Nut upsdev - Dec 2010 - windows_port NUT branch

Hi,

Frederic, big thanks for your work on this, by a coincedence it's exactly
what
i need at work :)

I'm compiling your branch on Debian GNU/Linux with mingw and having some
minor
troubles worth mentioning and (i hope) fixing.

I configure with
CC=586-mingw32msvc-gcc ./configure --host=i586-mingw32msvc --prefix=/c/winnut/

Interestingly, there's scripts/Windows/Makefile in the repo but it gets
overwritten by the configure process.

Other minor tweaks i had to do (probably not the most correct way, but i know
just about nothing wrt windows programming):

On jeu., 2010-12-16 at 12:55 +0300, Paul Fertser wrote:
> Hi,
> 
> Frederic, big thanks for your work on this, by a coincedence it's
exactly what
> i need at work :)
> I'm compiling your branch on Debian GNU/Linux with mingw and having
some minor
> troubles worth mentioning and (i hope) fixing.
> 
> I configure with
> CC=586-mingw32msvc-gcc ./configure --host=i586-mingw32msvc
--prefix=/c/winnut/
> 
> Interestingly, there's scripts/Windows/Makefile in the repo but it gets
> overwritten by the configure process.
The Makefile in the repo is currently not used. It won't prevent nut to
work but needs to be addressed.
> 
> Other minor tweaks i had to do (probably not the most correct way, but i
know
> just about nothing wrt windows programming):
> 
> From 26ff6d7fb1ffc6421f1f36243ab9166b8544a9ab Mon Sep 17 00:00:00 2001
> From: Paul Fertser <fercerpav at gmail.com>
> Date: Thu, 16 Dec 2010 12:48:07 +0300
> Subject: [PATCH] mingw32 cross-compilation tweaks
> 
> This adds defines for usleep (windows lacks that) and also missing
> define for SVC_EVENT.
I have no problem compiling with usleep here. Can you give me the error
string you get on this ?
As a side note, beware that your implementation of usleep could lead to
call Sleep(0) (immediate return) if n<1000. usleep is supposed to round
up the time spent depending on system activity and timers granularity.

Concerning SVC_EVENT, it is declared in winevent.h which is a generated
file from winevent.mc as well as a DLL required for the Windows Event
log. Here is how it works:

windmc winevent.mc
windres winevent.rc winevent.o
dllwrap --output-lib=libwinevent.a --dllname=winevent.dll
--driver-name=gcc winevent.o

Note that this should be automated soon in the Makefile.
> Thanks again and happy hacking :)
> 
Thanks for your interest and do not hesitate to send your feedback on
working or non working drivers.

Fred


--------------------------------------------------------------------------

On Dec 16, 2010, at 7:16 AM, Fr?d?ric Boh? wrote:
> As a side note, beware that your implementation of usleep could lead  
> to
> call Sleep(0) (immediate return) if n<1000. usleep is supposed to  
> round
> up the time spent depending on system activity and timers granularity.
Related: http://trac.networkupstools.org/projects/nut/changeset/2729

To fix the rounding, I would recommend adding a constant to the  
numerator:

timeout = (d_sec*1000) + (d_usec+999)/1000;

versus:

timeout = (d_sec*1000) + (d_usec/1000) + (((d_usec % 1000)!=0)?1:0);

The same could be applied to the usleep -> Sleep function. It saves an  
extra call to the mod function.

FWIW, I have mingw-w32api 3.3 here, and it does not have usleep in the  
headers (only Sleep).

On jeu., 2010-12-16 at 09:07 -0500, Charles Lepple
wrote:
> On Dec 16, 2010, at 7:16 AM, Fr?d?ric Boh? wrote:
> 
> > As a side note, beware that your implementation of usleep could lead  
> > to
> > call Sleep(0) (immediate return) if n<1000. usleep is supposed to  
> > round
> > up the time spent depending on system activity and timers granularity.
> 
> Related: http://trac.networkupstools.org/projects/nut/changeset/2729
> 
> To fix the rounding, I would recommend adding a constant to the  
> numerator:
> 
> timeout = (d_sec*1000) + (d_usec+999)/1000;
> 
> versus:
> 
> timeout = (d_sec*1000) + (d_usec/1000) + (((d_usec % 1000)!=0)?1:0);
> 
> The same could be applied to the usleep -> Sleep function. It saves an  
> extra call to the mod function.
good point, changed. Thanks.



--------------------------------------------------------------------------