On 2024/03/19 11:55, Tao Liu wrote:
Hi Alexey,
On Tue, Mar 19, 2024 at 6:00 AM Alexey Makhalov
<alexey.makhalov(a)broadcom.com> wrote:
>
> Hi Tao,
>
> I like the idea of having just one thread at a time from the gdb side, and it can be
replaced by a "set" crash command.
> Single semantics will work for both live and offline debugging and for both types of
tasks: active or scheduled out.
> Need to be careful with lifetime for the data (gdb caches). I'm experimenting
with live debugging these days.
> Aditya, Tao, let me know what is the latest patchset I should use from you folks?
You can use
https://github.com/liutgnu/crash-dev/commits/one-thread
directly. The one-thread branch takes Aditya's v10 patchset and my v1
patchset, as well as the trial "one gdb thread/CPUS" patch. You can
revert the trial one patch for testing.
I also agree to the idea of one gdb thread, it seems nice and simple!
I'm looking at your patch set, is it going to be cleaned up with
Aditya's patch set somehow? especially it looks like your patch removes
the gdb patches added by Aditya's patch set again, that can be reduced
and we have to avoid "modifying gdb patch" [1]. also it will be better
to change gdb side first, to simplify the following patches.
I think you're going to do this later though, just in case.
[1]
https://github.com/crash-utility/crash/wiki#writing-patches
Thanks,
Kazu