Marcel Raad writes:

Also, there is a new internal compiler error in boost/msm/back/dispatch_table.hpp, struct make_chain_row_from_map_entry.

I have created a Microsoft Connect bug report for the crash: https://connect.microsoft.com/VisualStudio/feedback/details/915061/vc12- update-3-rc-boost-1-56-compiler-crash

I would like to divide these problems into three bins:
1) Missing files/files in the wrong place in the build
* WinAPI
2) General differences between build structure for 1.55 and 1.56
* Underscores vs. periods
* No 7z version
* Broken links to the docs
* Weird metadata in the zip file
3) Other problems introduced by the build process
4) Problems with the release
* Broken links in the docs
* Multiply defined symbols
* Problems with smart_ptr in debug mode on VC 2013

Marcel - can you verify that the problems you saw do (or do not) exist in
the master branch of the git repo? If they don’t, please let me know. If
they do, please file bugs so they can get fixed.
Paul - I’ll be making a 7z version “soon”; this build was to make sure
everything was in the right place, etc. There will also be a .tar.bz2
version as well.

I’ll post an RC2 version with some of these things fixed in a bit.

— Marshall


On Mon, Jul 7, 2014 at 7:12 AM, Paul A. Bristow 
wrote:

>
>
> > -----Original Message-----
> > From: Boost [mailto:boost-bounces@lists.boost.org] On Behalf Of
> Marshall Clow
> > Sent: 06 July 2014 16:28
> > To: boost
> > Subject: [boost] Test build for 1.56 available
> >
> > I have made a test build for 1.56 and would appreciate help in finding
> problems with
> > it.
> >
> > boost.1.56.0.tar.gz:
> https://dl.dropboxusercontent.com/u/11048326/boost/boost-
> > 1.56.0a1c1.tar.gz
> > boost.1.56.0.zip:
> >       https://dl.dropboxusercontent.com/u/11048326/boost/boost-
> > 1.56.0a1c1.zip
> >
> > The tar.gz file builds w/o errors on my Mac.
> > The zip version contains windows line endings.
> >
> > I'm looking for:
> >       * missing files/libraries
> >       * files/folders that don't belong in the release
> >       * files/folders in the wrong places
> >       * build failures
> >       * Anything else that is wrong.
> >
> > One thing I've noticed is that these are *significantly* larger than the
> 1.55.0
> > archives.
> > A bunch of that is because the Boost.Ublas documentation has exploded in
> size
> - but
> > there are probably other reasons as well.
> >
> > Looking forward to feedback.
>
> Download and unzip
> Some library docs missing (align, algorithm...)
>
> But as short of time, I didn't build libraries to /lib from scratch but
> just
>
> ran b2 on /math/test and /mat/example and they built and ran OK  (VS Pro
> 2013 on
> Windows 8.1).
>
> Math/docs look OK at a glance.  (and have also probably increased in size
> ;-(
>
> It looks slightly odd that the master says
>
> http://www.boost.org/doc/libs/master/doc/html/Assignable.html
>
> has a banner saying "This is documentation for a development version of
> boost".
>
> But perhaps this will be changed for the final release?
>
> FWIW
>
> Paul
>
> PS OK  (what happened to the 7z version?)
> ---
> Paul A. Bristow
> Prizet Farmhouse
> Kendal UK LA8 8AB
> +44 01539 561830
>
>
>
>
>
>
>
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost
>

On Mon, Jul 7, 2014 at 10:56 PM, Marshall Clow wrote:

Ok, here are rc2s. Fixed (I believe) are: * Now includes winapi library * Underscores instead of periods in the names * No more weit metadata in the tar file.

https://dl.dropboxusercontent.com/u/11048326/boost/boost-1_56_0a1c2.zip https://dl.dropboxusercontent.com/u/11048326/boost/boost-1_56_0a1c2.tar.gz

Please check them out, let me know what problems you have.

