extconf.rb spitting out SH Makefile on windows?

M

Mr Eiland

Title says it all, I'm running ruby extconf.rb in a visual studio 2008
pro command prompt and the resulting Makefile is not for nmake.

All I'm wanting is to create a single function that I can call into.
I'm not willing to use SWIG, I see no reason to add more moving parts
for something that should be so simple.

Ultimately, if I could find documentation on how to set up the visual
studio C++ environment so that I could compile the extension without the
use extconf.rb, I would prefer that, but using extconf.rb is the next
best thing.
 
E

Eric Christopherson

Title says it all, I'm running ruby extconf.rb in a visual studio 2008
pro command prompt and the resulting Makefile is not for nmake.

All I'm wanting is to create a single function that I can call into.
I'm not willing to use SWIG, I see no reason to add more moving parts
for something that should be so simple.

Ultimately, if I could find documentation on how to set up the visual
studio C++ environment so that I could compile the extension without the
use extconf.rb, I would prefer that, but using extconf.rb is the next
best thing.

Since no one's chimed in yet, I thought I would just post to say you'd
be best served installing the RubyInstaller DevKit, which includes gcc
and GNU make. Others might know of ways to integrate Ruby with VS, but
the conventional wisdom seems to be to use DevKit.
 
Q

Quintus

Am 04.03.2011 07:10, schrieb Eric Christopherson:
Since no one's chimed in yet, I thought I would just post to say you'd
be best served installing the RubyInstaller DevKit, which includes gcc
and GNU make. Others might know of ways to integrate Ruby with VS, but
the conventional wisdom seems to be to use DevKit.

I'd wonder if it was possible with Visual Studio anyway--All recent
Rubies have been compiled with gcc (via MinGW/MSYS), and incorporating a
C extension compiled with another compiler is not possible I think. The
OP would have to compile Ruby with MSVC as well.

Valete,
Marvin
 
L

Luis Lavena

Title says it all, I'm running ruby extconf.rb in a visual studio 2008
pro command prompt and the resulting Makefile is not for nmake.

What version of Ruby are you using?

From what you say sounds like you're using RubyInstaller, which used
GCC -- not Visual Studio -- as compiler toolchain.
All I'm wanting is to create a single function that I can call into.
I'm not willing to use SWIG, I see no reason to add more moving parts
for something that should be so simple.

SWIG has nothing to do with compiler toolchain.
Ultimately, if I could find documentation on how to set up the visual
studio C++ environment so that I could compile the extension without the
use extconf.rb, I would prefer that, but using extconf.rb is the next
best thing.

extconf.rb is required to create a Makefile compatible with Ruby.

Please read README.EXT that is bundled in Ruby source code.

Also, if you want Ruby with Visual Studio, build it or use the ones
from GarbageCollect page:

http://www.garbagecollect.jp/ruby/mswin32/en/

Be aware to instal the missing binaries:

http://www.garbagecollect.jp/ruby/mswin32/en/documents/install.html
 
M

Mr Eiland

I ended up moving everything to the mingw toolchain, but I don't
understand why you're saying it's impossible to use Visual Studio.

At the end of the day the windows loader is the one doing the work, not
the Ruby runtime. Unless you're telling me the dll being spit out by
mingw isn't really a DLL? I would need a more in depth explanation
before I would buy that Windows can load the appropriate DLL created by
mingw, but not the equivalent DLL created by Visual Studio. We're not
talking about compiler specific object files.
 
L

Luis Lavena

I ended up moving everything to the mingw toolchain, but I don't
understand why you're saying it's impossible to use Visual Studio.

I'm saying that using Ruby build with MinGW is not possible to compile
extensions with Visual Studio.
At the end of the day the windows loader is the one doing the work, not
the Ruby runtime.  Unless you're telling me the dll being spit out by
mingw isn't really a DLL?  I would need a more in depth explanation
before I would buy that Windows can load the appropriate DLL created by
mingw, but not the equivalent DLL created by Visual Studio.  We're not
talking about compiler specific object files.

Ruby 'mkmf' uses rbconfig, which is generated at compilation time of
Ruby, and is going to use the toolchain defined when Ruby was
compiled.

