[shedskin] dropping windows support

15 views
Skip to first unread message

srepmub

unread,
Mar 11, 2010, 9:12:59 AM3/11/10
to shedskin-discuss
hi all,

I'm considering dropping windows support with shedskin 0.4. if you'd
like to volunteer to maintain the windows version of shedskin, please
let me know.


mark.

Jérémie Roquet

unread,
Mar 12, 2010, 4:28:22 AM3/12/10
to shedskin...@googlegroups.com

Mark Dufour

unread,
Mar 12, 2010, 6:55:58 AM3/12/10
to shedskin...@googlegroups.com

2010/3/12 Jérémie Roquet <arka...@gmail.com>

essentially, yes, but also testing to see if (most) tests and example programs still work. since everything worked fine with 0.3, doing a 0.4 release _should_ only take a few hours. it would be useful to also upgrade MingW at some point, because the version currently shipped with shedskin is getting a bit outdated, but that would take some more time..


thanks,
mark.
--
"Overdesigning is a SIN. It's the archetypal example of what I call 'bad taste'" - Linus Torvalds

Zahari Zahariev

unread,
Mar 23, 2010, 3:43:07 AM3/23/10
to shedskin...@googlegroups.com
Yeah!

I really enjoy cool piece of software like ShedSkin that is really easy to get and use in Linux but has no windows version.

You rock!



--
You received this message because you are subscribed to the Google Groups "shedskin-discuss" group.
To post to this group, send email to shedskin...@googlegroups.com.
To unsubscribe from this group, send email to shedskin-discu...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/shedskin-discuss?hl=en.

Mark Dufour

unread,
Mar 23, 2010, 5:09:56 AM3/23/10
to shedskin...@googlegroups.com
I really enjoy cool piece of software like ShedSkin that is really easy to get and use in Linux but has no windows version.

yes, I enjoy that too, although marketing-wise it's probably a disaster.. though not as big a disaster as one might expect considering windows market share: only about 1/3 of the downloads is currently for windows. this is just not enough for me anymore to justify going through the misery each release of testing and fixing things under windows.

in any case, I've made the final decision to remove windows support, and have already removed the windows instructions from the tutorial.


Paul Boddie

unread,
Mar 23, 2010, 5:04:53 PM3/23/10
to shedskin-discuss
On 23 Mar, 10:09, Mark Dufour <mark.duf...@gmail.com> wrote:
>
> yes, I enjoy that too, although marketing-wise it's probably a disaster..
> though not as big a disaster as one might expect considering windows market
> share: only about 1/3 of the downloads is currently for windows. this is
> just not enough for me anymore to justify going through the misery each
> release of testing and fixing things under windows.

This is where the Windows people just have to step up and help out. If
people use a platform, whether it's Windows, Linux, Mac OS X, Solaris
or even the delights of AIX or HP-UX, they can't expect you to run
around making sure it works on all of them, let alone provide a one-
click installer. Open source is about having to unroll the red carpet
yourself sometimes. :-)

Paul

P.S. I had some helpful advice on Shed Skin packaging for Debian, but
I'm currently stalled doing other things. One important thing which is
hopefully resolved is the licensing: is everyone's code now accounted
for and licensed suitably? That would be a big Debian show-stopper
otherwise.

Mark Dufour

unread,
Mar 24, 2010, 5:15:59 AM3/24/10
to shedskin...@googlegroups.com
P.S. I had some helpful advice on Shed Skin packaging for Debian, but
I'm currently stalled doing other things. One important thing which is
hopefully resolved is the licensing: is everyone's code now accounted
for and licensed suitably? That would be a big Debian show-stopper
otherwise.

I think everything is suitably licensed now. the licensing problems that were found by the Fedora team have been solved in any case.

something else I feel should be done (though perhaps not strictly necessary to get shedskin into debian/fedora), is to check if all the library module implementations have appropriate copyright statements..


thanks,
mark.

Oyster

unread,
Mar 29, 2010, 11:43:02 AM3/29/10
to shedskin-discuss
bad news :(

Andy

unread,
Apr 18, 2010, 9:58:19 AM4/18/10
to shedskin-discuss
I've done some hacking over the weekend and have shedskin working
'well' under windows using MSVC C++ 2010 Express edition (almost
certain that MSVC 2008 will NOT work due to template and STL support).

I've posted a patch file (http://code.google.com/p/shedskin/issues/
detail?id=63) -- it makes a number of changes to function names etc so
need Mark's approval..

If anyone gets a chance to try it (as I've probably messed something
up) I'm happy to help out.

Andy

Jérémie Roquet

unread,
Apr 19, 2010, 4:13:54 AM4/19/10
to shedskin...@googlegroups.com
2010/4/18 Andy <nzmi...@gmail.com>:
> I've done some hacking over the weekend and have shedskin working
> 'well' under windows using MSVC C++ 2010 Express edition (almost
> certain that MSVC 2008 will NOT work due to template and STL support).
> I've posted a patch file (http://code.google.com/p/shedskin/issues/
> detail?id=63) -- it makes a number of changes to function names etc so
> need Mark's approval..

