Experimental Physics and
Industrial Control System

Mark Rivers <[email protected]> · Tue, 12 Dec 2017 02:15:09 +0000

I think we have said the VS2010 is the oldest supported, and that is reasonable.  Since the Community editions are free now we could even move the minimum up to more recent versions if there is a good reason?

I wonder how useful Cygwin is any more?  I used to support it.  But then some version of Cygwin 2-3 years ago stopped being able to link with vendor supplied .lib files built with Visual Studio.  This makes it pretty useless for any hardware control, since one needs to link with vendor libraries and they are normally built with Visual Studio.

Mark

________________________________
From: [email protected] <[email protected]> on behalf of Michael Davidsaver <[email protected]>
Sent: Monday, December 11, 2017 7:43 PM
To: Ralph Lange; EPICS Core Talk
Subject: Re: Appveyor-CI builds are back

Ok, so appveyor finished chewing on this and I see three failures.

> ..\pv/pvType.h(38) : fatal error C1083: Cannot open include file: 'stdint.h': No such file or directory

The 9.0 (aka VS 2008) builds fail trying to include "stdint.h".  As this is the oldest
version I first ask the obvious question.  Should we just drop 9.0?

The others are cygwin problems...

> ../../src/misc/pv/anyscalar.h: In instantiation of 'epics::pvData::AnyScalar::AnyScalar(T) [with T = double]':
> ../../src/misc/pv/anyscalar.h:146:9:   required from 'epics::pvData::AnyScalar& epics::pvData::AnyScalar::operator=(T) [with T = double]'
> ../pv/pvData.h:502:11:   required from 'void epics::pvData::PVScalarValue<T>::getAs(epics::pvData::AnyScalar&) const [with T = double]'
> ../../src/pv/valueBuilder.cpp:295:2:   required from here
> /cygdrive/c/projects/epics-base-5e851/include/epicsAssert.h:48:29: error: static assertion failed: sizeof(TT)<=sizeof(_wrap.blob)
>  #define STATIC_ASSERT(expr) static_assert(expr, #expr)
>                              ^
> ../../src/misc/pv/anyscalar.h:96:9: note: in expansion of macro 'STATIC_ASSERT'
>          STATIC_ASSERT(sizeof(TT)<=sizeof(_wrap.blob));

2. This could only be failing if the c++11 implementation is used and "sizeof(double) > sizeof(std::string).
I'm curious to know how this can be true at all, and why only on cygwin x86.

FYI. I would like to use std::aligned_union here, but switched to std::aligned_storage for compatibility.
One of the gcc versions/variants (mingw I think) is missing aligned_union.

Anyway the fix is clear.

> /usr/lib/gcc/x86_64-pc-cygwin/6.4.0/libstdc++.a(cow-stdexcept.o): In function `std::logic_error::logic_error(std::logic_error const&)':
> /usr/src/debug/gcc-6.4.0-4/libstdc++-v3/src/c++11/cow-stdexcept.cc:59: multiple definition of `std::logic_error::logic_error(std::logic_error const&)'
> testBaseException.o:/usr/lib/gcc/x86_64-pc-cygwin/6.4.0/include/c++/stdexcept:113: first defined here

3. Maybe some inline-ing failure?  Someone with a local cygwin install will have to take the lead here.

On 12/11/2017 09:36 AM, Ralph Lange wrote:
> The builder for the epics-base project currently hangs. I have opened a support ticket for this; it seems to be a common issue and is always fixed very quickly.
> Until we're up again ... the build against my (identical) personal clone is progressing: https://ci.appveyor.com/project/ralphlange/epics-base-5e851
>
> Cheers,
> ~Ralph
>
>
> On Mon, Dec 11, 2017 at 11:30 AM, Ralph Lange <[email protected] <mailto:[email protected]>> wrote:
>
>     Due to popular demand,
>     I fixed the Appveyor builds for EPICS Base 7.0.1. You will find results for 48 different Windows builds soon at
>         https://ci.appveyor.com/project/epics-base/epics-base <https://ci.appveyor.com/project/epics-base/epics-base>
>
>     Cheers,
>     ~Ralph
>
>

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System