Beyond altering rbconfig to detect and use Visual Studio, is a matter
of CRT mismatch. This has been discussed in the list and the risk of
memory allocation and free of it due Ruby's GC design can generate
segfaults.

And if you need better answers, you can google for them, described the
problem above the big one, but there are others associated with
compilation flags.
 
M

Mr Eiland

It sounds to me like it has more to do with some of Ruby's design
decisions, which is something I can believe.

This will definitely be the last Ruby project I do on Windows, the Ruby
team needs to work on their windows support. I'm a unix weenie so it's
natural for me to use the mingw toolchain, but others aren't so
fortunate.

And no, I'm not willing to put forth my time to do it, I'll just be sure
and avoid Ruby on windows from now on. Ruby is a beautiful language,
but as far as I'm concerned it's hardly usable on Windows.
 
L

Luis Lavena

It sounds to me like it has more to do with some of Ruby's design
decisions, which is something I can believe.

Yes and no, see what I've mentioned about CRT mismatch.
This will definitely be the last Ruby project I do on Windows, the Ruby
team needs to work on their windows support.  I'm a unix weenie so it's
natural for me to use the mingw toolchain, but others aren't so
fortunate.

And so the same for Visual Studio.
And no, I'm not willing to put forth my time to do it, I'll just be sure
and avoid Ruby on windows from now on.  Ruby is a beautiful language,
but as far as I'm concerned it's hardly usable on Windows.

Then stop complaining. If you're not willing to contribute back, then
shut up.
 
M

Mr Eiland

Luis Lavena wrote in post #987573:
Yes and no, see what I've mentioned about CRT mismatch.

Which is a technical challenge, not an insurmountable hurdle. See
Python for an example.
And so the same for Visual Studio.

That's a shitty response. Windows dev's know Visual Studio,and the
context of this entire discussion has been about building extensions in
Windows, which makes your point facetious.
Then stop complaining. If you're not willing to contribute back, then
shut up.

Roger that, send me a private email and I'll let you know how much it's
going to cost you to influence my behavior.
 
P

Phillip Gawlowski

Roger that, send me a private email and I'll let you know how much it's
going to cost you to influence my behavior.

What Luis meant to say is that Ruby is open source. You are free to
spend the time and effort necessary to fix the flaws you see. This is
generally the case if a single person moves outside of a project's
supported build chain (which is MinGW on Windows), a choice was made
deliberately, since Visual Studio is only free in the Express
versions, and that cannot be bundled with a third party download, as
is necessary to support native extension building via RubyGems.

The problem you have is a result of your own choice of tools, and you
can't expect the community to fix problems that come about because of
your specific situation. Not unless you are willing to provide
incentives. Incentives that go beyond threatening to take your ball
and never, ever again use Ruby on Windows. As far as threats go, it's
a weak one, anyway: I'm not specially bothered that you won't use Ruby
on Windows any more, just because it doesn't work with a tool not
supported by the community.

--
Phillip Gawlowski

Though the folk I have met,
(Ah, how soon!) they forget
When I've moved on to some other place,
There may be one or two,
When I've played and passed through,
Who'll remember my song or my face.
 
M

Mr Eiland

Phillip Gawlowski wrote in post #987872:
What Luis meant to say is that Ruby is open source. You are free to
spend the time and effort necessary to fix the flaws you see. This is
generally the case if a single person moves outside of a project's
supported build chain (which is MinGW on Windows), a choice was made
deliberately, since Visual Studio is only free in the Express
versions, and that cannot be bundled with a third party download, as
is necessary to support native extension building via RubyGems.

mingw is not bundled with either Ruby Windows installer, or the rubygems
installer, making your entire point moot. In order to build an
extension
via ruby gems, you have to:

1. install ruby
2. install ruby gems
3. install rubydevkit
4. execute a .bat or .ps1 file every single time you want to install or
update a gem that needs to be compiled.

So don't give me this bullshit about MS not allowing you to bundle 3rd
party software, they don't bundle gcc either.

The problem you have is a result of your own choice of tools, and you
can't expect the community to fix problems that come about because of
your specific situation. Not unless you are willing to provide
incentives. Incentives that go beyond threatening to take your ball
and never, ever again use Ruby on Windows. As far as threats go, it's
a weak one, anyway: I'm not specially bothered that you won't use Ruby
on Windows any more, just because it doesn't work with a tool not
supported by the community.

