GMP, MPFR and VS version compatibility
Michael.Abshoff at fsmath.mathematik.uni-dortmund.de
Fri Jul 13 00:36:34 CEST 2007
Jim White wrote:
> "Michael"'s description of his MSVS build problems
> lend further support to my argument that using Visual
> Studio to build libraries like GMP and MPFR is asking
> for trouble.
Well, developing non-trivial software on Windows that is supposed to be
cross platform is no walk in the park, and having to deal with Visual
Sutdio 200[2|3|5] on top of that doesn't make things any easier.
Microsoft's tendency to force upgrades with incompatible changes doesn't
> Brian will, I hope, forgive me (again) for "dissing"
> (again) this approach, but having to re-tool your
> projects because of changes in MSVS would seem to me
> to be an ominous portent for ongoing maintenance!
This whole issue could be sidestepped by using an nmake based build system
or integrating support for MSVC and yasm into the autotool toolchain. LyX
has been doing that (using cl.exe with autotools) for well over a year, it
would just require for the user to have MinGW+Msys installed on the
system. I am certainly not going to do that any time soon (and I doubt
anybody else would volunteer) - at least as long as Brian does the great
work he is doing with the Visual Studio port.
I really like gcc and MinGW, but the main problem with gcc in all its
forms on Windows remains the same for the last couple years: No 64 bit
support (except for Cygwin/Itanium and who cares about that) and the C++
ABI issue when linking libraries compiled with MSVC.
I get payed to work in an open source project with a big, big corporate
client that runs our software (that uses gmp) on Windows and 64 bit
support is essential. Suggesting that they switch to Linux or MacOSX would
get us laughed out of the door. The technical people don't like the fact
that it has to run on Windows, but it remains a fact. Hopefully I will get
the company to buy a gmp support contract in the future, because they
understand that open source people have to pay a mortgage, too.
> Jim White
More information about the gmp-discuss