GNU Emacs 28.0.50 crash (emacs-w32)

GNU Emacs 28.0.50 crash (emacs-w32)

[Jim Reisert AD1C]

[-- Attachment #1: Type: text/plain, Size: 372 bytes --]

The test release of GNU Emacs 28.0.50 crashed on me when using
emacs-w32.  I can repeat this every time.  I attached the trace and
stackdump.

Windows 11 Pro 64-bit
Take Command 27.01.24 x64
Opened a text file using emacs-w32
Ctrl-X Ctrl-C to exit Exit Emacs
I did not make any changes to the text file

-- 
Jim Reisert AD1C, <jjreisert@alum.mit.edu>, https://www.ad1c.us

[-- Attachment #2: emacs_backtrace.txt --]
[-- Type: text/plain, Size: 432 bytes --]


Backtrace:
0x100608b80
0x1005fe058
0x1004bf5ea
0x1006001ca
0x10060029b
0x100602776
0x10060278a
0x1800640dd
0x18018f508
0x8020f6600

Backtrace:
0x100608b80
0x1005fe058
0x1004bf5ea
0x1006001ca
0x10060029b
0x100602776
0x10060278a
0x1800640dd
0x18018f508
0x10060a258

Backtrace:
0x100608b80
0x1005fe058
0x1004bf5ea
0x1006001ca
0x10060029b
0x100602776
0x10060278a
0x1800640dd
0x18018f508
0x10060a258

[-- Attachment #3: emacs-w32.exe.stackdump --]
[-- Type: application/octet-stream, Size: 1884 bytes --]

Stack trace:
Frame        Function    Args
000FFFF7B40  001800620A7 (000FFFF7D48, 00010000002, 00000000002, 000FFFFDE50)
00000000000  001800640E5 (000FFFF8390, 00000000000, 000000004E0, 00000000000)
000FFFF8250  0018012FE0B (0018012B2B3, 00000000000, 00000000000, 00000000000)
00000000000  00180157AD9 (00000000000, 00000000000, 00000000000, 00000000020)
00000000000  0018018F3EB (00000000000, 00000000000, 00000000000, 00000000020)
00000000000  001006002E5 (00000000002, D1365D5C0000, D1365D5C41FE, 00180157C70)
00000000000  0010060277E (00000000001, 00000000000, 00000000000, 00010000000)
00000000000  001004D8A84 (00000000000, 000FFFF8610, 00010000000, 000FFFFCE00)
00000000000  001004D8BB9 (000FFFF8668, 000FFFFCE00, 00180064300, 00000000000)
00000000000  001800640E5 (000FFFF8D80, 00000000000, 000000004E0, 00000000000)
000FFFF8C40  0018012FE0B (0018018F435, 00000000000, 001801302FB, 00000000000)
00000000000  00180157AD9 (00000000006, 000FFFF8EA0, 001009995B3, 0018018F435)
00000000000  001802130D8 (000FFFF909C, 0000000000D, 00000000000, 0010060A258)
00000000000  00100608B09 (00000000000, 00000000028, 00000000000, 00000000000)
00000000000  001005FE060 (0010051EAF8, 6FFFFF3DEC18, 0010051E778, 00000000053)
00000000000  001004BF5F2 (00000000053, 00000000053, 001803241A0, 001803241A0)
00000000000  001006001D2 (0010051F74E, 00000000004, 00800FDE920, 00100A30440)
00000000000  001006002A3 (00800FE6450, 00100A30440, 00100A291C0, 00180157C70)
00000000000  0010060277E (0010051F74E, 00000000004, 00800FE5B98, 00100A30440)
00000000000  00100602792 (00100A291C0, 00000000000, 00100A310A0, 00800FE5B98)
00000000000  001800640E5 (00100A291C0, 00000000000, 00001002015, 000615DB00B)
000FFFFAB30  0018018F510 (00100A291C0, 00000000000, 00001002015, 000615DB00B)
000FFFFAB30  0010051E6EF (00100A291C0, 00000000000, 00001002015, 000615DB00B)
End of stack trace

Re: GNU Emacs 28.0.50 crash (emacs-w32)

[Jim Reisert AD1C]

> The test release of GNU Emacs 28.0.50 crashed on me when using
> emacs-w32.  I can repeat this every time.  I attached the trace and
> stackdump.

