[MLton] x86_64 port status + powerpc

Wesley W. Terpstra terpstra@gkec.informatik.tu-darmstadt.de
Wed, 10 May 2006 16:32:40 +0200


On May 10, 2006, at 3:36 AM, Matthew Fluet wrote:
>>> So far, I've needed the attached patch for the runtime.
>>
>> Wesley, you have svn commit access, and we're perfectly fine with you
>> making commits.  So, feel free to commit this patch and to  
>> continue to
>> commit without explicit approval.
>
> I've committed the patch, but feel free to commit in the future.

Thanks.

I just feel a bit uncomfortable about committing to your branch as I  
might break it! Also, I thought you guys had some sort of changelog  
history / documentation scheme I don't know about..?

To test your resolve, I've just committed a patch (to the x86_64  
branch) which fixes filename handling in the bin/* files and  
Makefiles. In general, MLton shell scripts (independent or in  
Makefiles) have been pretty sloppy about quoting strings or  
preserving parameters. For instance, if you tried to run:
	mlton @MLton "foo bar" -- ...
the 'foo bar' gets turned into '"foo' 'bar"' by the script. I fixed  
this for the mlton-script and the regression command.

The main reason I went through and systematically escaped things was  
so that building in a path with a ' ' in it works. This was really  
bothering me under MinGW (more on the mingw port in another email) as  
a windows home directories include usernames with spaces. I've tested  
that a build works for me, but as there are so many different script  
cases, I have probably broken one or two.

As a side-note, most of this wouldn't have been necessary if MLton  
didn't use absolute paths everywhere. Also, since MLton makefiles  
allow the user to override MLTON=..., those needed to be escaped too  
(in case I give a path with specials in it). More information is  
included in the commit message.

I'll send another email about powerpc once I've tracked down a couple  
more problems in the build.