[AD] plugins.h on MinGW32 2

["Matthew Leverton" <matthew@xxxxxxxxxx>]

When using the new version 2 of MinGW32 (gcc 3.2), the make process dies on
the plugins.h file, due to a garbage character at the end. Looks like a
stray CR/LF, and you can see it in notepad.

The following patch of "makefile.mgw" on line 553 fixes it (at least on
Win2000):

--
   define GENERATE_PLUGINS_H
      copy /B tools\plugins\*.inc obj\mingw32\plugins.h
   endef
--

It seems that by not specifying the /B (binary) switch, extra characters are
getting added? I'm pretty sure that all versions of `copy' ranging from DOS
to WinXP have the /B switch. Also, any other makefiles (DJGPP / MSVC / BCC
?) that use the the `copy *.inc >' syntax should be updated as well.

PS: I'm using the 4.1.1 version, I didn't check to see if the latest version
via CVS is fixed or not...

--
Matthew Leverton - matthew@xxxxxxxxxx
http://www.allegro.cc

----- Original Message -----
From: "Bob" <ohannessian@xxxxxxxxxx>
To: "Sven Sandberg" <sven.sandberg.1977@xxxxxxxxxx>
Cc: "Allegro Developer" <conductors@xxxxxxxxxx>
Sent: Wednesday, September 11, 2002 7:32 AM
Subject: Re: [AD] Some questions


> Sven Sandberg wrote:
>
> >
> > There is probably a very small overhead with bmp_write_line(). But
> > note that you only need to call bmp_write_line() once per scanline. So
> > if you are drawing one scanline at a time then the overhead is
> > negligible. AFAIK Allegro never makes a special case for pure memory
> > bitmaps in order to use line[] instead of bmp_write_line().
>
> I tried making a special case for memory bitmaps in FBlend, by jumping
> accross the call to bmp_write_line and reading line[] directly (this
> should be an easily predictable jump, so the overhead is quite small).
> However, I have not been able to measure any speed difference in between
> that code and simply calling bmp_write_line.
>
> I think Allegro would benefit more from a region locking mechanism, so
> that the calls only need to made once "per region" instead of once per
> scan line. Of course, you can always use a region of height 1 to
> simulate the old behavior if needed.
> I've put up something like that for the Allegro 5 API, but I think 4.2
> would benefit from having it too.
>