gcc 4.8 on Linux x86_64, builds fine in C++03 and C++11 modes. Boost.Signals fails to build with MSVC 12: .\boost/iterator/iterator_facade.hpp(526) : error C2039: 'decrement' : is not a member of 'boost::signals::detail::named_slot_map_iterator' .\boost/signals/detail/named_slot_map.hpp(80) : see declaration of 'boost::signals::detail::named_slot_map_iterator' .\boost/iterator/iterator_facade.hpp(677) : see reference to function template instantiation 'void boost::iterator_core_access::decrement<Derived>(Facade &)' being compiled with [ Derived=boost::signals::detail::named_slot_map_iterator , Facade=boost::signals::detail::named_slot_map_iterator ] .\boost/iterator/iterator_facade.hpp(676) : while compiling class template member function 'boost::signals::detail::named_slot_map_iterator &boost::iterator_facade::operator --(void)' with [ Value=boost::signals::detail::connection_slot_pair ] .\boost/signals/detail/named_slot_map.hpp(84) : see reference to class template instantiation 'boost::iterator_facade' being compiled with [ Value=boost::signals::detail::connection_slot_pair ] .\boost/iterator/iterator_facade.hpp(544) : error C2039: 'advance' : is not a member of 'boost::signals::detail::named_slot_map_iterator' .\boost/signals/detail/named_slot_map.hpp(80) : see declaration of 'boost::signals::detail::named_slot_map_iterator' .\boost/iterator/iterator_facade.hpp(690) : see reference to function template instantiation 'void boost::iterator_core_access::advance<Derived>(Facade &,__w64 int)' being compiled with [ Derived=boost::signals::detail::named_slot_map_iterator , Facade=boost::signals::detail::named_slot_map_iterator ] .\boost/iterator/iterator_facade.hpp(689) : while compiling class template member function 'boost::signals::detail::named_slot_map_iterator &boost::iterator_facade::operator +=(__w64 int)' with [ Value=boost::signals::detail::connection_slot_pair ] Master branch is also affected. Somehow develop is not affected although Boost.Signals code is the same. I suspect the compiler is behaving unstable. I'll create a pull request to adjust the compiler version check for a workaround for this problem. There is also the problem with Boost.Atomic described in the nearby thread "[atomic] Structs with default constructor". The fix is in develop and will be merged to master in a few days. Will I need to explicitly ask to include this fix into the final release? The question about the packages revision is still relevant.

2014-07-07 22:56 GMT+04:00 Marshall Clow :

Ok, here are rc2s.

<...>

https://dl.dropboxusercontent.com/u/11048326/boost/boost-1_56_0a1c2.zip https://dl.dropboxusercontent.com/u/11048326/boost/boost-1_56_0a1c2.tar.gz

Please check them out, let me know what problems you have.