Emacs in an X window with this same file did NOT crash.  It seems to
be something particular to emacs-w32 or a non-X environment.

-- 
Jim Reisert AD1C, <jjreisert@alum.mit.edu>, https://www.ad1c.us

Re: GNU Emacs 28.0.50 crash (emacs-w32)

[Ken Brown]

On 10/6/2021 10:22 AM, Jim Reisert AD1C wrote:
> The test release of GNU Emacs 28.0.50 crashed on me when using
> emacs-w32.  I can repeat this every time.  I attached the trace and
> stackdump.
> 
> Windows 11 Pro 64-bit
> Take Command 27.01.24 x64
> Opened a text file using emacs-w32
> Ctrl-X Ctrl-C to exit Exit Emacs
> I did not make any changes to the text file

Could this be the following problem that I mentioned in the release announcement?

> Compilation is done asynchronously, with a log in a buffer called
> *Async-native-compile-log*.  If you run emacs-w32 and exit while a
> compilation is in progress, you might see a dialog box saying that
> emacs has aborted and asking if you want to attach a debugger.  Just
> say No.  If this annoys you, check the compilation buffer before
> exiting, and wait for the "Compilation finished" message.

I'll add one more thing that I didn't know when I wrote that: You can customize 
the variable native-comp-async-query-on-exit if you want to be warned that there 
are compilations in process when you exit.

Ken

Re: GNU Emacs 28.0.50 crash (emacs-w32)

[Ken Brown]

On 10/6/2021 11:38 AM, Ken Brown via Cygwin wrote:
> On 10/6/2021 10:22 AM, Jim Reisert AD1C wrote:
>> The test release of GNU Emacs 28.0.50 crashed on me when using
>> emacs-w32.  I can repeat this every time.  I attached the trace and
>> stackdump.
>>
>> Windows 11 Pro 64-bit
>> Take Command 27.01.24 x64
>> Opened a text file using emacs-w32
>> Ctrl-X Ctrl-C to exit Exit Emacs
>> I did not make any changes to the text file
> 
> Could this be the following problem that I mentioned in the release announcement?
> 
>> Compilation is done asynchronously, with a log in a buffer called
>> *Async-native-compile-log*.  If you run emacs-w32 and exit while a
>> compilation is in progress, you might see a dialog box saying that
>> emacs has aborted and asking if you want to attach a debugger.  Just
>> say No.  If this annoys you, check the compilation buffer before
>> exiting, and wait for the "Compilation finished" message.
> 
> I'll add one more thing that I didn't know when I wrote that: You can customize 
> the variable native-comp-async-query-on-exit if you want to be warned that there 
> are compilations in process when you exit.

Having said that, I'll still try to track down the cause of the crash.

Ken

Re: GNU Emacs 28.0.50 crash (emacs-w32)

[Jim Reisert AD1C]

> Could this be the following problem that I mentioned in the release announcement?
>
> > Compilation is done asynchronously, with a log in a buffer called
> > *Async-native-compile-log*.  If you run emacs-w32 and exit while a
> > compilation is in progress, you might see a dialog box saying that
> > emacs has aborted and asking if you want to attach a debugger.  Just
> > say No.  If this annoys you, check the compilation buffer before
> > exiting, and wait for the "Compilation finished" message.
>
> I'll add one more thing that I didn't know when I wrote that: You can customize
> the variable native-comp-async-query-on-exit if you want to be warned that there
> are compilations in process when you exit.

Sure, that could be the problem.  I'll try the workaround.

I do notice that there are separate compilation directories for emacs
and emacs-w32.  Is this necessary?

