[Scilab-Dev] Scilab 6.1.0 is available!
Sylvain Corlay
sylvain.corlay at gmail.com
Mon Apr 13 11:55:58 CEST 2020
Hello All,
I am following up on this.
Is there any hope for a patch Scilab release that would support gcc < 8 ?
This regression makes packaging the latest version of scilab impossible on
conda-forge, which is probably the largest scientific software distribution.
Thanks,
Sylvain
On Thu, Mar 5, 2020 at 11:56 AM Sylvain Corlay <sylvain.corlay at gmail.com>
wrote:
> Hi Clément,
>
> Indeed, this is what I ended up doing - and there are a couple of
> instances of the same error in various places.
>
> This appears to be a legitimate error C++ even though this passes with GCC
> 8.
>
> On the choice of dropping GCC<8, this seems a bit early, even if the
> latest flavors of Ubuntu, or Debian Sid already have GCC8 available. GCC6
> and GCC7 are recent compilers and people target older distributions in
> their builds.
>
> Best,
>
> Sylvain Corlay
>
> On Thu, Mar 5, 2020 at 11:41 AM Clément David <Clement.David at esi-group.com>
> wrote:
>
>> Hello Sylvain,
>>
>> I took a look at the trace, it looks like g++ 7 is confused with the
>> xml.h include ; as libxml2 is already protected with ifdef __cplusplus
>> extern "C" you might just move the #include <xml.h> from XMLDocument.hxx:28
>> out of the extern "C".
>>
>> Sorry about that but gcc-7 is not our target for official releases,
>> Debian/Ubuntu/Fedora all ship gcc-8 with a more stable c++17 support. I
>> could only help to debug issues on the conda side.
>>
>> Regards,
>>
>> --
>> Clément
>>
>> > -----Original Message-----
>> > From: dev <dev-bounces at lists.scilab.org> On Behalf Of Sylvain Corlay
>> > Sent: Thursday, March 5, 2020 10:15 AM
>> > To: List dedicated to the development of Scilab <dev at lists.scilab.org>
>> > Subject: Re: [Scilab-Dev] Scilab 6.1.0 is available!
>> >
>> > Hello,
>> >
>> > Is there any update on the support of GCC 7 for Scilab 6.1?
>> >
>> > This is a major blocker for us to package it for the conda ecosystem.
>> >
>> > Best,
>> >
>> > On Thu, Feb 27, 2020 at 4:12 PM Sylvain Corlay <
>> sylvain.corlay at gmail.com
>> > <mailto:sylvain.corlay at gmail.com> > wrote:
>> >
>> >
>> > Patching the source to use <experimental/filesystem> resolves that
>> > particular issue with C++17 in GCC7.
>> >
>> > However, the Scilab build now fails with many occurrences of the
>> error:
>> >
>> > error: template with C linkage
>> >
>> >
>> > - You can see the raw build logs here:
>> https://dev.azure.com/conda-
>> > forge/84710dde-1620-425b-80d0-
>> > 4cf5baca359d/_apis/build/builds/125787/logs/7
>> > - The scilab recipe update including the patches for <filesystem>
>> is
>> > available here: https://github.com/conda-forge/scilab-feedstock/pull/11
>> >
>> > Officially supporting older versions of GCC than GCC 8 (which is
>> from
>> > May 2018) would be greatly appreciated.
>> >
>> >
>> > Best,
>> >
>> >
>> > On Thu, Feb 27, 2020 at 2:28 PM Sylvain Corlay
>> > <sylvain.corlay at gmail.com <mailto:sylvain.corlay at gmail.com> > wrote:
>> >
>> >
>> > Hi Clément,
>> >
>> > Regarding the Java, we already skip xcos in the build but
>> we
>> > were really hoping that we could start including it with this version.
>> >
>> > For the filesystem thingy, I will be patching the source
>> as part of
>> > the conda recipe to use <experimental/filesystem> instead and see how
>> it goes -
>> > although it would be really nice if we could support more compilers out
>> of the
>> > box.
>> >
>> > Sylvain
>> >
>> > On Thu, Feb 27, 2020 at 2:25 PM Clément David
>> > <Clement.David at esi-group.com <mailto:Clement.David at esi-group.com> >
>> > wrote:
>> >
>> >
>> > Hello Sylvain,
>> >
>> > First, thank you for your work on the conda
>> packaging.
>> > The c++17 requirements is only needed for a single file that is used to
>> implement
>> > fullpath() (named fullpath.cpp). The used API is reduced to
>> > std::filesystem::weakly_canonical and std::filesystem::absolute [1]. I
>> guess using
>> > a light patch might relax the use of the filesystem header, for example,
>> > something like [2].