That's a nice story, but you missed the part where I explicitly stated I
had moved everything over to the mingw toolchain, so take your strawmen
and stuff them up your ass.

I'm not using Ruby for Windows development because the windows support
is shit, not because I have to use the mingw toolchain. That's a
symptom, the problem is the community thinking it's acceptable to
pretend Windows is Unix. Jackasses and fools on both sides of that
argument. The jackasses who want to stay in unix land, even in windows,
and the fools who think windows is the only OS out there.

Next time, just acknowledge the point and state no one has any interest
in fixing it instead of launching into the OSS party line. You'll stop
misinterpreting statements of intention as threats.
 
L

Luis Lavena

mingw is not bundled with either Ruby Windows installer, or the rubygems
installer, making your entire point moot.  In order to build an
extension
via ruby gems, you have to:

1. install ruby
2. install ruby gems
3. install rubydevkit
4. execute a .bat or .ps1 file every single time you want to install or
update a gem that needs to be compiled.

Incorrect.

1) Install RubyInstaller. RubyGems is already bundled
2) Download DevKit. Installation is a manual process for now,
contributions are welcome to provide a real installer.
3) There is no step 3, gem install will trigger the gem compilation
without polluting your normal environment.
So don't give me this bullshit about MS not allowing you to bundle 3rd
party software, they don't bundle gcc either.

Dude, chill out, MS EULA forbids you to bundle Visual Studio with any
tool.

This is an old discussion about MinGW versus Visual Studio for Ruby
for Windows, and you've reach to it 3 years later.

You can still use Visual Studio, sure, go ahead download Ruby source
code, download all dependencies and compile them with Visual Studio,
and rejoice, be happy, go whatever you want.

But don't come here to criticizes the work done by others.

Windows != Visual Studio. There are other languages beyond C or C# and
of course there are other compilers beyond Microsoft one.

So all the offensive stuff you wrote is completely pointless.
Next time, just acknowledge the point and state no one has any interest
in fixing it instead of launching into the OSS party line.  You'll stop
misinterpreting statements of intention as threats.

Fixing? what is broken? Please, show it to me, wait, even better, send
some patches, I will happily apply to Ruby itself.
 
J

Jon Forums

mingw is not bundled with either Ruby Windows installer, or the rubygems
installer, making your entire point moot. In order to build an
extension via ruby gems, you have to:

1. install ruby
2. install ruby gems
3. install rubydevkit
4. execute a .bat or .ps1 file every single time you want to install or
update a gem that needs to be compiled.

Get your facts right.

1. Installing Ruby via RubyInstaller also includes RubyGems.
2. Once you've installed the RubyInstaller DevKit you don't need to run
a .bat or .ps1 file every time you need to install or update a native
gem. The DevKit install process integrates with RubyGems by providing a
pre-install hook that temporarily brings the DevKit onto your PATH. If
you _want_ to use the DevKit toolchain outside of Ruby/RubyGems, by all
means have at it by taking advantage of the .bat/.ps1 helper scripts.

Glad you asked; there are very good reasons why the DevKit is not, nor
should be, bundled by default with the RubyInstaller. However, if you do
need/want the toolchain, it's fairly straightforward to integrate it.

I'm not using Ruby for Windows development because the windows support
is shit, not because I have to use the mingw toolchain.

Your call, but what specifically do you mean by "...the windows support
is shit...?"

Seems from your posts you aren't happy with issues regarding your choice
of toolchains, and you're not interested in understanding the tradeoffs
made and solutions implemented for MRI Ruby on Windows users.

And you've obviously not browsed through the Ruby/RubyGems source trees
to see how much Windows specific support has been built in, or the
number of bugfixes, etc.

Any you likely haven't investigated JRuby's Windows support.

Nor have you done the most basic of searches to discover the work by
many of the gem authors (Nokogiri, FFI, Thin, EventMachine, hitimes,
sqlite3, gherkin, do_sqlite3, ...) to support Windows users.

That's OK, but stop chumming this list.