Links to the docs are broken for the following libraries in libs/libraries.htm: Algorithm, Align, CompressedPair, Context, Core, Coroutine, EnableIf, Functional/factory, Functional/forward Log, Odeint, Phoenix, Range, Spirit, Swap, TTI, TypeIndex, Units Library TypeIndex is totally missing (no headers, no folder in libs). Previous release was called: boost_1_55_0/ New release has folder: boost-1_56_0a1c2 Do we want to keep that minus? Some problems with bzip2 headers detection during build: libs/iostreams/src/bzip2.cpp:20:56: fatal error: bzlib.h: No such file or directory #include "bzlib.h" // Julian Seward's "bzip.h" header. ^ compilation terminated. "g++-4.8" -ftemplate-depth-128 -O3 -finline-functions -Wno-inline -Wall -pthread -fPIC -DBOOST_ALL_NO_LIB=1 -DBOOST_IOSTREAMS_DYN_LINK=1 -DBOOST_IOSTREAMS_USE_DEPRECATED -DNDEBUG -I"." -c -o "bin.v2/libs/iostreams/build/gcc-4.8/release/threading-multi/bzip2.o" "libs/iostreams/src/bzip2.cpp" Warning in locale: ./boost/locale/boundary/segment.hpp:30:35: warning: typedef ‘right_iterator’ locally defined but not used [-Wunused-local-typedefs] typedef RightIterator right_iterator; Compilation of Python failed: In file included from ./boost/python/detail/prefix.hpp:13:0, from ./boost/python/numeric.hpp:8, from libs/python/src/numeric.cpp:6: ./boost/python/detail/wrap_python.hpp:75:24: fatal error: patchlevel.h: No such file or directory #include ^ compilation terminated. "g++-4.8" -ftemplate-depth-128 -O3 -finline-functions -Wno-inline -Wall -pthread -fPIC -DBOOST_ALL_NO_LIB=1 -DBOOST_PYTHON_SOURCE -DNDEBUG -I"." -I"/usr/include/python2.7" -c -o "bin.v2/libs/python/build/gcc-4.8/release/threading-multi/numeric.o" "libs/python/src/numeric.cpp" ...failed gcc.compile.c++ bin.v2/libs/python/build/gcc-4.8/release/threading-multi/numeric.o... gcc.compile.c++ bin.v2/libs/python/build/gcc-4.8/release/threading-multi/list.o In file included from ./boost/python/detail/prefix.hpp:13:0, from ./boost/python/list.hpp:8, from libs/python/src/list.cpp:5: ./boost/python/detail/wrap_python.hpp:75:24: fatal error: patchlevel.h: No such file or directory #include ^ compilation terminated. "g++-4.8" -ftemplate-depth-128 -O3 -finline-functions -Wno-inline -Wall -pthread -fPIC -DBOOST_ALL_NO_LIB=1 -DBOOST_PYTHON_SOURCE -DNDEBUG -I"." -I"/usr/include/python2.7" -c -o "bin.v2/libs/python/build/gcc-4.8/release/threading-multi/list.o" "libs/python/src/list.cpp" ...failed gcc.compile.c++ bin.v2/libs/python/build/gcc-4.8/release/threading-multi/list.o... gcc.compile.c++ bin.v2/libs/python/build/gcc-4.8/release/threading-multi/long.o In file included from ./boost/python/detail/prefix.hpp:13:0, from ./boost/python/long.hpp:8, from libs/python/src/long.cpp:5: ./boost/python/detail/wrap_python.hpp:75:24: fatal error: patchlevel.h: No such file or directory #include ^ compilation terminated. Warnings in Spirit: ./boost/spirit/home/classic/core/primitives/primitives.hpp:50:68: warning: typedef ‘result_t’ locally defined but not used [-Wunused-local-typedefs] typedef typename parser_result::type result_t; ^ ./boost/spirit/home/classic/core/non_terminal/impl/grammar.ipp:285:68: warning: typedef ‘iterator_t’ locally defined but not used [-Wunused-local-typedefs] typedef typename helper_list_t::vector_t::reverse_iterator iterator_t; ^ ./boost/spirit/home/classic/tree/parse_tree.hpp:79:42: warning: typedef ‘attr_t’ locally defined but not used [-Wunused-local-typedefs] typedef typename match_t::attr_t attr_t; ^ -- Best regards, Antony Polukhin

On Tuesday 08 July 2014 09:58:11 Oliver Kowalke wrote:

2014-07-08 9:53 GMT+02:00 Antony Polukhin :

...

Links to the docs are broken for the following libraries in libs/libraries.htm: Algorithm, Align, CompressedPair, Context, Core, Coroutine, EnableIf, Functional/factory, Functional/forward Log, Odeint, Phoenix, Range, Spirit, Swap, TTI, TypeIndex, Units

does this issue require action on the side of the library authors (like generating and check-in of html etc.)?

No autogenerated files in git, please. I believe these test archives simply don't include all docs (not sure if that was intentional).

On 8 July 2014 10:26, Antony Polukhin wrote:

2014-07-08 12:44 GMT+04:00 Andrey Semashev :

...

On Tuesday 08 July 2014 09:58:11 Oliver Kowalke wrote:

...

2014-07-08 9:53 GMT+02:00 Antony Polukhin :

...

Links to the docs are broken for the following libraries in libs/libraries.htm: Algorithm, Align, CompressedPair, Context, Core, Coroutine, EnableIf, Functional/factory, Functional/forward Log, Odeint, Phoenix, Range, Spirit, Swap, TTI, TypeIndex, Units

does this issue require action on the side of the library authors (like generating and check-in of html etc.)?

