[sane-devel] sane runs in windows without cygwin

[m. allan noah]

On Wed, Jan 14, 2009 at 10:45 AM, Ilia Sotnikov <hostcc at gmail.com> wrote:
> On Wed, Jan 14, 2009 at 4:25 PM, m. allan noah <kitno455 at gmail.com> wrote:
>> 2009/1/13 Michael Cronenworth <mike at cchtml.com>:
>>> Hello,
>>>
>>> I've had success in building sane with MinGW, MSYS, and libusb-win32-0.12.
>>> During my initial work I was unable to find much of anything about such a
>>> feat as most people settled for Cygwin.
>>
>> Excellent news!
>>
>>> Some caveats:
>>> 1. I ripped out all backends except for "fujitsu" as this is the scanner I
>>> have. The "fujitsu" code required no modifications to compile.
>>
>> Did you try any others, or start with only fujitsu?
>>
>>> 2. Signals, threading, and scsi are also obviously not available. I put
>>> #ifdef 32 or #ifndef WIN32 around the problem areas.
>>
>> sane-fujitsu does not use signals or threading, so you made a good
>> choice of backend :)
>>
>>> 3. I changed sleep() calls to usleep() calls or used the Windows method of
>>> sleep.
>>
>> I only see usleep calls in sane, but there is alot of code...
>>
>>> 4. getpass() is used?!?
>>
>> By scanimage, yes. SANE includes a rudimentary form of authentication
>> checking, in case a resource requires more security. I don't know of
>> anything that uses it.
>>
>>> In retrospect:
>>> 1. The rest of the backends seemed to need extensive rework as they use
>>> signals and ioctl calls that do not translate at all.
>>
>> There are a few others that dont use threading or signals (canon_dr,
>> cardscan, ...), it would be nice to see if those compile as well.
>>
>>> 2. Threading and scsi could be made available quite easily by adding Windows
>>> equivalents.
>>
>> Well, we would need a windows expert to do it :)
>>
>>> 3. Platform independent sleep calls should be put in.
>>> 4. Obsolete function! This should be replaced with similar functionality.
>>
>> Certainly these two are easy fixes.
>>
>>> I have a GTK+2 app that used sane in Linux and it works in Windows without
>>> any modification to the app code. I scanned successfully and my program
>>> rendered a beautiful PNG to the screen. It should be noted that in order for
>>> the scanner to be detected, I had to install the Fujitsu TWAIN driver. I
>>> guess libusb-win32 could not detect the device until the appropriate driver
>>> was installed? Artefact of using Windows inside of VirtualBox?
>>
>> That is odd. I don't know how windows works in that regard. Makes me
>> wonder if we could produce some sort of prog that would read the sane
>> .desc files and build a .inf or something to tell windows to talk to
>> the device?
>>
>>> If anyone is interested in the changes I've made, I'd be interested in
>>> making them available.
>>
>> yes, very much interested. there has been some talk lately about
>> updating the sane build system, it might be possible to coordinate
>> with that effort, so that at least minimal win32 support is added to
>> mainline sane.
>>
>
> Well, what about experimental/sane-backends-win32 sitting in CVS for
> about 18 months? To be honest, we are using that version of SANE
> backends (dll + net) in native Win32 form for more than 2 years. The
> version comes even with Inno setup installer support ;) Anyone
> interested?

Yes- I had forgotten about that. So now we have dll, net, and fujitsu
backends, and scanimage natively building?

allan
-- 
"The truth is an offense, but not a sin"