That's a symptom, the problem is the community thinking it's
acceptable to pretend Windows is Unix.

What the hell are you talking about? Who in the "community" is
pretending Windows is Unix?

Next time, just acknowledge the point and state no one has any interest
in fixing it instead of launching into the OSS party line.

Once again, what specifically is "..fixing it..." from your point of
view?

Jon
 
P

Phillip Gawlowski

mingw is not bundled with either Ruby Windows installer, or the rubygems
installer, making your entire point moot. =A0In order to build an
extension
via ruby gems, you have to:

1. install ruby
2. install ruby gems
3. install rubydevkit
4. execute a .bat or .ps1 file every single time you want to install or
update a gem that needs to be compiled.

Step for is unnecessary, as has been pointed out.
So don't give me this bullshit about MS not allowing you to bundle 3rd
party software, they don't bundle gcc either.

The Ruby DevKit *does* bundle a full compiler toolchain, not just what
is required to have a compiler run (which would be ruby.h, and that'd
be it, more or less).

The DevKit download, however, cannot, and will not, include Visual C++
Express, since the EULA forbids redistribution by third parties. So,
it's not possible to bundle VS in any way, shape, or form, and thus
the community need for an NMake-compatible toolchain is pretty low.

That's a nice story, but you missed the part where I explicitly stated I
had moved everything over to the mingw toolchain, so take your strawmen
and stuff them up your ass.

I refer you to your OP, where you said that Ruby's extconf.rb doesn't
generate NMake compatible makefiles.

Everything else follows from that. Including your outbursts.
I'm not using Ruby for Windows development because the windows support
is shit, not because I have to use the mingw toolchain. =A0That's a
symptom, the problem is the community thinking it's acceptable to
pretend Windows is Unix. =A0Jackasses and fools on both sides of that
argument. =A0The jackasses who want to stay in unix land, even in windows= ,
and the fools who think windows is the only OS out there.

That's why the RubyInstaller project created the DevKit. It makes, for
all intents and purposes, the MinGW tools invisible to the average
Ruby user. They are only noticeable if a gem blows up during
compilation. Which is neither Ruby's nor Windows' fault, nor the fault
of MinGW.

BTW: Notice that Ruby mingw32 lacks any dependency on cygwin.dll. It's
a fully native Ruby binary, not using any sort of compatibility layers
to map *NIX system calls to Windows equivalents.
Next time, just acknowledge the point and state no one has any interest
in fixing it instead of launching into the OSS party line.

Fact is: OSS is volunteer driven.
Fact is: OSS is "scratch your own itch" driven.
Fact is: *you* want NMake compatible makefiles.
Fact is: Unless you either do the patch yourself, or offer incentives
for somebody else to provide a patch, you don't get to complain. You
can file a bug, but otherwise you are left to the community's fancy
and what is seen as a necessary feature (First: Get better Windows
compatibility with Ruby).
=A0You'll stop misinterpreting statements of intention as threats.

Your tone is leaving very little to the imagination.

That you aren't offering anything beyond "Ruby on Windows is shit",
and "OSs party line", among curses, pushes you towards the "troll" end
of things.

Offer some suggestions or, better yet, patches to fix what you
perceive as issues, and we are talking. But unless your bug reports
become more verbose than "Windows support by Ruby is shit", nothing
will be done.

--=20
Phillip Gawlowski

Though the folk I have met,
(Ah, how soon!) they forget
When I've moved on to some other place,
There may be one or two,
When I've played and passed through,
Who'll remember my song or my face.
 
M

Mr Eiland

Phillip, until you stop with the strawmen, I'm going to ignore you from
here on out. I know it makes you feel productive, but it's useless.
You're explicitly ignoring later posts to attack what you perceive to be
a slight on your romanticized view of OSS.
Incorrect.

1) Install RubyInstaller. RubyGems is already bundled
2) Download DevKit. Installation is a manual process for now,
contributions are welcome to provide a real installer.
3) There is no step 3, gem install will trigger the gem compilation
without polluting your normal environment.

Fair enough, I installed in that order based upon everything I had read,
but I don't claim to be an authority. In particular the documentation
for DevKit left me with the impression that the environment had to be
loaded up every time you wanted to use it. It's been long enough that I
can't remember exactly where it was documented, so lets just call it a
misunderstanding on my part.

