PyCrypto builds neither with MSVC nor MinGW

[Chris Angelico]

On Mon, Feb 6, 2012 at 6:39 PM, Terry Reedy <tjreedy at udel.edu> wrote:
> On 2/6/2012 1:53 AM, Chris Angelico wrote:
>> I suppose there's no chance of moving to a free compiler?
>
> VC express is free-as-in-beer. The whole V. Studio is free to core
> developers. MS may not *like* open-source software, but they have decided
> they would like it even less if everyone compiled it with non-MS compilers.

Oh, that's something at least. I wasn't aware of what exactly they
charge for and what they don't.

>> Windows work, I've generally used the Open Watcom compiler; that's not
>> to say it's the best, but it does the job, and it's free software.
>
> Would it build CPython, including the +- dependent libraries like tcl/tk?
> How would the speed compare?

I can't answer that question without grabbing the sources, going
through the whole work of porting makefiles etc, and finding out
whether there's failures - in other words, doing the whole job. It's
entirely possible that there'll be some dependency failure; but I
would posit that, on balance, it's more likely there won't be.

As to speed - I've not done a lot of compiler benchmarking. (Not sure
whether you mean compilation speed or the efficiency of the resulting
code; either way, I've not tried.) Never actually had multiple
compilers on any one platform for long enough to do serious testing.
It's hardly fair to compare Borland C++ for Windows 3.1, icc on OS/2
32-bit, Open Watcom on XP, and gcc on Debian 64-bit!

It's probably not worth the hassle of changing compilers, although I
do wonder whether changing compiler _versions_ isn't sometimes nearly
as much work. ("What? All that legacy code doesn't compile any more?
Ohh... it doesn't like #include <iostream.h> any more...")

ChrisA