Windows 9x <-> Windows 11 gdbserver remote debugging, break/interrupt does not work

[emxd]

Hello,

Thank you for all the work on this distribution.

I was trying to make remote debugging work between a Windows 98 VM and Windows 11 and it mostly works except the host debugger (gdb) can't remotely stop the execution of the debugged program, on ctrl+c the debugger issues "The target is not responding to interrupt requests." This is a common problem for remote debuggers on Windows 9x, but I was hoping that maybe you'd have some insights on how to make it work. Windbg has the same problem and only very old Windbg versions can interactively break-in.

I used gdb+gdbserver from this release -> https://github.com/redpanda-cpp/mingw-lite/releases/download/13.4.0-r5/mingw32_686-msvcrt_win98-13.4.0-r5.7z
on a Windows 98 guest (Windows 98 + KernelEx) + Windows 11 host.

The host debugger connects, it's just that the interrupt signal does not break execution. I understand that this might be beyond the scope, only posting here just in case you had any ideas.

Edit: a small addition, this also concers normal gdb debugging on Win9x, not just remote. But gdb.exe, at least in the release build I linked contains a lot more imports, it would be difficult to make it run in Win9x. Which is why I opted for remote debugging.

[CyanoHao]

I cannot reproduce the error "The target is not responding to interrupt requests." But I encountered another error: gdbserver correctly forward Ctrl+C to the debugged program. Instead of getting paused, the program is terminated with exit code -1.

Test environment: Windows 7 host (Celeron 847E, VMware 7.1.6), Windows 98 SE guest (with VMware tools, no extra patches).

(The error I encountered is due to limitation of Windows 9x. Ctrl+C terminates debugged process, instead of generating a DBG_CONTROL_C event. I looked into gdb's git history but didn't find any clue -- it's using Ctrl+C event since the creation of repository in 1999.)

P.S. Do you have any idea which version of windbg is working? Maybe I can investigate how it works.

[emxd]

no extra patches

Interesting. If I run remote debugging without KernelEx (and therefore without system-wide Unicode via unicows.dll), I'm getting this error on connecting:

1.mp4

With KernelEx, I can connect successfully.

Do you have any idea which version of windbg is working? Maybe I can investigate how it works.

This is the version where ctrl+c works:

dbg.zip

You should also be able to download it from archive.org (https://web.archive.org/web/20020212054128if_/http://mssjus.www.conxion.com:80/download/platformsdk/windbg/Feb2000/NT45/EN-US/dbg.exe)

I only tried local debugging with it, the version is so old I'm not even sure how remote works.

I looked into gdb's git history but didn't find any clue

Yes, I also tried old gdb versions from mingw.org distribution, but ctrl+c break did not work.

[CyanoHao]

Interesting. If I run remote debugging without KernelEx (and therefore without system-wide Unicode via unicows.dll), I'm getting this error on connecting:

Workaround: set remotetimeout 5 (in gdb).

Thanks for your information about KernelEx. I compared my implementation of socket APIs with KernelEx's, and finally found the root cause of the remote delay.