[PATCH 1/4] winedbg: Set thread->name from MSVC exception when in GDB proxy mode.

[Eric Pouech]

Le 16/03/2022 à 21:13, Brendan Shanks a écrit :
>
>> On Mar 16, 2022, at 12:20 AM, Eric Pouech <eric.pouech at orange.fr> wrote:
>>
>> Le 15/03/2022 à 20:05, Brendan Shanks a écrit :
>>> Signed-off-by: Brendan Shanks <bshanks at codeweavers.com>
>>> ---
>>>  programs/winedbg/gdbproxy.c | 9 +++++----
>>>  1 file changed, 5 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/programs/winedbg/gdbproxy.c b/programs/winedbg/gdbproxy.c
>>> index 0268a288481..bcde120adeb 100644
>>> --- a/programs/winedbg/gdbproxy.c
>>> +++ b/programs/winedbg/gdbproxy.c
>>> @@ -482,7 +482,6 @@ static BOOL handle_exception(struct gdb_context* 
>>> gdbctx, EXCEPTION_DEBUG_INFO* e
>>>      {
>>>          const THREADNAME_INFO *threadname = (const THREADNAME_INFO 
>>> *)rec->ExceptionInformation;
>>>          struct dbg_thread *thread;
>>> -        char name[9];
>>>          SIZE_T read;
>>>            if (threadname->dwType != 0x1000)
>>> @@ -494,10 +493,12 @@ static BOOL handle_exception(struct 
>>> gdb_context* gdbctx, EXCEPTION_DEBUG_INFO* e
>>>          if (thread)
>>>          {
>>>              if (gdbctx->process->process_io->read( 
>>> gdbctx->process->handle,
>>> -                threadname->szName, name, sizeof(name), &read) && 
>>> read == sizeof(name))
>>> +                threadname->szName, thread->name, 
>>> sizeof(thread->name), &read) &&
>>> +                read == sizeof(thread->name))
>>
>> it looks a bit strange to me that we can always expect being able to 
>> read sizeof(thread->name) here...
>>
>> using existing memory_get_string helper might be a better idea
>
> Thanks, I hadn’t seen that function before, I’ll use it. I think the 
> end result will be the same though, since memory_get_string() uses the 
> same read() that's implemented with ReadProcessMemory(), which doesn’t 
> do partial reads.
> ReadProcessMemory() could fail if sizeof(thread->name) would overflow 
> into an inaccessible page, but for a convenience feature like thread 
> names (and this is the old/deprecated way of setting them) I’m not 
> sure it’s worth handling that rare case.

sounds reasonable (we could evolve memory_get_string if that ever 
happens for any kind of string)

A+
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.winehq.org/pipermail/wine-devel/attachments/20220317/b36e6029/attachment.htm>