X-Authentication-Warning:	delorie.com: mail set sender to djgpp-bounces using -f
X-Recipient:	djgpp AT delorie DOT com
Subject:	Re: Death of DJGPP support on Win10 32 bit? [WAS: Re: Max value of
	DpmiLimit registry setting in Windows 10 32 bit]
To:	djgpp AT delorie DOT com
References:	<410d8515-84ba-c47e-1f90-f60a944d8668 AT iki DOT fi>
	<1477a7ba-2388-2fa5-9de9-70a261cc4a6c AT iki DOT fi>
	<910ca294-6f13-fa68-a5ed-01e22d549f6c AT iki DOT fi>
From:	"Andris Pavenis (andris DOT pavenis AT iki DOT fi) [via djgpp AT delorie DOT com]" <djgpp AT delorie DOT com>
Message-ID:	<77600078-c9a6-dfbd-7a75-0ea30e828bf0@iki.fi>
Date:	Tue, 15 May 2018 22:09:29 +0300
User-Agent:	Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
	Thunderbird/52.7.0
MIME-Version:	1.0
In-Reply-To:	<910ca294-6f13-fa68-a5ed-01e22d549f6c@iki.fi>
X-MIME-Autoconverted:	from Quoted-Printable to 8bit by delorie.com id w4FJ9a0f019637
Reply-To:	djgpp AT delorie DOT com
Errors-To:	nobody AT delorie DOT com
X-Mailing-List:	djgpp AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

On 04/14/2018 05:36 PM, Andris Pavenis (andris DOT pavenis AT iki DOT fi) [via djgpp AT delorie DOT com] wrote:
> On 03/27/2018 07:37 PM, Andris Pavenis (andris DOT pavenis AT iki DOT fi) [via djgpp AT delorie DOT com] wrote:
>> On 03/21/2018 07:55 PM, Andris Pavenis (andris DOT pavenis AT iki DOT fi) [via djgpp AT delorie DOT com] wrote:
>>> Some experience with DJGPP under Windows 10 Home 32 bit lately (VirtualBox VM under ArchLinux)
>>>
>>> Attempt to run DJGPP program from another one failed:
>>>
>>> C:\Users\andris>sh
>>> sh-2.05b$ sh
>>> Exiting due to signal SIGSEGV
>>> Stack Fault at eip=00000a11
>>> eax=00010001 ebx=000f0080 ecx=0009f200 edx=000007bf esi=00000c19 edi=00000c09
>>> ebp=0026773a esp=0026773a program=C:\DJGPP\BIN\SH.EXE
>>> cs: sel=00cf  base=000d3a40  limit=00000db0
>>> ds: sel=00b7  base=000cc150  limit=0000ffff
>>> es: sel=0040  invalid
>>> fs: sel=0000
>>> gs: sel=0000
>>> ss: sel=00b7  base=000cc150  limit=0000ffff
>>> App stack: [00263aa0..000e3aa4]  Exceptn stack: [000e3190..000e1250]
>>>
>>> It  worked (earlier setup) in the begin of February 2018 as I built gcc-8.0.1 development 
>>> version then.
>>>
>>> I got however it working by decreasing 
>>> HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\WOW\DpmiLimit from 0x7FFF0000 to 0x7EFF0000 (no 
>>> restart needed after change). No attempts done to find more exact value.
>>>
>>> The change of max value of DpmiLimit is most likely caused by 2018-02 cumulative update 
>>> (noticed problem before 2018-03 cumulative update were installed).
>>>
>>> Let's see how build of gcc-8.0.1-20180318 will go (it at least do not fail immediately as with 
>>> DpmiLimit=0x7FFF0000)
>> It built OK then.
>>
>> Unfortunately all this looks like end of DJGPP support under Win10 32 bit.
>> 1) problem went away when  DpmiLimit was decreased as mentioned above and 2018-03 cumulative 
>> update (KB4088776) uninstalled
>> 2) yesterday evening left building gcc-8.0.1 20180325. M$ voluntarily reinstalled KB4088776 
>> later before build ended and as result I'm getting this failure always for any value of 
>> DpmiLimit (also when it is not specified in registry)
>> 3) un-installing KB4088776 again fixes the problem
>>
>> So do not see any way to keep DJGPP support rather than trying to prevent Win10 from accessing 
>> Windows update (local network only like host only networking in VirtualBox or specifying static 
>> IP address and no gateway)
>>
>> I am not optimistic that submitting error report to Microsoft will do any good.
>
> Tried with insider builds of Windows 10 32 bit (does not harm as that Win10 32 bit VM is used for 
> DJGPP builds only). No improvement even when both slow and fast ring was tried. Switching from 
> slow to fast ring required Google searches and messing with windows registry...
>
> Submitted feedback to M$. I'm not very optimistic that we'll get anything from it though
>
> I'm also afraid that some M$ update may break DJGPP support for WIn7 32 bit (I hope that it 
> works. I have not tested it myself)
>
>
Current fast ring insider build (calls itself as OS build 17134.48) shows very slight but 
insufficient improvement:
- test case I provided in feedback no more crashes
- trying to do something real (like running some script) causes same error to happen randomly but 
often enough to prevent any useful use

I have not tested with any other builds.

I guess I should make a new failing test case and feed it back to Microsoft.

Andris

- Raw text -