[JJR:/cygdrive/e/] $ new ~/.emacs.d/eln-cache/
total 28
drwxrwx---+ 1 jjrei jjrei 0 Oct  3 15:44 28.0.50-eb740d6f
drwxrwx---+ 1 jjrei jjrei 0 Oct  6 08:17 28.0.50-891fd08c

-- 
Jim Reisert AD1C, <jjreisert@alum.mit.edu>, https://www.ad1c.us

Re: GNU Emacs 28.0.50 crash (emacs-w32)

[Ken Brown]

On 10/7/2021 10:16 AM, Jim Reisert AD1C wrote:
>> Could this be the following problem that I mentioned in the release announcement?
>>
>>> Compilation is done asynchronously, with a log in a buffer called
>>> *Async-native-compile-log*.  If you run emacs-w32 and exit while a
>>> compilation is in progress, you might see a dialog box saying that
>>> emacs has aborted and asking if you want to attach a debugger.  Just
>>> say No.  If this annoys you, check the compilation buffer before
>>> exiting, and wait for the "Compilation finished" message.
>>
>> I'll add one more thing that I didn't know when I wrote that: You can customize
>> the variable native-comp-async-query-on-exit if you want to be warned that there
>> are compilations in process when you exit.
> 
> Sure, that could be the problem.  I'll try the workaround.

Thanks.  Please report back.

I'm about to announce a new test release that tries to work around the problem 
in a different way.  I'd be interested in hearing how that works too.

> I do notice that there are separate compilation directories for emacs
> and emacs-w32.  Is this necessary?

Yes.  Each build environment for emacs requires its own compiled libraries. 
Here "build environment" includes source files, system, and configuration 
options.  In this case, the build of emacs-w32 used the configuration option 
"--with-w32" that wasn't used in the build of emacs-X11.

Thanks very much for testing!

Ken

Re: GNU Emacs 28.0.50 crash (emacs-w32)

[Jim Reisert AD1C]

> >> I'll add one more thing that I didn't know when I wrote that: You can customize
> >> the variable native-comp-async-query-on-exit if you want to be warned that there
> >> are compilations in process when you exit.
> >
> > Sure, that could be the problem.  I'll try the workaround.
>
> Thanks.  Please report back.

Yes, it now warns that active processes are running.

I originally typo'd the variable (see below, I wasn't setting it to
true), and Emacs just silently exited.  That doesn't sound right.  But
this isn't your issue, it seems like an Emacs issue.

 '(version-control t)
 '(native-comp-async-query-on-exit))

-- 
Jim Reisert AD1C, <jjreisert@alum.mit.edu>, https://www.ad1c.us

Re: GNU Emacs 28.0.50 crash (emacs-w32)

[Jim Reisert AD1C]

> I'm about to announce a new test release that tries to work around the problem
> in a different way.  I'd be interested in hearing how that works too.

I updated Emacs and turned off "native-comp-async-query-on-exit".
This time, there was no error report from emacs-w32, it just silently
exited while the compilation was ongoing.  I don't know if that was
the desired outcome.

-- 
Jim Reisert AD1C, <jjreisert@alum.mit.edu>, https://www.ad1c.us

Re: GNU Emacs 28.0.50 crash (emacs-w32)

[Ken Brown]

On 10/7/2021 4:21 PM, Jim Reisert AD1C wrote:
>> I'm about to announce a new test release that tries to work around the problem
>> in a different way.  I'd be interested in hearing how that works too.
> 
> I updated Emacs and turned off "native-comp-async-query-on-exit".
> This time, there was no error report from emacs-w32, it just silently
> exited while the compilation was ongoing.  I don't know if that was
> the desired outcome.

Yes.  All I did was to suppress the error report.  It turns out that there was 
actually a Cygwin bug underlying this, which will be fixed in Cygwin 3.3.0:

   https://cygwin.com/pipermail/cygwin-developers/2021-October/012418.html

So I'll be able to remove the workaround once Cygwin 3.3.0 is released.

Ken