[HEADSUP] Phasing out old Windows versions and 32 bit support

[Thomas Wolff]

Hi Peter,

Am 27.10.2021 um 18:46 schrieb Peter A. Castro:
> On Wed, Oct 27, 2021 at 11:37:26AM +0200, Thomas Wolff wrote:
>
> Greetings, Thomas,
>
>> Am 27.10.2021 um 10:49 schrieb Corinna Vinschen via Cygwin:
>>> On Oct 27 09:24, Takashi Yano via Cygwin wrote:
>>>> On Tue, 26 Oct 2021 22:55:01 +0200
>>>> Corinna Vinschen wrote:
>>>>> We're also planning to drop Support for the 32 bit release of Cygwin in
>>>>> 2022, thus Cygwin 3.4.0 won't come in 32 bit anymore, and the package
>>>>> maintainers won't have to update 32 bit packages anymore.  If you're
>>>>> still running Cygwin under WOW64, consider to move to 64 bit in the next
>>>>> couple of months.
>>>> I agree with you that 32 bit cygwin under WOW64 is not worth to
>>>> support any more. However, 32 bit version of Windows 10 will be
>>>> still supported at least until Oct. 2025. Personally, I think it
>>>> would not be nice to exclude the supported windows version from
>>>> cygwin support.
>>> Well, it's not much effort to support WOW64 if we support 32 bit anyway.
>>> The problem is that Cygwin is somehow outgrowing 32 bit systems in terms
>>> of the available memory.  Also, 32 bit Cygwin is still using a 32 bit
>>> time_t, https://en.wikipedia.org/wiki/Year_2038_problem
>>>
>>> Per the download statistics, as far as those statistics are trustable,
>>> 32 bit systems are less than 5% of the installed base, with the majority
>>> of them being WOW64 installations.  Those can move over to 64 bit Cygwin
>>> easily.
>>>
>>> Less than 1% are real 32 bit systems.
>> I think roughly 1% is still a community to consider. Working old machines
>> shouldn't be trashed just because they are missing a few bits :)
>>
>>> Dropping 32 bit support will reduce code complexity in Cygwin and it will
>>> reduce the workload of the package maintainers.
>> Code complexity was also an argument when dropping XP support, but there was
>> quite some discussion at its time.
>> For `egrep "# *if.*(32|64)"` I'm counting roughly 160 matches in winsup, but
>> only in a few files. Is it really necessary?
>>
>>> Those few still running
>>> Cygwin on a real 32 bit system will still have a chance to run Cygwin
>>> by utilizing Peter's time machine.
>> Peter's time machine is a very appreciable effort. It's a bit fiddly though to figure out how to use it, particularly to identify the "latest XP version". Maybe some explicit howto could be published on the cygwin pages?
> Could you please give an example of the "fiddly" bit?  I list the URLs
> to use with the install and it's clearly labeled "The last version of
> Cygwin that supports XP is 2.5.2-1".  Or were you, perhaps, refering to
> the actual usage of the URL in the Setup program?
On http://www.crouchingtigerhiddenfruitbat.org/Cygwin/timemachine.html, 
I see a setup link and a repository URL.
If I run that setup and enter the URL on the mirror selection page, I 
get an error popup
---------------------------
Cygwin Setup
---------------------------
Can't open /software/windows/cygwin32/x86_64/setup.xz.sig for reading: 
No such file or directory
---------------------------
OK
---------------------------
So I should have deselected the preselected repository explicitly. If I 
fix that or click the popup off a few times, there's another popup
---------------------------
Cygwin Setup
---------------------------
Unable to get 
http://ctm.crouchingtigerhiddenfruitbat.org/pub/cygwin/circa/64bit/2016/08/30/104235/x86_64/setup.xz.sig 
from 
<http://ctm.crouchingtigerhiddenfruitbat.org/pub/cygwin/circa/64bit/2016/08/30/104235/>
---------------------------
OK
---------------------------
repeatedly.
The command line you mentioned in the other mail does not seem to work 
either. The 32-bit version works (from the command line only). Also 
setup suggests my existing cygwin installation as its installation 
target which needs to be fixed carefully to avoid destruction...

> Another user, Michel, responded that perhaps a more explicit message
> with exact steps for install this might be helpful (as the "Dead Simple
> Instructions" are generic), but I'm not sure it's really necessary.  Is
> that, perhaps, what you are refering to in that the instructions aren't
> explicit enough?
An explicit quote of a safely working command line invocation would 
certainly help.
Best greetings
Thomas