Dude, chill out, MS EULA forbids you to bundle Visual Studio with any
tool.


We're talking past each other.

What I meant is the Ruby installer doesn't bundle gcc or Visual Studio,
so the idea that it's built with gcc because Visual Studio doesn't allow
3rd party distribution is fallacious.

I could *maybe* see your point if the installer came bundled with
everything needed to build the gems that are compiled, but it doesn't.
Once you've installed Ruby on Windows, you now get to go download
another piece of software, so it becomes a matter of choice. DevKit, or
*Microsoft* Visual C++ Express.

Windows != Visual Studio. There are other languages beyond C or C# and
of course there are other compilers beyond Microsoft one.

Which is another shitty response. We're talking about C development on
a Windows OS. The de facto standard is Visual Studio, in much the same
way that there are other compilers on Unix, but the de facto standard is
gcc, and you can't really talk yourself past that point.

This is an old discussion about MinGW versus Visual Studio for Ruby
for Windows, and you've reach to it 3 years later.

Exactly, which is why I won't bother with Ruby on Windows anymore. It's
unnecessarily difficult, and the community has already decided it isn't
important enough to change that. That isn't a threat, it's me removing
obstacles between my clients and their solutions.

But don't come here to criticizes the work done by others.

I can, and I will. If that hurts your feelings in any way, grow a
thicker skin.


Nor have you done the most basic of searches to discover the work by
many of the gem authors (Nokogiri, FFI, Thin, EventMachine, hitimes,
sqlite3, gherkin, do_sqlite3, ...) to support Windows users.

That's because I'm not talking about gems, I'm talking about Ruby.
That's OK, but stop chumming this list.

No, and you guys seriously need to stop that. If my style offends you,
blacklist me and be done with it.


As for fixing it, fixing it means not asking someone on windows to learn
the gcc toolchain just to build an extension. The express versions are
free to download, asking the user to download an MS compiler or the gcc
compiler should be a no brainer.

Standardize on a version of the Express compilers and work with the
rubygem maintainers to let gem authors offer different installs for
Windows. Suddenly, the user doesn't have to install a dev environment
just to install a ruby gem, the complexity is pushed off to the gem
maintainers, where it should be. Windows is not Unix, stop pretending
it is, and do things the Windows way when in Windows.

update mkfm to SUPPORT WINDOWS, and that means more than simply using
dlltools and then spitting out a Makefile. Piggyback off of CMake if
you have to.

http://www.rubyinside.com/is-windows-a-first-class-platform-for-ruby-823.html
"Ruby development on Windows? The environment sucks (the language rocks
though). Is that Windows fault? No. The community overall is very *nix
based, so development focuses on that environment, and very anti-Windows
as is."

This man nailed the problem, and the problem hasn't changed. The
Windows support may have improved, but not enough to compete with other
languages on Windows. And "use JRuby" is not a valid alternative.
 
L

Luis Lavena

Fair enough, I installed in that order based upon everything I had read,
but I don't claim to be an authority.  In particular the documentation
for DevKit left me with the impression that the environment had to be
loaded up every time you wanted to use it.  It's been long enough that I
can't remember exactly where it was documented, so lets just call it a
misunderstanding on my part.

Please, wiki is open, introduce your edits to make it more easy to
understand for newcomers. Let us know on our mailing list once you do
it:

http://groups.google.com/group/rubyinstaller

The page is here:

https://github.com/oneclick/rubyinstaller/wiki/Development-Kit

If something is not clear, we would like to correct it.
We're talking past each other.

What I meant is the Ruby installer doesn't bundle gcc or Visual Studio,
so the idea that it's built with gcc because Visual Studio doesn't allow
3rd party distribution is fallacious.

I could *maybe* see your point if the installer came bundled with
everything needed to build the gems that are compiled, but it doesn't.

Not everybody needs a compiler bundled. Not every needs to install
gems or compile extensions to use Ruby, why increase installers 20MB
(GCC) or 2GB (Visual Studio + PSDK) for something you don't want?