Thanks Andy!

--
Jérémie

Mark Dufour

unread,
Apr 19, 2010, 10:45:15 AM4/19/10
to shedskin...@googlegroups.com
On Sun, Apr 18, 2010 at 3:58 PM, Andy <nzmi...@gmail.com> wrote:
I've done some hacking over the weekend and have shedskin working
'well' under windows using MSVC C++ 2010 Express edition (almost
certain that MSVC 2008 will NOT work due to template and STL support).

yep, thanks! :-) good work. I will try to have a good look at your patch later this week. some comments already:

- someone else has almost finished building a new MinGW package for shedskin. it would be nice if we could support both MSVC and GCC.
- would using three underscores conflict with MSVC, so for example, ___box..? somehow the 'T' looks a bit ugly..
- have you tried to build an extension module with MSVC yet (shedskin -e)..?


thanks again!

Andy Miller

unread,
Apr 19, 2010, 11:20:15 PM4/19/10
to shedskin-discuss
To confirm the triple underscores work fine as does building as an extension (small changes needed to cpp.py to generate the correct makefile)..

Andy

On 20 April 2010 06:44, Andy Miller <nzmi...@gmail.com> wrote:
- I don't see an issue with supporting MinGW and MSVC -- I'd simply go back and change some of the 'ifdef WIN32's to ifdef _MSC_VER in places that are specific to MSVC
- 3 underscores should be fine and yes would probably look nicer :) 
- Haven't tried an extension module -- will do so and let you know how it goes.

Andy

Andy Miller

unread,
Apr 19, 2010, 6:44:31 PM4/19/10
to shedskin-discuss
- I don't see an issue with supporting MinGW and MSVC -- I'd simply go back and change some of the 'ifdef WIN32's to ifdef _MSC_VER in places that are specific to MSVC
- 3 underscores should be fine and yes would probably look nicer :) 
- Haven't tried an extension module -- will do so and let you know how it goes.

Andy


On 19 April 2010 22:45, Mark Dufour <mark....@gmail.com> wrote:

Mark Dufour

unread,
Apr 20, 2010, 12:06:29 PM4/20/10
to shedskin...@googlegroups.com
jeremie,

have you seen the problem under windows with test 191, where open('rU') doesn't seem to work.. ? do you think that would be difficult to fix?


thanks,
mark.

Jérémie Roquet

unread,
Apr 20, 2010, 12:56:10 PM4/20/10
to shedskin...@googlegroups.com
2010/4/20 Mark Dufour <mark....@gmail.com>:
> jeremie,
> have you seen the problem under windows with test 191, where open('rU')
> doesn't seem to work.. ? do you think that would be difficult to fix?

Sorry, I haven't looked at the patch yet.

I guess it may be the result of the 'U' not being removed from the
parameter string given to fopen()...
Unfortunately, I can't test on Windows. May someone test the attached
patch, please?

Note: I also fixed two compilation issues in builtin.hpp

Best regards,

--
Jérémie
open-universal.2.patch

Andy Miller

unread,
Apr 20, 2010, 1:58:50 PM4/20/10
to shedskin-discuss
Mark

If you are OK with the direction I'm taking with the Windows support then I'm happy to review the library support in general as there are certainly issues to be resolved (I 'def' out quite a bit of code in os and sys)...

Also happy to do regular testing etc, only issue is that due to real work I'm often out of touch for a week or two at a time.. :(

Andy

2010/4/21 Jérémie Roquet <arka...@gmail.com>

Mark Dufour

unread,
Apr 21, 2010, 6:47:45 AM4/21/10
to shedskin...@googlegroups.com
I guess it may be the result of the 'U' not being removed from the
parameter string given to fopen()...
Unfortunately, I can't test on Windows. May someone test the attached
patch, please?

thanks for the patch! I tested it, but I think I see the same problem: the python version adds two empty lines to the output (for the last with-test), while the shedskin version leaves them out:

---
hello world

bye

==


Note: I also fixed two compilation issues in builtin.hpp

thanks, committed :)

btw, another small thing you might want to look into, is that in builtin.cpp there are some '-1' expressions, where we'd probably want to use std::string::npos (iirc). this leads to signed/unsigned warnings under the new windows/mingw package (without enabling all warnings). problem is, when I change them to use std::string::npos, some tests start failing, and I'm not sure how to fix this correctly.

in any case, thanks for looking into the open('U') problem,

mark.
--
"Overdesigning is a SIN. It's the archetypal example of what I call 'bad taste'" - Linus Torvalds

Mark Dufour

unread,
Apr 21, 2010, 7:00:15 AM4/21/10
to shedskin...@googlegroups.com
If you are OK with the direction I'm taking with the Windows support then I'm happy to review the library support in general as there are certainly issues to be resolved (I 'def' out quite a bit of code in os and sys)...

well, I think the easiest way to support Windows is using MinGW, because it's UNIX and uses GCC, which is in many ways closer to GNU/Linux, and because the current libraries are also already tested against MinGW. so in that sense, I'm not so sure MSVC would be the "right" way to support windows (if that's what you mean by direction). but of course, I'd be happy to incorporate patches for an MSVC version, if you're passionate about that, and wouldn't mind taking the time to get the patches past me.. :)


thanks!

mark.
--
"Overdesigning is a SIN. It's the archetypal example of what I call 'bad taste'" - Linus Torvalds

Andy Miller

unread,
Apr 21, 2010, 11:46:06 PM4/21/10
to shedskin-discuss
I'll take a deeper look into the libs and if I can get them working under MSVC without a huge amount of work then I'm happy to help with supporting it longer term as I think this is a more suitable solution for typical windows users.. 

