Re: [Plplot-devel] Intermittent Ada example build failures are now back again (SOLVED by gnat-8)

[Alan W. I. <Ala...@gm...>]

On 2018-09-14 12:19-0000 Arjen Markus wrote:

> Hi Alan,
>
>> -----Original Message-----
>> From: Alan W. Irwin [mailto:Ala...@gm...]
>> To Jerry and Arjen:
>>
>> On Sunday (just before commit 7ec926fae) I discovered that my installs of gdc-7
>> and gnat-7 were probably not good choices because although I had both gcc-7 and
>> gcc-8 installed, PLplot was being built with gcc-8 rather than gcc-7 on my Debian
>> Buster system.  So in this case at least Debian does not have tight version
>> consistency requirements between either gdc or gnat and gcc.  And perhaps that is
>> justified since previously I was getting perfect results with the combination of gdc-7
>> and gcc-8.
>> However, the above parallel build issue with gnat-7 encouraged me to try gnat-8,
>> and I decided to install gdc-8 as well just to be consistent with gcc-8.
>> It turns out (see the above commit message for the details) the combination of gdc-
>> 8, gnatmake-8, and gcc-8 produces perfect massively
>> (-j10) parallel build and test results
>> for a comprehensive set of of configurations and build trees, and for three different
>> CMake versions.
>>
>> So it seems once again that on Linux at least, parallel builds of our Ada bindings
>> and examples work well, and I advise on other platforms to avoid gnat-7 in all cases,
>> to use parallel builds with caution, and at least for the sake of consistency use the
>> same version of gdc, gnat, and gcc (e.g., gdc-8, gnat-8, and gcc-8).
>>
>
> I just checked the Cygwin set-up: there is at this moment at least no version 8 available within that environment. I do not know if there are any plans to incorporate it, but I will keep an eye out.

Indeed, from <https://cygwin.com/cgi-bin2/package-grep.cgi> a search
for gcc shows that currently the maximum gcc version available is
gcc-7 and similarly gnat-7 is the maximum version of gnat that is
available.  So you might want to try those two (in fact your previous
tests might have already been done with those two), but from my Linux
experience I would not bother with parallel builds with gnat-7.  But
as I recall, Cygwin has other troubles with parallel builds so that might
be a moot constraint.

In contrast to the Cygwin case, <http://repo.msys2.org/mingw/x86_64/>
shows for the MinGW-w64/MSYS2 case that the maximum gcc and gnat
versions are 8.2.0 which is also the versions of those two software
packages I used recently for my successful Debian Buster tests.  So
the next time you test this platform I suggest you try version 8 of
gcc and gnat if you are not doing that already.  And since parallel
builds and tests worked so well for the Linux case you might want to
try those on this platform especially if those have been successful
before.

By the way, it is a bit of a concern for the Cygwin rolling release
that the CMake, gcc, and gnat versions are getting pretty far behind
what is available with MinGW-w64/MSYS2, but, of course, those
packaging delays could all be special cases.  Does the overall health
of the Cygwin project seem fine (e.g., all legitimate questions
answered in a timely way by Cygwin developers, and packaging progress
continuing to be made for a large number of packages) from what you
have been reading on the Cygwin mailing list?

Alan
__________________________
Alan W. Irwin

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________