[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Question: Windows Build - feature branch
From: |
Conrad T. Pino |
Subject: |
RE: Question: Windows Build - feature branch |
Date: |
Wed, 7 Apr 2004 13:16:04 -0700 |
Hi Derek,
> From: Derek Robert Price [mailto:derek@ximbiot.com]
>
> >>>Is adding a complete set distinctly named VC 6.0 build files and
> >>>keeping VC 5.0 files in a semi-maintained state an option?
>
> I suppose I wouldn't really have any objection to this if it appears
> most convenient to the folks who have to use the files.
I asked you first since you're a gatekeeper. I plan to email Dennis
Jones to develop a consensus for consideration.
Double checking, who else beyond the lists should be solicited?
> >Do you have any thoughts about MinGW http://www.mingw.org/?
>
> Hadn't even heard of it before. It looks like it is trying to do what
> we want. Have you tried to build CVS with it?
Not yet (I'm avoiding lost time on unacceptable options) but that's why
I asked. My preliminary reading looks promising so far.
> >>I haven't tried the DJGPP version of GCC yet. It is possible that it
> >>doesn't suffer from the same limitations as Cygwin. If you have time to
> >>research that and would like to report back, I would like to hear.
> >
> >DJGPP http://www.delorie.com/djgpp/ looks problematic since it's target OS
> >is DOS, MS-DOS and compatibles. The current Windows build targets the Win32
> >API and DJGPP targets the DOS API. I'm concerned we lose too munch function
> >with the DOS API.
>
> Windows still has support for DOS, and I'm fairly certain that the
> command line version of CVS won't be making much use of Windows-specific
> calls. Some of the networking stuff might be tricky, but I'll be
> suprised if that won't build just from what I heard when I used to lurk
> on the Autoconf lists.
Yes, Windows NT supports the DOS API but it still imposes constraints.
The 8.3 size limit and upper case only for file names being the worst.
I assume the directory name "windows-NT" implies a minimum API level
which has to be the Win32 API. The code in "windows-NT" seems to rely
heavily on the NT POSIX layer but it's not fully compliant and I expect
Microsoft will deprecate or drop it at some point.
Right now I'd lean towards using the Win32 API directly and avoiding
the NT POSIX layer but that is NOT a proposal for the CVS project.
I hope MinGW lives up to it's initial promise.
> Derek
Conrad