I believe I have stated this in the wiki in relation to this and other
tools:

https://github.com/oneclick/rubyinstaller/wiki/FAQ#bundled_short
http://groups.google.com/group/rubyinstaller/msg/34dd829ebe0e4567
Once you've installed Ruby on Windows, you now get to go download
another piece of software, so it becomes a matter of choice.  DevKit, or
*Microsoft* Visual C++ Express.

Over the past 5 years I've been actively working with Ruby for Windows
and 4 years since I took over One-Click Installer, *many* have
proposed make Ruby agnostic of the toolchain used to build extensions.

But the truth is that nobody decided to do it, everybody complains and
is a hard task.

We are talking about changing an utterly non-OOP unstructured and
commentless tool named 'mkmf'.

Is a daunting task, specially since pretty much all extensions base on
that tool to compile and introduce something like Pythons' distutil
could introduce many many regressions.

Instead of that, I decided to focus my efforts in reduce the
complications while sticking to only one toolchain.

Luckily for us Ruby is written in C, but if any of the parts of the
language was written in C++ and exported, that will be impossible to
deal with due GCC vs Visual Studio C++ different ABIs

Again, the free/xfree malloc and mismatch of CRT you already know.

Having all those drawbacks I think you will value what you can get now
compared to previous Ruby for Windows situation which was the strict
dependency on Visual C 6.0

You can search the web or read my my blog about that: http://blog.mmediasys..com/
Which is another shitty response.  We're talking about C development on
a Windows OS.  The de facto standard is Visual Studio, in much the same
way that there are other compilers on Unix, but the de facto standard is
gcc, and you can't really talk yourself past that point.

Again man, I believe I treated you with respect all this time. Calling
shitty to responses you don't like is plainly aggressive and idiotic.

Before Windows NT codebase there were different compilers for Windows,
not just Visual Studio, seems you can't see beyond that point either.

What you call "defacto" standard is what you want be your standard.
Visual Studio Express not only complicates the compilation nature of
Ruby and it's dependencies but also can't be automated to the point we
use to deliver RubyInstaller.

Please, I encourage you do the same thing we do with RubyInstaller
project and how we build all the dependencies with Visual Studio and
let me know.

Excluding for a second that Visual Studio express lacks advance
optimizations like PGO only found in paid version, which is a
restrictive point for someone that works with Open Source Software
want to contribute back but is not paid by it.

Heck, over the past years there has been ZERO emails from Microsoft
encouraging me use their tools, and pretty much everybody knows my
email address.
Exactly, which is why I won't bother with Ruby on Windows anymore.  It's
unnecessarily difficult, and the community has already decided it isn't
important enough to change that.  That isn't a threat, it's me removing
obstacles between my clients and their solutions.

To change that, work is required, clearly you're not interested in
work on it. How you can get engagement from others if not even you are
committed to it?

If you want dinner served, go to a restaurant. I believe you're coming
from a different mindset in relation to development, but Open Source
works by actions, not by words. You can write down the most beautiful
haiku but is pointless if you can provide a code that solves the
problem.
I can, and I will.  If that hurts your feelings in any way, grow a
thicker skin.

You're not providing constructive criticism. Your words can't be
backed by your work, and that is how things roll in Open Source.

But again, this whole discussion is moot since clearly you've
repeatedly said don't care about Ruby for Windows, so why bother
answering?

Seems you're the one can't move forward.
As for fixing it, fixing it means not asking someone on windows to learn
the gcc toolchain just to build an extension.  The express versions are
free to download, asking the user to download an MS compiler or the gcc
compiler should be a no brainer.

If you consider it needs to be fixed, please, fix it and showed to us.
Standardize on a version of the Express compilers and work with the
rubygem maintainers to let gem authors offer different installs for
Windows.  Suddenly, the user doesn't have to install a dev environment
just to install a ruby gem, the complexity is pushed off to the gem
maintainers, where it should be.  Windows is not Unix, stop pretending
it is, and do things the Windows way when in Windows.

MinGW is not Unix, is Minimalistic GNU for Windows and is GNU tools
for Windows, is not emulating anything.

It uses same libraries and similar headers than Visual Studio (ported
to GCC syntax)
update mkfm to SUPPORT WINDOWS, and that means more than simply using
dlltools and then spitting out a Makefile.  Piggyback off of CMake if
you have to.

