Reference implementation currently too hard to install in Windows
Thanks for the feedback; I agree the Windows build is too hard to install. Last I checked it wasn't possible to build Windows binaries with SML/NJ. I will talk to my contacts who work on SML/NJ and see if there's something we can do to make this more seamless.
Thanks again for the detailed feedback!
,
Dave Herman wrote:
Thanks for the feedback; I agree the Windows build is too hard to install. Last I checked it wasn't possible to build Windows binaries with SML/NJ. I will talk to my contacts who work on SML/NJ and see if there's something we can do to make this more seamless.
Thanks Dave, that would be great. Could you let us know how things stand once you know more?
Thanks again, Jonathan
Thanks Dave, that would be great. Could you let us know how things stand once you know more?
It's looking good. Despite the fact that SML/NJ doesn't appear to have upcoming plans to support Windows binaries, MLton (mlton.org) does and has for a long time. We have written pretty exclusively to the standard of SML so porting to MLton has remained pretty easy. Last night Graydon and I successfully built Windows binaries with MLton. We should be pushing out another release ASAP that will include a Windows (Cygwin) binary.
I have to look into licensing, but I believe if we can include the Cygwin DLL with the Windows binary, there should be no dependencies whatsoever and you can just download and run.
For the future, however, I do want to improve the build process to make it easier for the uninitiated. SML/NJ and MLton both tend to be a little tricky to install, though, so that goal may be trickier.
Dave Herman wrote:
Thanks Dave, that would be great. Could you let us know how things stand once you know more?
It's looking good. Despite the fact that SML/NJ doesn't appear to have upcoming plans to support Windows binaries, MLton (mlton.org) does and has for a long time. We have written pretty exclusively to the standard of SML so porting to MLton has remained pretty easy. Last night Graydon and I successfully built Windows binaries with MLton. We should be pushing out another release ASAP that will include a Windows (Cygwin) binary.
I have to look into licensing, but I believe if we can include the Cygwin DLL with the Windows binary, there should be no dependencies whatsoever and you can just download and run.
Fantastic! Hopefully Windows users will soon be able to provide useful feedback based on first hand experience. :-)
For the future, however, I do want to improve the build process to make it easier for the uninitiated. SML/NJ and MLton both tend to be a little tricky to install, though, so that goal may be trickier.
Good luck with that, and thanks for your efforts guys!
Jonathan
Graydon Hoare wrote:
Thanks for the patience, and continuing to push. The RI continues to lag the spec in a number of places -- there is a bug tracker to follow these -- but it's always great to have more hands playing with it.
Of course. I was very happy with the quick response to my feedback and the extra information on where things stand. Given the extremely short time in which it's been developed, if the reference implementation wasn't buggy, I'd think I'd somehow jumped to some sort of parallel, Utopian universe. :-)
Thanks again, Jonathan
Jonathan Watt wrote:
I have to look into licensing, but I believe if we can include the Cygwin DLL with the Windows binary, there should be no dependencies whatsoever and you can just download and run.
Fantastic! Hopefully Windows users will soon be able to provide useful feedback based on first hand experience. :-)
For the future, however, I do want to improve the build process to make it easier for the uninitiated. SML/NJ and MLton both tend to be a little tricky to install, though, so that goal may be trickier.
Good luck with that, and thanks for your efforts guys!
Thanks for the patience, and continuing to push. The RI continues to lag the spec in a number of places -- there is a bug tracker to follow these -- but it's always great to have more hands playing with it.
First of all, I think making the reference implementation available to everyone is a great idea. Thanks for that. Unfortunately, at this early stage, I think the barriers too getting it running are too high for your average ECMAScripter with only Windows experience. If one of the aims is to collect feedback on, and encourage experience with, ES4, then these really need to be lowered. Some of these barriers include:
No link to cygwin on the download page.
No mention of which options to choose when installing cygwin (just say "accept the defaults"? Or not? (I already had it installed, so I'm not sure if the defaults are enough)).
The Microsoft Windows section on the SML/NJ page links to a zip and WININSTALL file but, after messing around with that, I figure out it's not the one to use with cygwin.
I notice there's also a CYGWININSTALL file linked to on the SML/NJ page (and it could be clearer about the fact I should use it and then the Unix download and INSTALL instructions)
The instructions in CYGWININSTALL assume too much from the perspective of your average Windows ECMAScripter (what is "binmode", "manpage" and where and how do you set up a cygwin environment to automatically "export the environment variable").
The contents of the INSTALL file are themselves a barrier to your average Windows user.
On following the CYGWININSTALL and then Unix INSTALL instructions I get the error:
/usr/share/smlnj/bin/.link-sml: line 45: /home/jwatt/smlnj/bin/.arch-n-opsys: No such file or directory : unable to determine architecture/operating system config/install.sh !!! Boot code failed, no heap image (sml.x86-cygwin).
And yes, I had SMLNJ_CYGWIN_RUNTIME exported.
Since I personally have a little cygwin/unix shell experience I could continue on and try to figure out what I did wrong, but that's besides the point I think. Making all these problems go away by providing a native Windows installer would be ideal, but until that happens, could someone add step-by-step instructions for Windows dummies who don't care to experience the pain required to figure out how to get things working in a Unix environment? I assuming most ECMAScripters on Windows fall into this camp.
Again, thanks for your efforts to make the reference implementation available to all, and I hope the above is helpful in that regard.
, Jonathan