rust generate contrib and support doco make m::b::ffi subclassable version number support for mangled languages have to pass in the libs to the mangler add FFI_PLATYPUS_DLERROR as an env var doco typo Module::Build::FFI#ffi_include_dir can now be a array reference ffi_{libtest,ffi}_dir can now be an array reference unused var fix build assembly for the future v0.18 accept other extensions in addition to $Config{dlext} (#29) update changes v0.19 next version will require latest Alien::FFI do not need this anymore attach wrapper colot class doco correct assembly specific doco menton FFI in the abstract itself use house style forgot to mention the version when #package was added example improvements shortcut for FFI::CheckLib::find_lib (#30) doco and meta (#30) another pointer to #find_lib wrapper example with commentary v0.20 add alignof method forgot the test records workaround for unable to detect alignment with gcc (#21) record members for integer and float types looks weird to have this commented out. opaque accessor doco use check_default_headers before check_alignof_type (#21) support for arrays in records but needs to be unborked This unborks my last commit Do not allow two members of the same name in a record test for array alignment use : for members that you don't want accessors for keep track of alignment for records record ro string support for fixed length strings acknowledgements spelling fixed length strings read/write string. use with care. (#21) doco allow hash initializer for records doco simplify uuid example using records and fixed length strings optimize for 80 cols v0.21 fix typo and simplify links for fortran and pascal Add Caveats section, working with non C languages (#20) probe for size and alignment of exotic floats (#32) bump Config::AutoConf version (#33) Java example, via gcj move examples to the bottom this is a crazy Java example for the heck of it s/an/a/ var length array update changes more changes v0.22 test with Lang::Fortra travis: test acme::ford::prefect::ffi and lang::pascal fp compiler for precise (i think) travis test more packages should not have included this yet travis test exampls fix package name example test has to come after TinyCC test example needs Convert::Binary::C travis add libgtop2 shesh travis re-enable old versions of perl; other build configs rm comment travis fix yaml error travis need recent Time::Seconds for examples start of exotic float support this variable was not assigned! strings are types too long double as argument and return type (#32) keep a separate argument_pointers pointer add G_EVAL seems to fix 5.12 and earlier cleanup for longdouble support (#32) caveat this includes partial support for complex types some big endian fixes rm accidnetal binary checkin closure big endian fixes final big endian fixes update changes complex eg probe for complex support (#32) if a probe compile fails do not die update changes support Math::Complex as an argument (#32) move the set exotic logic out of call.h document limited complex support better new type doco v0.23 typo should return 2 on failure fix windows issues refix ppport.h v0.24 add probe for longdouble this makes longdouble a probe failure for cygwin, which seems to be the right thing more long double fixes update changes v0.25 pointer to long double support only test types we can test properly skip tests on tests that arent there single element accessors for arrays tied array interface save guard against bad array index programmers note array of long double use debug flags for libtest array of long double return add notes about Alt::Alien::FFI::System more work arounds for Strawberry we do actually need the extra compiler/linker flags for the libtest step mark it as experimental v0.26 simplify win32 probe have to use the fullly qualified name probe for ABIs onyl do it once probe order fix use the right ffitarget.h different type of configure interface to the list of ABIs new non-default ABI interface doco windows types line numbers for eval fix scoping issue in Perl 5.8.8 fix #39 update changes lang Win32 defaults to stdcall on 32 bit windows fix for #35 v0.27 fix Win32::ErrorMode prereq workaround for openbsd64 issue solaris is apparently picky about this sort of thing more coplaints from solaris cc v0.28 rm mental helth reference see also P5NCI doco delay loading of Win32::ErrorMode v0.29 more diagnostics and the version number pass opaque in place of a closure expand tabs into spaces note the gh issues in Changes use house style fix for #5 v0.30 allow multiple CUSTOM_PERL arguments with argument counts > 1 (include/ffi_platypus_call.h) reduce the number of add operations that we need to do cleanup build.bat on cygwin win32 beep example skip test requiring snprintf on platforms that do not provide it ... and i forgot the conditional changes remove overly aggressive skip on test for snprintf v0.31 C++ free zone use win32 types remove generated file include/ppport.h travis: make sure Devel::PPPort is installed try this but skeptical more stops words for hunspell put -L flags from Alien::FFI in front doco fixes for MS Visual C++ (#16) ints are usually 4. longs are usually the word size. in modern times at least. ms version of alloca is _malloca accdiental checkin _malloc may allocate on the heap alter dignostic as does not necessary relate to strawberry anymore update contribs travis to test ZMQ::FFI internally, longdouble cannot be the same as double for us support for system perl on OS X v0.32 remove dup cases check for global destruction rm debug update changes v0.33 update contributors some cleanup forgot to add this file v0.34 use constant to avoid warnings in Perl 5.22 v0.35 constant was being removed because it was in core. but we require a more recent version. v0.36 rm xt/release/unused_vars.t require plugin version 1.74 for recent fixes yet another constant fix v0.37 add a simple thread test also test forks alter thread tests so that the FFI object gets destroyed irc doco rm explicit dep on constant more perls spelling v0.37_01 v0.37_02 convert::binary::c can be recommended now for now we need sudo on travis :( bump version up to latest improve thread safety more threading safety improvements v0.37_03 require more recent version of constant clone on creating new thread also update travis for new constant v0.37_04 older Perls need MY_CXT_KEY apparently v0.37_05 check for stdc headers v0.37_06 croak helpful here for diagnostics travis status forgot this one squish dev releases in the change log v0.38 see also for c::blocks fix unhelpful warning v0.39 spacing parser fix for function pointers v0.40 Update .gitattributes does this make linguist work? test against 5.22 update changes updates for recent [@Author::Plicease] try gfortran-4.6 travis update contributors update diag include libffi version in diagnostic tweak diagnostic catch a execute fail on the ABI probe (#64) do not test abi in /tmp (#64) spelling v0.41 support for msys2 v0.42 update release tests add see also to NativeCall start to discourage the declarative interface migrate more doco to oo interface fix spelling mistakes (gh#66) update readme include probe data in standard diagnostic. make build a little bit faster big endian 64 specific conversions rm trailing space more big endian 64 fixes fix checking endianness on wrong wordsize spelling fix v0.42_01 update contributors add appveyor also need this additional windows deps need newer constant after all spelling v0.43 Philip (9): allow integers to be passed as function pointers test new behaviour document the ability to cast closures to opaque and use them as function arguments Cache and reuse FFI closures generated for a FFI::Platypus::Closure test case which fails with v0.30 C file for testing multiple custom arguments add another test case which is slightly more realistic fix: allow multiple CUSTOM_PERL arguments with argument counts > 1 (Function.xs) fix compiler warning Vickenty Fesunov (1): Avoid unnecessary downgrade on when NV==longdouble bakkiaraj (1): Add getSystemTime.pl in to examples gregor herrmann (1): Imported Upstream version 0.43 ----------------------------------------------------------------------- No new revisions were added by this update. -- Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/pkg-perl/packages/libffi-platypus-perl.git _______________________________________________ Pkg-perl-cvs-commits mailing list Pkg-perl-cvs-commits@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-perl-cvs-commits