If by Windows you mean Visual Studio, see the mswin32 builds from
garbagecollect:
http://www.garbagecollect.jp/ruby/mswin32/

Use that, use Visual Studio with it and be happy.

As for Ruby and RubyInstaller: Patches are welcome.

Everybody is a critic but no one is willing to put code where their
mouth is.

Show us -- with real code -- that all that we managed to achieve today
is possible with Visual Studio Express, that it can be automated and
that usage can be simplified.

And I'm willing to take the chance of doing a RubyInstaller based on
Visual Studio.

And no, is not up to me to do it, you seem very interested in state
that I'm wrong and you're right, so is up to you.

As for the RubyInside article, ego on a side, see who is being quoted
in it.
 
J

Jon Forums

Nor have you done the most basic of searches to discover the work by
That's because I'm not talking about gems, I'm talking about Ruby.

When you talk about Ruby, practically speaking, you implicitly talk
about gems/extensions for desired/required functionality outside of
what's in core+stdlib. That said, I don't know your use cases.

No, and you guys seriously need to stop that. If my style offends you,
blacklist me and be done with it.

You missed my point. I could care less about your style. Say what you
want to say, how you want to say it, as most of the "cost" is yours.

But I do care when people state things as fact that just aren't so. For
a list like this I think it's crucial to be clear about facts vs.
speculation. It's OK to say things like "IIRC" or "(??)" don't you
agree?

As for fixing it, fixing it means not asking someone on windows to learn
the gcc toolchain just to build an extension. The express versions are
free to download, asking the user to download an MS compiler or the gcc
compiler should be a no brainer.

Now we're getting somewhere...specifics! And I agree.

However, wrt the rubyinstaller installer .exe and .7z archives and build
recipes I'm biased that this already is the case as we've abstracted
much of the build toolchain.

For example, to build mri ruby from source you typically:

* clone the rubyinstaller github repo
* rake ruby19

and if you wan't to go a bit deeper (say build from your local ruby
source with mods for vtune profiling and build with mingw 4.5.2) you may
do something like:

* rake ruby19 local=c:\users\me\ruby-trunk vtune_build=1
dkver=mingw-32-4.52

And as a user who want's to build a native gem, say rdiscount, you
typically do something like:

* download the devkit
* install it via the dk.rb ruby script (can be tweaky, but that's
another topic)
* gem install rdiscount [--platform=ruby]

Or if you want to build a gem that has external dependencies:

* gem install curb --platform=ruby -- --with-curl-lib="c:\curl\bin"
--with-curl-include="c:\curl\include"

So many of the typical use cases are abstracted away already. That said,
(a) there's areas that are a bit rough and could use some updates, and
(b) the current recipes don't address your desire for the msft express
toolchains or for the msft windows sdk.

As I said, I'm biased since I've been using and refining rubyinstaller
for awhile. That said, we can always benefit from fresh eyes, fresh
determination, and fresh horsepower.

Swing by our google group and weigh in with specifics. Hopefully you can
tell by now that Luis (project lead) is always open for discussing how
to make the rubyinstaller more useful for ruby windows users.

Standardize on a version of the Express compilers and work with the
rubygem maintainers to let gem authors offer different installs for
Windows. Suddenly, the user doesn't have to install a dev environment
just to install a ruby gem, the complexity is pushed off to the gem
maintainers, where it should be.

I agree with you if we were in a perfect world and had unlimited time
and $$. Wouldn't it be great if a Window's user could type "gem install
farfalle" and it automagically downloads a binary gem (1.8 or 1.9
dependent upon what their using) with the correct runtime linkage for
their system? And a *nix user downloads the gem and it builds for their
system? Same for OSX, same for ___.

There are a lot of dependency details with this one.

Even though most people contribute in their spare time, without pay to
fund their habits, a usable-but-not-perfect situation has evolved:

* gem authors on *nix can support their windows users via
cross-compiling with mingw, taking advantage of Luis' rake-compiler
helper.
* gem authors can build "fat" binary gems so a user doesn't have a
1.8-only gem explode when they install it on their 1.9 system.
* native interop capabilities such as DL and FFI often help one
integrate with windows libs.
* mri ruby and rubygems do windows-specific things to help
* rubyinstaller users can use the devkit to build and install native
gems from source if there's not a correct binary gem available.

