tracepoint.c memory range requests

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

tracepoint.c memory range requests

jimb (Bugzilla)
Looking at the code in tracepoint.c:stringify_collection_list, it
seems that memory ranges with absolute addresses end up having the
form MFFFFFFFF,<start>,<len>.  This is because "%X" always treats its
argument as unsigned.  So the number of F's depends on sizeof
(unsigned int).

I know that we spent some effort keeping the packets small.  Since
there are, as far as we know, no stubs that support tracepoints in
active use, may I change that to send '-' instead of the F's?
Reply | Threaded
Open this post in threaded view
|

Re: tracepoint.c memory range requests

Michael Snyder
Jim Blandy wrote:
> Looking at the code in tracepoint.c:stringify_collection_list, it
> seems that memory ranges with absolute addresses end up having the
> form MFFFFFFFF,<start>,<len>.  This is because "%X" always treats its
> argument as unsigned.  So the number of F's depends on sizeof
> (unsigned int).
>
> I know that we spent some effort keeping the packets small.  Since
> there are, as far as we know, no stubs that support tracepoints in
> active use, may I change that to send '-' instead of the F's?

Actually, there may be such a stub -- there's been at least
one guy posting questions about target tracepoint implementation
this year.

Is the purpose just to make the packet smaller?

The intent was for that number to be the value negative one,
so changing it to "M-1,<start>,<len>" wouldn't be entirely
unreasonable.  The comments in gdb refer to it as "-1",
(when they don't mistakenly refer to it as "0").

Reply | Threaded
Open this post in threaded view
|

Re: tracepoint.c memory range requests

jimb (Bugzilla)
On 11/18/05, Michael Snyder <[hidden email]> wrote:
> Actually, there may be such a stub -- there's been at least
> one guy posting questions about target tracepoint implementation
> this year.
>
> Is the purpose just to make the packet smaller?

Yeah.

> The intent was for that number to be the value negative one,
> so changing it to "M-1,<start>,<len>" wouldn't be entirely
> unreasonable.  The comments in gdb refer to it as "-1",
> (when they don't mistakenly refer to it as "0").

Okay.  I'll make the code behave that way, and document it that way.