No autogenerated files in git, please. I believe these test archives simply don't include all docs (not sure if that was intentional).

Looks more like missing redirects or infinite loop of redirects.

It's because it doesn't include the documentation from: http://boost.cowic.de/rc/boost-master-docs.7z It needs to be downloaded and extracted inside the tree before creating the archive.

A few more broken links in tools/index.html: Boost.Build, Release.

Those are 2 of the 766 broken links listed in the inspect report: http://boost.cowic.de/rc/docs-inspect-master.html

On Tuesday 08 July 2014 13:10:19 Paul A. Bristow wrote:

...

-----Original Message----- From: Boost [mailto:boost-bounces@lists.boost.org] On Behalf Of Antony

Polukhin

...

Sent: 08 July 2014 10:27 To: boost@lists.boost.org List Subject: Re: [boost] Test build for 1.56 available

2014-07-08 12:44 GMT+04:00 Andrey Semashev :

...

On Tuesday 08 July 2014 09:58:11 Oliver Kowalke wrote:

...

2014-07-08 9:53 GMT+02:00 Antony Polukhin :

...

Links to the docs are broken for the following libraries in libs/libraries.htm: Algorithm, Align, CompressedPair, Context, Core, Coroutine, EnableIf, Functional/factory, Functional/forward Log,

Odeint,

...

...

Phoenix, Range, Spirit, Swap, TTI, TypeIndex, Units

does this issue require action on the side of the library authors (like generating and check-in of html etc.)?

No autogenerated files in git, please. I believe these test archives simply don't include all docs (not sure if that was intentional).

Looks more like missing redirects or infinite loop of redirects.

Andrey is absolutely right about autogenerated files in git. About a half of the

...

archive is consumed by the autogenerated docs.

It would be good to have "./b2 docs" in next release and remove all the

generated

...

docs from git.

I agree - I've found that any generated items in git make maintenance disastrously messy.

But we also need to ensure that the setup of the tools, Quickbook, Autoindex and Inspect are all automatically built to a 'standard' location, and in the users path, so that the

./b2 docs works slickly for everyone.

It will of course take quite a long time - hours ? - to regenerate all the docs :-(

Personally, I think distributing pre-built docs as a separate package is a better option. Building docs locally require toolchain that is not distributed or built with Boost, so offering users to build the docs themselves is not quite practical. At the same time, I'm sure most users won't use bre-built docs and just go online, so packaging docs with the sources seems like a waste.

On Tuesday 08 July 2014 09:42:37 Edward Diener wrote:

On 7/8/2014 9:04 AM, Andrey Semashev wrote:

...

Personally, I think distributing pre-built docs as a separate package is a better option. Building docs locally require toolchain that is not distributed or built with Boost, so offering users to build the docs themselves is not quite practical. At the same time, I'm sure most users won't use bre-built docs and just go online, so packaging docs with the sources seems like a waste.

I believe most users will execute the index.html in the top-level directory and expect to be able to view the documentation for any Boost library in which they are interested from the Libraries link there. To think that we will distribute versions of Boost without this is foolish IMO.

Do you actually use pre-built docs? I don't. Neither do most of my coworkers. I don't see why cutting down the archive I need to download by more than a half is foolish.

On Tuesday 08 July 2014 11:58:56 Edward Diener wrote:

On 7/8/2014 10:07 AM, Andrey Semashev wrote:

...

On Tuesday 08 July 2014 09:42:37 Edward Diener wrote:

...

On 7/8/2014 9:04 AM, Andrey Semashev wrote:

...

Personally, I think distributing pre-built docs as a separate package is a better option. Building docs locally require toolchain that is not distributed or built with Boost, so offering users to build the docs themselves is not quite practical. At the same time, I'm sure most users won't use bre-built docs and just go online, so packaging docs with the sources seems like a waste.

I believe most users will execute the index.html in the top-level directory and expect to be able to view the documentation for any Boost library in which they are interested from the Libraries link there. To think that we will distribute versions of Boost without this is foolish IMO.> Do you actually use pre-built docs? I don't. Neither do most of my coworkers. I don't see why cutting down the archive I need to download by more than a half is foolish.