>> >
>> > About the Java8 requirement, I guess you could
>> > ./configure --without-xcos as a first approach. I started porting the
>> Java code
>> > out of javax.xml.bind [3] but that's very repetitive and error prone
>> work.
>> >
>> > [1]:
>> >
>> https://codereview.scilab.org/#/c/21041/25/scilab/modules/fileio/src/cpp/fullp
>> > ath.cpp
>> > [2]:
>> > https://stackoverflow.com/questions/45867379/why-does-gcc-not-seem-to-
>> > have-the-filesystem-standard-library
>> > [3]: https://codereview.scilab.org/#/c/20630/
>> >
>> > Regards,
>> >
>> > --
>> > Clément
>> >
>> > > -----Original Message-----
>> > > From: dev <dev-bounces at lists.scilab.org
>> > <mailto:dev-bounces at lists.scilab.org> > On Behalf Of Sylvain Corlay
>> > > Sent: Thursday, February 27, 2020 1:50 PM
>> > > To: List dedicated to the development of Scilab
>> > <dev at lists.scilab.org <mailto:dev at lists.scilab.org> >
>> > > Subject: Re: [Scilab-Dev] Scilab 6.1.0 is
>> available!
>> > >
>> > > Congratulations on the release.
>> > >
>> > > I am the author of the conda package for
>> scilab, and
>> > unfortunately, it does not
>> > > seem that 6.1.0 can be successfully packaged for
>> > conda-forge with the new
>> > > requirement for the <filesystem> C++17 header,
>> > which requires GCC 8.
>> > >
>> > > Conda-forge is still based on GCC 7, which is
>> fairly
>> > recent, with C++17 enabled by
>> > > default. Would you consider not using the
>> > <filesystem> from the C++17 standard
>> > > so that Scilab can be made available to a wider
>> > audience?
>> > >
>> > > Another blocker to the packaging of Scilab is
>> the
>> > outdated version of java that is
>> > > required by the GUI. Is there any plan to
>> support a
>> > more recent version of
>> > > OpenJDK?
>> > >
>> > > Best,
>> > >
>> > > Sylvain Corlay
>> > >
>> > > On Tue, Feb 25, 2020 at 2:10 PM Clément David
>> > <Clement.David at esi-
>> > > group.com <http://group.com>
>> > <mailto:Clement.David at esi-group.com <mailto:Clement.David at esi-
>> > group.com> > > wrote:
>> > >
>> > >
>> > > Dear Scilab-ers,
>> > >
>> > >
>> > >
>> > > A brand new Scilab 6.1.0
>> > <https://www.scilab.org/download/6.1.0> is
>> > > released today!
>> > >
>> > >
>> > >
>> > > This version includes further improvement
>> atop
>> > Scilab 6.0 for better
>> > > stability and increased algorithm performance.
>> It also
>> > includes a reworked
>> > > display for more compact and meaningful value
>> > printing; web tools for HTTP,
>> > > JSON support; better debug support and various
>> > algorithm rewrite/extension.
>> > >
>> > >
>> > >
>> > > This first iteration of the 6.1 branch
>> fixes up to 245
>> > bugs and implements
>> > > missing features from the 6.0.2 version. We
>> would
>> > like to give a special thanks
>> > > to Samuel and Stephane who have been very active
>> > this year.
>> > >
>> > >
>> > >
>> > > If you find any critical issue or
>> instability that might
>> > need a 6.1.x release
>> > > please alert us <https://bugzilla.scilab.org/>
>> . If you
>> > are a toolbox maintainer,
>> > > please rebuild your code, upgrade it when
>> needed and
>> > publish it to
>> > > atoms.scilab.org <http://atoms.scilab.org>
>> > <https://atoms.scilab.org/> .
>> > >
>> > >
>> > >
>> > > For the complete list of changes and bugs
>> fixed,
>> > please take a look at
>> > > the CHANGES <https://help.scilab.org/CHANGES>
>> > file.
>> > >
>> > >
>> > >
>> > > --
>> > >
>> > > Clément on behalf of the Scilab team
>> > >
>> > >
>> > _______________________________________________
>> > > dev mailing list
>> > > dev at lists.scilab.org <mailto:
>> dev at lists.scilab.org>
>> > <mailto:dev at lists.scilab.org <mailto:dev at lists.scilab.org> >
>> > >
>> http://lists.scilab.org/mailman/listinfo/dev
>> > >
>> >
>> >
>> > _______________________________________________
>> > dev mailing list
>> > dev at lists.scilab.org <mailto:dev at lists.scilab.org
>> >
>> > http://lists.scilab.org/mailman/listinfo/dev
>> >
>>
>> _______________________________________________
>> dev mailing list
>> dev at lists.scilab.org
>> http://lists.scilab.org/mailman/listinfo/dev
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/dev/attachments/20200413/24ee4b47/attachment.htm>
More information about the dev
mailing list