This is the mail archive of the
mailing list for the Java project.
Re: Experimental UNICODE-only MinGW Build
- From: Mohan Embar <gnustuff at thisiscool dot com>
- To: João Garcia <jgarcia at uk2 dot net>
- Cc: java at gcc dot gnu dot org, Bryce McKinlay <bryce at mckinlay dot net dot nz>, tromey at redhat dot com, João Garcia <jgarcia at uk2 dot net>, Ranjit Mathew <rmathew at hotmail dot com>, luciano at virgilio dot it
- Date: Thu, 20 Nov 2003 09:26:32 -0600
- Subject: Re: Experimental UNICODE-only MinGW Build
- Reply-to: gnustuff at thisiscool dot com
>Thank you! Nice work. This can be useful for many people already, and it
>is a step forward.
Like I said, I felt bad about doing nothing about this for so long.
>Could you post the relevant code (using the mailing list or on your web
>site)? Thank you. :-)
...I'm still testing this, so it's not final yet. I suspect it won't
change much, though. I will submit an official patch today or tomorrow.
If you'll notice, I've used the T macros so that someone wanting to
do ANSI build could simply remove the [_]UNICODE defines. I haven't
gotten around to testing this yet, though.
>But AFAIK this is only the static link portion for the dll (and it is
>not small). Am i right?
The static, open-source libunicows.a is 399K, the non-free MS unicows.dll
>You still have to have the MS dll for this to work on Win9X.
>And you cannot allow re-distribution of the DLL and can only distribute
>it bundled with your code that should be exclusively for Windows platform...
>In other words: MS forces you to limit your own distribution license and
>bundled code (see Open Office project for details)...
>And it is not easy to set a link for the UNICOWS distribution in the MS
>web site (it changes all the time)...
I agree with you on all counts. I was able to find it easily enough with
a Google search, but like you said, this could change tomorrow. Win9X users
would be at MS' mercy.
During this experiment, I changed my mind about UNICOWS. My stance is that
making libgcj UNICODE-only and supporting WinNT / Win2K / WinXP elegantly
and out-of-the-box is more important than inconveniencing Win9X users. I am
also capitalizing on Bryce's post:
...which would appear to condone this, even though I'm not necessarily
sure that there aren't a lot of Win9X installations out there.
When I submit my patch, I'm going to disqualify myself from reviewing
it and let someone like Bryce, Tom or Andrew make the final decision,
with the necessary lobbying from you, me and others who might have an
interest in this. Hopefully, Ranjit will chime in too.
For the record: are you very much against a UNICOWS solution or would
you tolerate it?
>Besides that, you do not need the major part of the UNICOWS bundle just
>to make character conversions. It would be better to make our own
>"mini-unicows" for this purpose, based on
>WideCharToMultiByte()/MultiByteToWideChar() and A-function calls!
>Anyway... this would make things for Win32 to far apart from POSIX in
>the IO library... I would not recommend this way at all...
>But you could do this just for sport! :-)
This would be interesting, but the time I allotted myself for this
is already depleted and in the negative zone. :) Besides, I'd rather
leverage MS' work, even though it's admittedly overkill for our purposes.