I will go back and look at the changes I made as they probably (in places) break MinGW as I need to check for _MS_VER instead of WIN32..

Cheers
Andy

John Yeung

unread,
Apr 22, 2010, 2:43:01 AM4/22/10
to shedskin...@googlegroups.com
On Wed, Apr 21, 2010 at 11:46 PM, Andy Miller <nzmi...@gmail.com> wrote:
> I'll take a deeper look into the libs and if I can get them
> working under MSVC without a huge amount of work then
> I'm happy to help with supporting it longer term as I think
> this is a more suitable solution for typical windows users..

First of all, let me say that I am very glad someone with both the
interest and the ability to support Shed Skin on Windows is taking up
the cause. (Perhaps even more than one someone, based on comments
Mark made earlier.)

Obviously the *most* typical Windows user, even one who uses Python,
isn't going to have any C or C++ compiler at all. But between MinGW
and MSVC, I'd say both are pretty popular. A good deal of the Windows
C programmers learned C on Unix, with gcc. Many of these developed a
disdain for MS, but put up with Windows because it's what is available
at work, or it's what's cheapest to buy for their home, or they have
come to like or depend on various pieces of third-party Windows
software. Even those who don't actively hate MS but are old enough to
have cut their teeth on Unix/gcc may simply not realize MS offers
no-cost versions of their language products these days. (I was in
this group for quite some time. When I was a student, my options were
gcc at school or low-cost Borland products at home. MS products were
way out of my price range. I became conditioned to avoid MS not out
of hate but simply out of cost considerations.)

I do recognize that there's a whole generation of programmers now who
have no *nix experience, and are very familiar with Visual Studio
Express, and this group is likely to grow. Also, MSVC is the
canonical CPython compiler for the Windows builds. (Though Martin v.
Loewis has gone out of his way recently to remind people that they
will be sticking with 2008 for some time yet.)

So, obviously, it would be ideal if both MinGW and MSVC can be
supported. (And *ideally* ideal if the MinGW support doesn't require
MSYS.) Please do not take this to be me asking for things or being
demanding. My purpose is just to offer my observations and
experiences. I'm genuinely happy for whatever Windows support Shed
Skin can get, just for Shed Skin's sake. I do not have the time or
expertise (but mainly the expertise) to provide it myself. I was, and
still am for that matter, fully prepared to accept, without complaint,
a complete lack of Windows support if it comes to that.

John

Jérémie Roquet

unread,
Apr 22, 2010, 4:37:10 AM4/22/10
to shedskin...@googlegroups.com
2010/4/21 Mark Dufour <mark....@gmail.com>:
> thanks for the patch! I tested it, but I think I see the same problem: the
> python version adds two empty lines to the output (for the last with-test),
> while the shedskin version leaves them out:
> [...]

Could you please send me the generated cr.txt, lf.txt, crlf.txt and a
cpython.txt containing the exact output of the script when run with
cpython and a shedskin.txt containing the exact output of the script
when run with shedskin ? The cpython version is not expected to add
two consecutive empty lines at any point of the script...
Thanks in advance.

> btw, another small thing you might want to look into, is that in builtin.cpp
> there are some '-1' expressions, where we'd probably want to use
> std::string::npos (iirc). this leads to signed/unsigned warnings under the
> new windows/mingw package (without enabling all warnings). problem is, when
> I change them to use std::string::npos, some tests start failing, and I'm
> not sure how to fix this correctly.

I see, I'll look into this :-)

Best regards,

--
Jérémie

Oyster

unread,
May 1, 2010, 10:02:55 AM5/1/10
to shedskin-discuss

On 4月22日, 下午2时43分, John Yeung <gallium.arsen...@gmail.com> wrote:
> On Wed, Apr 21, 2010 at 11:46 PM, Andy Miller <nzmill...@gmail.com> wrote:
>
> Obviously the *most* typical Windows user, even one who uses Python,
> isn't going to have any C or C++ compiler at all.
if so, he does not need shedskin at all

for myself, I know only very little with C/C++, but I can use mingw to
build some extension if there is a Makefile for mingw

Oyster

unread,
May 1, 2010, 10:06:44 AM5/1/10
to shedskin-discuss
Thanks a lot from a windows users :)

On 4月18日, 下午9时58分, Andy <nzmill...@gmail.com> wrote:
> I've done some hacking over the weekend and have shedskin working
> 'well' under windows using MSVC C++ 2010 Express edition (almost
> certain that MSVC 2008 will NOT work due to template and STL support).

Reply all
Reply to author
Forward
0 new messages