Perfect? No. Usable? Yes. Could it be improved? Of course.

Windows is not Unix, stop pretending it is, and do things the Windows way when
in Windows.

Another pov is that we've just found creative ways and creative tools to
allow us to use MRI Ruby on Windows. It's a matter of pragmatism rather
than pretending.

This is already well past tl;dr and likely best continued on the
rubyinstaller group if you have specifics to contribute, but...

* as a Windows developer, what specifically are your "must haves" and
"like to haves" for mri ruby on windows?
* what specifically keeps you from running mri ruby in a "production"
environment?
* would your issues be the same from your user's (non-developers) pov?

Jon
 
M

Mr Eiland

Over the past 5 years I've been actively working with Ruby for Windows
and 4 years since I took over One-Click Installer, *many* have
proposed make Ruby agnostic of the toolchain used to build extensions.

But the truth is that nobody decided to do it, everybody complains and
is a hard task.

We are talking about changing an utterly non-OOP unstructured and
commentless tool named 'mkmf'.

Is a daunting task, specially since pretty much all extensions base on
that tool to compile and introduce something like Pythons' distutil
could introduce many many regressions.

Instead of that, I decided to focus my efforts in reduce the
complications while sticking to only one toolchain.

Which is fine, but I was asked what I thought should be changed. It
isn't easy, and I'm not willing to put forth the effort, not when I can
just use other tools. I love Ruby as a language, but when it starts
getting between me and the solution, I stop using it.

What you call "defacto" standard is what you want be your standard.
Visual Studio Express not only complicates the compilation nature of
Ruby and it's dependencies but also can't be automated to the point we
use to deliver RubyInstaller.

I call it de facto because most shops that develop for Windows are going
to be using MS tools to do so. I called your response shitty because you
act as if it's just as likely that a windows developer using C will be
familiar with the gcc toolchain as they will with the MS toolchain,
which is absolutely not true. Yes, there are other languages out there,
but they are not under discussion, C is. That's also why I made my
comment about jackasses and fools. When in Rome, do as the Romans do.
When in Windows, do as the Windows devs do.

You obviously have your reasons for doing things the way you have, but
don't act as if it's reasonable to expect a Windows developer to be
familiar with Unix tools. That's shitty. If you want to make the
argument that they can learn, then do so, but that isn't the argument
you made.


For the rest of your response, please read back over what I said. I
*specifically* stated that I would not be using Ruby on Windows for
another project, and I did *not* give a call to action for anything.
Phillip Gawlowski pigeon-holed my statements into the classic OSS
strawman, and you're continuing in that vein. You specifically asked me
what I thought was wrong, and I told you. It is unfair for you to treat
that as a call to action when you *asked* for the input. I've already
stated multiple times that my solution is to simply use other tools, I
don't know how much more clear I can be. I really like Ruby, but that
fondness no longer includes Windows.



Jon, The great thing about the criticisms being mine is that I get to
choose who is included in said criticism and I can tell you
unequivocally that I included no gem authors. If you don't stop, I'm
going to start thinking you dislike the gem authors.

* as a Windows developer, what specifically are your "must haves" and
"like to haves" for mri ruby on windows?
* what specifically keeps you from running mri ruby in a "production"
environment?
* would your issues be the same from your user's (non-developers) pov?

I am not a Windows developer, I have more experience with Unix, and I
prefer Unix. I'm simply a developer who is flexible enough to
comfortably develop in both environments.

There are no must haves and nothing is preventing me from running it in
a production environment. I haven't claimed either, so I'm confused as
to why you would assume that was the case. I stated I wouldn't use Ruby
on Windows in the future, I never stated it was unusable for me now.
Not all of that is directly related to this particular discussion, but
Ruby in Windows is subpar. This is not an isolated opinion.

As for my users, I solved most of the issues by using an MSI installer.
It feels like Windows to them, they're happy, and I get paid.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

Forum statistics

Threads
473,744
Messages
2,569,484
Members
44,906
Latest member
SkinfixSkintag

Latest Threads

Top