I am fine with the main distribution not having the docs as long as there is a separate docs download.

That is what I suggested.

On 9/07/2014 01:04, Andrey Semashev wrote:

Personally, I think distributing pre-built docs as a separate package is a better option. Building docs locally require toolchain that is not distributed or built with Boost, so offering users to build the docs themselves is not quite practical. At the same time, I'm sure most users won't use bre-built docs and just go online, so packaging docs with the sources seems like a waste.

+1. As a user I've always just looked at the online docs, so having them in the download is just a waste of space. But I expect that many users will want to have offline docs, so making it a separate package is the best of both worlds. (Ideally, when modular boost really kicks off, each library should have a separate docs package so that the auto-library-downloading-tool can install only those docs that are actually required for the library subset in use.)

Tom — The question I need to know from this is: * Is this a problem with the build process (I don’t think so) or * Does the problem exist in the git repo (probably) [ But - I’m not seeing any failures for VC12 in the test matrix. ] Are you set up to build from git? (the master branch, in particular)? If so, can you check and see if the problem exist there? [ We need to fix them in either case, but knowing where they came from will help a lot ] Thanks! — Marshall On Tue, Jul 8, 2014 at 6:18 AM, Tom Kent wrote:

On Mon, Jul 7, 2014 at 1:56 PM, Marshall Clow wrote:

...

Ok, here are rc2s.

https://dl.dropboxusercontent.com/u/11048326/boost/boost-1_56_0a1c2.tar.gz

...

Please check them out, let me know what problems you have. Thanks to people who checked out the previous builds and reported

problems.

...

Builds on windows (Azure VM, 8GB Ram, Win 2008r2 - 64bit):

msvc-8.0 - 32 - builds, links, runs msvc-8.0 - 64 - builds, links, runs msvc-9.0 - 32 - builds, links, runs msvc-9.0 - 64 - builds, links, runs msvc-10.0 - 32 - builds, links, runs msvc-10.0 - 64 - builds, links, runs msvc-11.0 - 32 - builds, links, runs msvc-11.0 - 64 - builds, links, runs msvc-12.0 - 32 - builds - with errors in signals (see below), links, runs msvc-12.0 - 64 - builds - with errors in signals (see below), links, runs

Errors on msvc-12.0 32&64

D:\local\BoostBuilding\boost-1_56_0a1c2\boost/iterator/iterator_facade.hpp(526) : error C2039: 'decrement' : is not a member of 'boost::signals::detail::named_slot_map_iterator'

D:\local\BoostBuilding\boost-1_56_0a1c2\boost/signals/detail/named_slot_map.hpp(80) : see declaration of 'boost::signals::detail::named_slot_map_iterator'

D:\local\BoostBuilding\boost-1_56_0a1c2\boost/iterator/iterator_facade.hpp(544) : error C2039: 'advance' : is not a member of 'boost::signals::detail::named_slot_map_iterator'

D:\local\BoostBuilding\boost-1_56_0a1c2\boost/signals/detail/named_slot_map.hpp(80) : see declaration of 'boost::signals::detail::named_slot_map_iterator'

These two seem to repeat a few times.

Full build logs are here: http://boost.teeks99.com/misc/boost-1_56_0a1c2-32bitlog.txt http://boost.teeks99.com/misc/boost-1_56_0a1c2-64bitlog.txt

Tom

_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

On 08.07.2014 18:05, wrote Marshall Clow:

The question I need to know from this is: * Is this a problem with the build process (I don’t think so) or * Does the problem exist in the git repo (probably) [ But - I’m not seeing any failures for VC12 in the test matrix. ]

Are you set up to build from git? (the master branch, in particular)? If so, can you check and see if the problem exist there? [ We need to fix them in either case, but knowing where they came from will help a lot ]

FWIW: The compile fails for a deprecated lib i.e. signals, while the newer signals2 seemingly does not trigger an error. I tried to compile with the switch: --without-signals Could this be an explanation while the regression does not show up as an error? BTW: How is a deprecated lib expected to be handled for a "new" compiler? Regards, Roland

Hmm - master should still be open for bug fixes. I assumed that commits pushed to master will be included in the release.