[PATCH 00/39] Windows port, base and archival/

[Denys Vlasenko]

On Saturday 17 April 2010 13:19, Nguyen Thai Ngoc Duy wrote:
> 2010/4/17 Denys Vlasenko <vda.linux at googlemail.com>:
> >>  Config.in: add target platform selection
> >>  Config.in: add target platform WIN32
> >
> > This looks like a correct approach for this.
> > Let's choose a good name for it. TARGET
> > often assumes architecture, maybe we should call it
> > TARGET_PLATFORM? Regarding WIN32 part, since there
> 
> Or even shorter, PLATFORM (i.e. PLATFORM_POSIX)?
> 
> > are Cygwin-like POSIX-like emulation layers, WIN32
> > might be  ambiguous. Maybe NATIVE_WIN[DOWS]?
> > OTOH, TARGET_PLATFORM_NATIVE_WIN is a bit long...
> > but for now, it's the name which looks best to me.
> 
> Maybe I should change it back to PLATFORM_MINGW32. It's clear. And if
> someone ports this to MS Visual Studio, we can deal with it later.

I still think NATIVE_WIN is mpore future-proof.
I want to forestall an explosion of various WIN platforms.
NATIVE_WIN can be meant as "bare Windows platform, no
posix compal layer provided by toolchain".

> >>  libbb: update messages.c to support Windows
> >>  win32: Import fnmatch source
> >
> > Starting from this, you have mostly things which definitely should not
> > go in yet.
> > First batch of patches should ONLY enable building on WIN,
> > even if the result can build only "true", "false" and "basename" applets.
> > (Not that it IS that bad, I think you will have more than that).
> > Let's merge things up to this intermediate point,
> > and then continue.
> 
> That was my intention (although this series has two parts: one to
> build on win, one to enable archival/, just to demonstrate how it
> changes other applets).
> 
> In order to build, it needs lots of stubs functions.

allnoconfig needs _lots_ of stub functions?! That's suprising.

> I could have made 
> true (empty) stubs for the first part (building on win), then replace
> stubs with implementation later. But I was lazy so I put impl in place
> where they should have been only stubs.
> 
> >>  win32: set binary I/O mode by default
> >>  win32: add sleep()
> >>  win32: add mkstemp()
> >>  win32: add gettimeofday()
> >>  win32: add pipe()
> >>  win32: add gmtime_r()
> >>  win32: add localtime_r()
> >>  win32: add getpwuid()
> >>  win32: add signal routines and SIGALRM support

allnoconfig needs SIGALRM??

> >>  win32: add function to map windows errors to posix ones
> >>  win32: add link()
> >>  win32: add strsep()
> >>  win32: add realpath()
> >>  win32: add get_busybox_exec_path(), which is bb_busybox_exec_path
> >>  win32: add mkdir()
> >>  win32: add waitpid()
> >>  win32: add fcntl()
> >>  win32: add poll()
> >>  win32: add getenv(), setenv(), unsetenv() and clearenv()
> >>  Makefile: support building executable with extension
> >>  Makefile: support building on Windows using MinGW compiler
> >>  Add README.win32
> >
> > After first part will be merged, next patchset should add a small set
> > of compat glue, and WIN-enable a few applets which are becoming buildable
> > because of that glue. This way, the patchsets will be easier to understand
> > and review.
> 
> OK.

-- 
vda